Documente Academic
Documente Profesional
Documente Cultură
Amazonian Languages
W
DE
G
Handbook of
Amazonian Languages
Volume 4
edited by
Desmond C. Derbyshire and
Geoffrey K. Pullum
1998
Mouton de Gruyter
Berlin · New York
Mouton de Gruyter (formerly Mouton, The Hague)
is a Division of Walter de Gruyter GmbH & Co., Berlin.
© Printed on acid free paper which falls within the guideline of the ANSI to ensure permanence
and durability.
This fourth volume of the Handbook of Amazonian Languages follows three earlier
volumes published in 1986, 1990, and 1991 respectively. In addition to the linguistic
contents of this volume—two grammatical descriptions of a Cariban and an
Arawakan language, a typological study of a Panoan language, and a historical and
comparative exploration of Tupian languages, all summarized and discussed in our
editorial introduction—we include a cumulative index to all four volumes, subsum-
ing the index that was included in volume 3.
There is a sense in which it is an encouraging sign to us that the project continued
by this volume should be so far from having succeeded in achieving complete
coverage of its subject matter. The Handbook of Amazonian Languages has not
surveyed all the languages of the Amazon basin, nor even a fair percentage of them.
There is much more to be done than we have been able to do. The reason we see
that as a very good sign is that it stems from the simple fact that the indigenous
inhabitants of Amazonia, despite the sometimes genocidal ferocity of the five hun-
dred years of assault on their very existence, are not extinct, nor close to being
extinct.
It is true that millions of Amerindian people in Amazonia have died unnecessarily,
in a story of conquest, slavery, disease, displacement, and outright mass murder that
brings great shame upon humankind. But it is nonetheless true that as the 20th
century ends, the speakers of the hundreds of lowland South American languages
(the Arawan, Arawakan, Cariban, Je, Jivaroan, Panoan, Tucanoan, Tupian, and many
other language families) still thrive in many different regions, from Ecuador in the
west to the mouth of the Amazon in the east, from Venezuela in the north to
Paraguay in the south, and across much of the vast area of Brazil.
Indigenous South Americans survived their migration to the Americas from Asia
some forty thousand years ago, and have coped in the last five hundred years with
violence and traumatic social change on a scale that is hard to conceive of. But down
to the present day, in many parts of the region, they still pass on their ancient
languages and traditions to their children, defend their own political interests with
intelligence and skill, and make their own decisions about interacting with the
complex web of modern social, political, religious, and economic life in South
America.
Down through the many millennia of human habitation of the Amazon basin,
despite the gathering momentum of the wave of extinctions that is sweeping away
so many of the world's languages today, many of the beautiful and intriguing
aboriginal languages of the Amazon have persisted in use and not given way to
Spanish or Portuguese. Indeed, one of the languages described in this volume is
spoken by a tribe (the Wai Wai) who have been increasing their population steadily
vi Derbyshire and Pullum
in recent years rather than fading away. This book is a further expression of our
admiration and wonder at the linguistic aspect of the many intellectual achievements
of the Indians of the Amazonian area.
We dedicate this volume to the memory of Grace Derbyshire, who died unexpect-
edly in 1997 as the manuscript was being readied for the press, and is missed by
everyone who knew her. She was a participant from the start in the mission to the
Hixkaryana people that began in the 1950s, residing among the Hixkaryana people
on the Nhamundä river for many years and learning their language. She was
Desmond Derbyshire's constant companion in his linguistic and bible translation
work for forty years, and was constantly supportive of the work that led to this book.
Volume 4
South America
Atlantic Ocean
Pacific
Ocean
See p. 226 for large scale map showing locations of Warekena and some
neighboring language groups.
See p. 492 for map showing locations of Tupi-Guarani language groups.
CONTENTS
Preface v
Introduction 1
Desmond C. Derbyshire and Geoffrey K. Pullum
Wai Wai 25
Robert E. Hawkins
Warekena 225
Alexandra Y. Aikhenvald
Desmond C. Derbyshire
and
Geoffrey K. Pullum
Background
The primary intention that has guided us in compiling the Handbook of Amazonian
Languages (HAL) has been to stimulate, encourage, and promulgate scholarly work
on the grammatical structures of the languages of Amazonia. When we began
working together on this project in the late 1970s, there was no one engaged in any
such task in a general way. There were scattered individuals in a number of
countries in South America, mostly members of Christian missions, who were
studying individual Amazonian languages, but general linguistics was being prac-
tised almost entirely without reference to even the existence of Amazonian
languages.
This may sound like an overstatement,but it is not. For example, the much-re-
spected MIT Press journal of generative grammar Linguistic Inquiry is unusual in
that by the end of its first two years of publication it had published a full-length
research article on an Amazonian language (Kaye 1971, on the phonology of nasal
harmony in Desano); but a quarter of a century later there had been no other.
Linguistic Inquiry has to the present day never again published a full-length research
article on an Amazonian language. Derbyshire (1977), a short note on the Cariban
language Hixkaryana in the Squibs and Discussion section, was the next piece of
original research on an Amazonian language to appear in Linguistic Inquiry, and in
subsequent years other squibs were published on Kamaiurä (Tupian), Piraha (Mu-
ran), and Sharanahua (Panoan). But that is the total.
The point we are making is not about Linguistic Inquiry in particular; we choose
it as our example simply because it publishes a good language index that permits the
foregoing assertions to be easily checked. A similar picture would emerge if the
back runs of almost any other journal of general linguistics were scanned (one
exception, because of its specialized subject area, would be the International Journal
of American Linguistics}. And it should be noted that in grammar the absence of
Amazonian languages has been more extreme than in phonology. Two decades ago,
essentially nothing at all was appearing in the linguistic literature about the grammar
of any Amazonian languages, though various phonological studies had appeared, and
today the situation is only very slightly better.
It is satisfying to be able to record signs pointing to a distinct improvement in the
situation. Since the last volume of the Handbook (HAL 3) was published in 1991,
there has been a very encouraging flow of high-quality research, fieldwork, and
publication on languages of Amazonia, and a number of scholars have commenced
energetic research work and graduate training in the area. We cite four examples:
(i) Professor Spike Gildea is now doing Amazonian linguistic research and training
graduate students at Rice University. His fieldwork was done in Venezuela
among the Panare, and his doctoral dissertation on Cariban morphosyntax was
completed in 1992. Soon after that he went to Brazil for a more extended period
4 Derbyshire and Pullum
of fieldwork on Cariban languages. His main focus since his doctoral research
has been on Kaxuyana, and he has a grammatical sketch in preparation. He has
re-written his dissertation in the light of the vast amount of additional material
he has gathered since it was written (Gildea 1998), and along the way has
published several other papers.
(ii) Dr Denny Moore, at the Museu Goeldi in Belem, was in part responsible for
enabling Gildea to accomplish his work in Brazil. The collaboration between
Moore and Gildea has resulted in two Brazilian students, Sergio Meira and
Petronila da Silva Tavares, doing fieldwork on Cariban languages in Brazil
(Meira on Trio (Tiriyo) and Bakairi, Tavares on Wayana) and graduate work
in linguistics in the United States. As this volume goes to press they are
graduate students at Rice University under Gildea's supervision.
We could add a number of other such examples, and are confident that much more
is in prospect. We stress particularly the importance of the involvement of scholars
from Brazil in this work. We believe that a large part of the future of Amazonian
linguistics will rest on whether Brazilian linguists adopt with enthusiasm and com-
mitment the scientific study of the indigenous languages of their country.
It is a startling fact that at the time we write these words, there appear to be only
two grammars of Brazilian Amazon indigenous languages—both now extinct—that
have ever been written by Brazilians, and the two publications were separated by
exactly four hundred years. The first was the pathbreaking Arte de grammatica da
lingua mais usada na costa do Brasil (Anchieta 1595), a grammar of the once widely
spoken but now extinct Tupinambä. Anchieta was born in the Canary Islands
(Spanish territory), but he died in Brazil in 1597 and is seen today as a Brazilian
hero, a founder of the national literary tradition of that country. His grammar was
Introduction 5
written in 1555 but not published until forty years later. The second grammar of a
Brazilian language by a Brazilian was Aikhenvald (1995). Aikhenvald was not born
in Brazil either, but was a Brazilian citizen by the time she did her work on the
language. Just like Tupinambä, the language she described, Bare, is now extinct.
As of the time of writing this, then, not a single book-length grammar of a
Brazilian indigenous language by a Brazilian-bom linguist has yet been published,
and no Brazilian has ever published a grammar of a language that is spoken in the
region today.
We are confident that the long period during which these things have been true is
likely soon to end. There are now significant numbers of Brazilian linguists working
on languages indigenous to Amazonia. To name just a few whose work we have
encountered: Tania Clemente de Souza, Raquel Costa, Carmen Teresa Dorigo,
Charlotte Emmerich, Bruna Franchetto, Nilson Gabas, Yonne Leite, Marcus Maia,
Sergio Meira, Filomena Sandalo, Marilia Faco Soares, Luciana Storto, Petronila da
Silva Tavares, and Marcia Maria Damaso Vieira. They are working at institutions
spread from Rio de Janeiro (where an active group is centered at the Museu
Nacional) to Cambridge, Massachusetts (where at least two Brazilian graduate stu-
dents have been enrolled in the doctoral program at MIT's Department of
Linguistics).
The first three volumes of HAL included no work by Brazilians, though we had
looked for and solicited such work. Everything published in HAL so far has been
contributed by North American and British scholars who have developed lasting
interests in the languages of Amazonia. The work of these foreign residents and
visitors to Brazil has been of inestimable value, but it was never our wish to limit
HAL to their work. We look forward to work by Brazilian and other South American
scholars presenting original research, and criticizing and correcting the preliminary
work on Amazonian languages that HAL has been able to make available so far. It
is with pleasure that we greet the publication of Aikhenvald's chapter in this fourth
volume (see our discussion below), which at last adds a Brazilian national to the list
of those whose work has appeared in the series. We hope that in the near future we
will see other South American linguists publishing full grammatical descriptions of
the many fascinating languages still spoken in Amazonia.
It seems to us from a review of the flow of recent nonlinguistic publications on
Amazonia that it is perhaps an auspicious time for renewed attention to recording
and analyzing the languages of the area, for there has recently been a very substan-
tial increase in the rate of appearance of books about nonlinguistic aspects of the
lives and the environment of Amazonian Indians, many of them aimed at quite a
wide public.
The vast area of the Amazon basin has always engaged the imagination of those
who heard about it, ever since the first arrival of Europeans there at the beginning
of the 16th century. It has perhaps provided more potent images of the original
Edenic state of man and nature than any other place on earth. There is a generic
vision of Amazonian Indians that has been put before the public through tales of
6 Derbyshire and Pullum
fierce and hostile tribes who make shrunken heads as mementos of their decapitated
enemies (a practice associated with the tribes speaking Jivaroan languages). The title
of Francis Huxley's Affable Savages (1956), about the Urubu (see HAL 1 for a
description of their language), suggests a deliberate repudiation of that image.
Huxley is at pains to have his reader understand that the Urubu were friendly and
treated him well; but his book strongly suggests throughout that his experiences
enable him to generalize about all the Indians of Brazil: they are all naked, all
practice slash-and-burn shifting agriculture, are all for the most part innocent of
science and technology, are all deeply absorbed in spiritual discussion and myth-
making, are all full of tales of past warfare and cannibalism.
In truth, of course, the Amazon's people are highly diverse: some traditionally
lived in hunter-gatherer bands of as few as a hundred people with a reputation for
shunning contact with all outsiders and killing intruders who approach them (the
Atroari; the Kren-Akorore); others live in large, fairly settled communities in which
there may be tens of thousands of speakers of a single language (the Macushi; the
Shuar Federation). Some have traditionally lived with little technology but arrows,
darts, and spears, but many have amassed considerable stores of scientific knowl-
edge about the environment and its products (e.g. the chemistry of dart-tip poisons,
fish-stunning compounds, herbal medicines), and in some areas such arts as boat-
building and pottery-making were developed to levels of accomplishment that
astonished early European travellers. Naturally, many have been in continuous
casual contact with Portuguese or Spanish speakers for hundreds of years.
Amazonian Indians tend to be thought of as inhabiting primeval jungle, but hardly
any of Amazonia is appropriately described in that way. The original inhabitants of
the Amazon basin have constantly modified it and adapted it over millennia, harvest-
ing its natural food products and planting gardens to grow more, burning patches to
open up new space for growing food crops, and so on, over a period that is almost
certainly to be measured in tens of millennia: human beings have certainly been in
Amazonia for 10,000 years, and could have been there for more than 30,000 years
(see Wolkomir 1991 for a survey of recent research on the antiquity of settlement
sites such as Pedra Furada in Brazil).
In recent years, the literature on such aspects of the lives of indigenous Amazonian
people has been burgeoning. Even if we keep the task within bounds by restricting
ourselves just to books in English published since we prepared HAL 3 (i.e. since
about 1990), there is a veritable flood of new material to review, both popular and
academic. The books that have appeared address numerous aspects of the cultures of
aboriginal Amazonia: their art (Nugent and Coelho 1991; Verswijver 1992b); their
oral literature (Wilbert and Simoneau 1990; Mindlin et al. 1995); their spirituality
(Perkins 1994; Henley 1995); their science and natural history (Milliken et al. 1992;
Descola 1994); and the linked topics of their internecine warfare (Verswijver 1992a;
Hendricks 1993; Redmond 1994; Ferguson 1995) and the continuing genocidal
attacks on them by others (Berwick 1992; Stannard 1992; Penglase 1994). In addi-
tion, popular works on the development of Amazonia containing generally
Introduction 7
sympathetic accounts of the negative effects on the Indians have continued to appear
(see e.g. Kimerling 1991; Margolis 1992; Meunier and Savarin 1994; Kane 1995).
It can hardly be said that there are general grounds for optimism about widespread
improvement in the conditions under which Amazonian Indians live. In the north of
Brazil, near the Venezuelan border, murderous attacks on the Yanomami Indians,
carried out by illegal gold miners encroaching on their land, have continued without
serious government hindrance, despite repeated coverage in the world's press. And
throughout Amazonia, the appalling damage from large-scale burning of the rain
forest—way beyond anything the aboriginal population could ever have conceived—
has continued, and recently began to increase. On January 26, 1998, the Brazilian
government issued statistics on the pace of environmental destruction (see Schemo
1998 for a news report). While deforestation had begun to tail off after 1988, falling
to a low of a little over 4,000 square miles in 1991, it then rose again, and in the
1994-1995 burning season a staggering total of 11,196 square miles of rain forest
was completely destroyed by fire. And that is the Brazilian government's figure,
published after a delay while Brazil participated in the 1997 Kyoto world conference
on global warming and the October 1997 meeting in Manaus at which Brazil asked
the world's seven wealthiest countries for further environmental support. The total
amount of forest destroyed may be higher. Taking into account logging and thinning
of virgin rain forest under the canopy as well, a United States Congressional
commission in 1997 claimed that the forest destruction was well above 22,000
square miles per year.
These figures may be put in perspective by noting that the Amazon rain forest area
is about the size of Western Europe. Destroying more than 11,000 square miles is
comparable to burning off the whole of Belgium. Destroying over 22,000 square
miles could be accomplished by deforestation of the whole of Denmark and the
entire area of Northern Ireland. A continuation of annual figures like these will
mean, clearly, that the Amazon rain forest is doomed. And far from taking steps to
discourage these developments, the government has been (perhaps unwittingly)
taking actions that encourage them. As of early 1998 its federal environmental
agency still had no legal authority to enforce laws relating to the environment
(though legislation to remedy this was under discussion), and a program of agrarian
reform had put previously landless peasants in control of tracts of virgin rainforest
(over 18,000 square miles of it). Clearing the land by burning in order to attempt
agricultural production on it is really these peasant's only immediate option. Much
of the recent deforestation has been not through massive clearings by agribusiness,
but through relatively small fires—a third of a square mile or less—apparently set
by smallholders.
Events the other side of the Pacific in 1997 showed that land clearing through
burning can get out of control on a multinational scale. The burning season in parts
of Indonesia (partly through illegal bum-offs by large palm plantation operations, but
also by small farmers) led to uncontrolled fires over such a gigantic area that a
blanket of smoke and smoggy haze descended not just on much of Indonesia but also
8 Derbyshire and Pullum
over Singapore, parts of Malaysia and Thailand, and even the Philippines. Illnesses
and deaths from respiratory ailments increased and airliners crashed in the murky
air. Yet the fires raged on, rendered unstoppable by the unusual dryness attributable
to a particularly strong El Nino ocean warming effect that year. A study by the
Woods Hole Research Institute has reported signs of unusual dryness in the Amazon
area (which normally contains some 20 percent of the fresh water in the world and
continually recycles it through the process of evapotranspiration in the wet forests).
The likelihood of out-of-control burning that would have consequences for the whole
continent of South America thus cannot be dismissed.
If environmental destruction on the kind of scale that is reported from Brazil
continues, the effects on the remaining viable Indian communities will be severe.
Their traditional life depends on being settled on land that is capable of sustaining
them. If the land is taken over and changed beyond recognition, whether by gold
miners, cattle ranchers, road gangs, or oil drilling crews, and the Indians become
marginalized inhabitants in the way Australian Aborigines were turned into semi-le-
gal squatters on their own traditional land in the past century, there is essentially no
likelihood of their cultures and languages remaining intact. And without some
recognition and enforcement of the rights of Indian groups to land on which they
can subsist, the environmental destruction is very likely to continue. The fate of the
Amazonian environment and the fate of the linguistic richness HAL has been at-
tempting to document are thus to a considerable extent interlinked.
However, we believe that they do provide a reliable description of the main charac-
teristics of a wide range of phonological, morphological, semantic, and syntactic
phenomena. Reviews of, citations from, and other reports we have received about
earlier volumes indicate that they have proved of value to linguists specializing in
various areas of research, including functional-typological, historical-comparative,
and formal theoretical models. (See also, for their comments and citations, Dixon
1994, xv; Everett 1997; Gildea 1998.) We have some confidence that this fourth
volume will also prove useful to scholars interested in these still very little-known
languages.
Wai Wai
Wai Wai is a member of the Cariban language family. Three Cariban languages
have now been described in HAL volumes: Apalai is described in HAL 1 and
Macushi in HAL 3. Putting these together with Derbyshire's detailed descriptions of
Hixkaryana (e.g. Derbyshire 1979, 1985) and the recent work of Spike Gildea (see
e.g. Gildea 1998), they begin to provide adequate materials available for serious and
detailed comparative syntactic work on Cariban languages. Wai Wai is spoken by
communities living on both sides of the Guyana-Brazil border, the main locations
being on the Rivers Essequibo (Guyana)and Mapuera (Brazil). There are also Wai
Wai speakers living among the Trio (Tiriyo) in Surinam, and others living with the
Hixkaryana in Amazonas state in northern central Brazil.
In contrast with some other Amazonian languages (see particularly Warekena,
discussed below), the Wai Wai language and culture is very much alive and well. It
is worth stressing this, for the Wai Wai offer an important illustration of why it
would be a mistake to see indigenous people of South America as doomed to
extinction. Despite the stark facts of population decline and ongoing assaults by
modern Brazilian society on indigenous peoples that we have outlined in the intro-
ductions to previous volumes (see especially HAL 7.2-10, HAL 2.3-5, and HAL
3.4-8), it is worth noting that Amazonian indigenous people are not without the
ability or resources to determine and improve their lives, and there is no reason at
all why their overall decline in numbers since the sixteenth century should be seen
as something inevitable. The Wai Wai population has been growing steadily for at
least the last 40 years, both by natural reproduction within the group and also by
absorption of remnants of other neighboring language groups who have learned to
speak Wai Wai. There are now about 1,800 speakers, and the number is continuing
to increase.
There have been more anthropological and other studies of the lives of the Wai
Wai than of many peoples of the Amazon area. For anthropological works, see e.g.
Fock 1963 and Yde 1965; more general travel books about the Wai Wai include
Guppy 1958 and Allen 1985; and there have also been two books on the Wai Wai
Christian culture (Dowdy 1963, 1995). Thus the Wai Wai people and culture have
become fairly widely known.
10 Derbyshire and Pullum
Their language also has a longer history of description than most Amazonian
languages. There are descriptions of the phonology and nominal and verbal morphol-
ogy of the language in a number of articles: W. N. Hawkins (1952, 1962); W. N.
Hawkins and R. E. Hawkins (1953); R. E. Hawkins (1962). However, Robert
Hawkins' grammatical sketch of Wai Wai in this volume is the first general and
more complete description of the language, including its syntax.
Robert Hawkins and his wife first began their missionary work among the Wai
Wai people in 1950. This resulted in a Wai Wai translation of the New Testament
in the early 1980s, and they are now completing the translation of the entire Bible.
This work has given them a deep understanding of the semantics of the language, as
will be evident in such sections as 18.1 on "tense, number and emotional involve-
ment" distinctions. Grimes (1975) has coined the term 'pesky little particle' for
uninflected words appearing to have a semantic or pragmatic function but one that
is extraordinarily difficult for the linguistic investigator to identify. Wai Wai
abounds in pesky little particles, and Hawkins' extended discussion of them in
section 21 is remarkably rich in insight and detail concerning their contributions to
utterance meaning. Hawkins also displays a thorough knowledge of the range of
syntactic constructions and the complex morphology of Wai Wai. Section 23 is by
far the most extensive and in-depth description of the morphological structures and
the conditioning factors of the many allomorphic variants that has appeared in any
of the ten grammatical sketches published in the four HAL volumes.
Wai Wai is probably the most closely related Cariban language to Hixkaryana
(Derbyshire 1979, 1985). There are many similarities of form and meaning in both
the grammar and the lexicon. As is common in Cariban languages, in both Wai Wai
and Hixkaryana subordinate constructions (i.e., the equivalent of complement, adver-
bial, and relative clauses) are expressed by nominalizations which are ergatively
organized, with the notional intransitive subject and transitive object taking posses-
sor marking and the notional transitive subject being marked by a postposition (wya
in Hixkaryana and ya ~ wya in Wai Wai). The forms of the nominalizers and of verb
derivational affixes are almost identical in the two languages. Nonetheless, these are
not two dialects of a single language. There are notable grammatical differences
between the two. For example:
(i) In Wai Wai main clauses (sect. 9.2), the sequencing of direct object before verb
is less rigid than it is in Hixkaryana (and most other Cariban languages).
Moreover, the subject can occur before or after the verb, apparently without
any sort of conditioning factors. In Hixkaryana clear discourse-pragmatic con-
ditions have to be met in order for a clause-initial subject to be permissible
(Derbyshire 1986).
(ii) In Wai Wai there are only two basic past tense categories, one being 'today
past' and the other 'any past time prior to today', and the latter category has
two forms depending on whether the speaker is or is not 'emotionally in-
volved'. The Hixkaryana tense system contrasts sharply. In Hixkaryana there
Introduction 11
are three basic categories: immediate past (earlier today), recent past (within the
past few weeks or months), and distant past; and in the recent and distant
categories there is an aspectual distinction between completive (including se-
melfactive and punctual) and continuative (including habitual and progressive),
making five sets of past tense forms in all. The Wai Wai tense system is both
simpler and organized by reference to rather different semantic categories.
Warekena
Warekena (frequently found in reference books under other spellings such as
'Guarequena') is a member of the Northern Branch of the Maipuran language
family, which is the core of the Arawakan family. David Payne's chapter in HAL
5.355—499 may be consulted for a discussion of some languages which are certainly
Arawakan (within the Maipuran subfamily) and some others which have been
classified as Arawakan by earlier linguists but which may not be Arawakan at all.
Warekena is the first grammatical description of a Maipuran Arawakan language to
appear in HAL. In the introductions to earlier volumes we expressed our hope to
include descriptions of two others, but these have not been forthcoming (though
there are two comparative studies of the main morphosyntactic features of the
Arawakan languages in HAL 1: Derbyshire's on Brazilian Arawakan, HAL 7.469-
566, and Wise's on the PreAndine Arawakan languages of Peru, HAL 7.567-642).
There are several communities of Warekena speakers on the Xie river, a tributary
of the Upper Negro in the extreme north of Brazil, very close to the Colombia and
Venezuela borders. In contrast to Wai Wai, Warekena, which is a dialect of Baniwa
of Guainia (spoken in Venezuela), appears to be in a situation likely to lead to
language death, and many speakers no longer use the language for everyday com-
munication (in Brazil, they all speak Nheengatu, once known as Lingua Geral, and
the speakers in Venezuela all know Spanish).
Alexandra Aikhenvald has done extensive fieldwork on several languages in the
area where Warekena is spoken. She has worked on Bare (now extinct, the last
fluent speaker having died in 1993; her description of the language was published as
Aikhenvald 1995), Baniwa of I9ana, and Tariana. As Aikhenvald shows, there are
many similarities between those three languages and Warekena (some are more
general Northern Maipuran traits), although she also reports that the Baniwa of I9ana
language is not mutually intelligible with Baniwa of Guainia, the language of which
Warekena is a dialect.
Basic constituent order in main clauses in Warekena is SVO, but in clauses with
intransitive verbs there are two possibilities: those with active intransitives are SV
and those with Stative intransitives are VS. Most of the ten morphosyntactic charac-
teristics listed in our introduction to HAL 1 (see p. 19) are found in Warekena,
including: verb agreement with both subject and object; subject and object NPs
occurring infrequently and only to express new information or for pragmatic mark-
ing; no agentive passive construction; heavy reliance on direct speech rather than
12 Derbyshire and Pullum
Amahuaca
In our introduction to HAL 1, we stated our intention to include a description of a
Panoan language. The complete grammatical sketch we had hoped for has not
materialized, but the Amahuaca paper by Margarethe Sparing-Chavez in this volume
helps to fill the gap. It provides a description of one of the more unique interclausal
reference systems to be found anywhere in the world, and it is a system which is
common to Panoan languages. It has the more general characteristics which are
usually found in switch-reference systems (i.e., signalling whether the subject is
"same or different" in a sequence of clauses), plus the feature of encoding corefer-
ence between subjects and objects.
Switch-reference is, however, only one function of this system of morphological
marking; other functions include the encoding of transitivity and of temporal and
logical relations between events. The set of suffixes which compose the system is a
combination of case markers (ergative, absolutive, and nominative) and tense-aspect
marker, which in some cases include a person-marking component.
Other parameters that are relevant to the system are constituent order in clauses,
temporal adverbs, discourse-pragmatic factors, and verbal suffixes which encode
motion/direction and potentiality of an event. Thus, a major part of the morphosyn-
tax of the language enters into the description of this complex system.
For other publications on Panoan languages see footnote 1 and the References in
the Sparing-Chavez chapter. For a survey of the Panoan language family see the
chapter by Eugene Loos in Dixon and Aikhenvald (forthcoming).
Tupi-Guarani
In HAL 1 there was a grammatical sketch of one Tupi-Guarani language (Urubu-
Kaapor) and a typological study of certain morphological and syntactic traits in
another (Guajajara). In this volume we present a comparative study of the Tupi-
Guarani family by Cheryl Jensen, who has done in-depth research in one
Introduction 13
Tupi-Guarani language (Wayampi) and has already published historical and com-
parative studies on the family as a whole (see Jensen 1989, 1990, and other items
listed under her name in the References of her chapter).
Tupi-Guarani is the best documented family in lowland South America. There are
descriptions of two (now extinct) languages that go back to the 16th and 17th
centuries: Tupinambä (Anchieta 1595, mentioned above) and Old Guarani (Ruiz de
Montoya 1639, 1640). In the past forty years, many more studies of the grammars
(or parts thereof) of individual languages have been published or archived. This has
led to increased interest in historical and comparative research, largely inspired by
Aryon Rodrigues, and this has resulted in the publication of generally reliable
reconstructions and internal classifications. It has also resulted, of course, in consid-
erable debate and diverse analyses of certain phenomena, as will be noted by anyone
who compares, for example, Lemle (1971), Rodrigues (1985a, 1985b), Dietrich
(1990), and Jensen (1989, 1990).
Jensen takes sixteen areas of morphosyntax and carefully traces the changes that
have taken place from the Pro to-Tupi-Guarani forms and constructions to those
found in the descendant languages. The proto-forms are based on the phonological
(Lemle 1971) and morphological (Jensen 1989) reconstructions that have become
generally accepted by Tupi-Guarani specialists. She relates these diachronic changes
to the eight sub-groups of Tupi-Guarani languages which have been postulated by
Rodrigues (1985a) on the grounds of closer affinities within each sub-group, and
shows how these changes in the morphosyntax have developed in a systematic way
at different historical stages, affecting some sub-groups and not others.
In section 17, Jensen focuses on the verb agreement cross-referencing systems and
presents a reasonable hypothesis on how these could have developed from an earlier
(Pre-Proto-Tupi-Guarani) ergative system, with consistent absolutive cross-refer-
encing that applied to all independent verb and dependent verb constructions,
through various stages in which some descendant languages lost the absolutive
agreement patterns on independent verbs, and on to the sub-set of languages to-day
(Chiriguano, Guarani Mbyä, Guarayu, Kaiwä, and Wayampi) in which absolutive
cross-referencing has been lost in some of the dependent verb constructions as well
as on all independent verbs, and finally to the case of Urubu-Kaapor (see HAL 1 for
a description), which has lost all absolutive marking in all constructions.
This Tupi-Guarani chapter is the second detailed historical and comparative work
that has appeared in the Handbook series. David Payne's Maipuran (Arawakan)
chapter in HAL 3 presents the phonological and lexical data which are the basis for
his reconstructions and internal classification, with only a brief section on grammati-
cal structures. Jensen, in contrast, begins with previously established reconstructed
forms and internal classification (Rodrigues' sub-groups) and traces the develop-
ments in many areas of the morphosyntax of Tupi-Guarani languages.
14 Derbyshire and Pullum
Areal characteristics
The papers in this volume tend to confirm some of the Amazonian typological
similarities we noted in earlier volumes (HAL 7.16-20; HAL 3.9-12). Warekena is
unusual with respect to one areal feature we noted in connection with basic constitu-
ent order, i.e., object preceding subject in main clauses. Warekena is SVO. SVO also
appears to predominate in Tupi-Guarani languages (although each of the other
possible orders, except VOS, occur as the basic order in at least one Tupi-Guarani
language). Jensen reports, however, that dependent clauses are usually verb final and
opts for SOV as the likely order in Pro to-Tupi-Guarani. In Wai Wai and Amahuaca
SOV and OVS are the more common orders. These give further support to the
hypothesis stated in HAL 1.20 that at an earlier stage in their history most Ama-
zonian languages were fairly rigidly SOV.
What is also characteristic of all the languages described in this volume is that the
order of clausal constituents is greatly affected by discourse-pragmatic factors. Most
clauses occurring in texts do not have noun phrase (or free pronoun) subjects and
objects; noun phrases tend to be used only to express new information or for other
such pragmatically-defined effects. Otherwise, person marking in the verb is suffi-
cient to refer to subject and/or direct object (in Amahuaca the person marking is a
component of some of the interclausal reference suffixes).
Ergativity is clearly in evidence, to a greater or lesser degree, in the languages
described here. In Warekena it has only a minor syntactic role relating to a constraint
on coreferential deletion of a subject constituent in certain coordinate and subordi-
nate constructions (see section 3.2 and footnote 3 of the Warekena chapter for
details). In Wai Wai, as in many other Cariban languages, many of the nominalized
subordinate constructions are ergatively organized in that the notional intransitive
subject or transitive direct object surfaces as the possessor of the nominalized verb;
the notional transitive subject is governed by the postposition (w)ya 'to, by', which
is a cognate of the main clause ergative marker in a few other Cariban languages,
e.g., Macushi, which is described in HAL 3. Amahuaca has main clause ergative
marking, and the ergative marker is incorporated into some of the interclausal
reference suffixes. Most of the Tupi-Guarani languages, as noted above, still have
absolutive verb cross-referencing in at least some of their constructions and, as
Jensen argues, it seems a reasonable hypothesis that this reflects a more complete
ergative system in the earlier stages of the family history which led up to Proto-
Tupi-Guarani.
Jensen's diachronic scenario for the loss of ergativity in the Tupi-Guarani family
is convincing. It gives more substance to Harrison's (1986) hypothesis of the erosion
of ergativity in Guajajara and other Central Brazil (Tupian and Je) languages. It may
also be significant in suggesting a similar scenario for the direction of change in
other Amazonian language families in which there are languages with varying
degrees of ergativity.
The Cariban family may or may not prove to be an exception. Derbyshire (1991,
1994) proposed a comparable hypothesis of progressive loss of ergativity for
Introduction 15
Cariban, but this faces a serious challenge mounted by Gildea (1998), who has
accumulated a large amount of data unavailable to Derbyshire, taken from many
Cariban languages. Gildea's conclusion is that main clause ergative marking is an
innovation in a few languages since Proto-Carib. His argumentation bears very
serious attention, though given the evidence of areal confluence of syntactic features
in Amazonia, it could conceivably be put in doubt if the direction of change in other
Amazonian language families proves to be gradual loss of ergativity since the
proto-stages. There are a significant number of families in the area with languages
that display varying degrees of ergativity, which suggests that ergativity has been
around for a long time. These include the following languages that have been
described in the Handbook series: Apalai, Macushi, Wai Wai (Cariban), Canela-
Kraho (Je), Guajajara (Tupi-Guarani), Sanuma (Yanomami), Paumari (Arawan), and
Amahuaca (Panoan). Other families in the area reported to have ergativity are:
Tacanan, and Tupi groups which are not part of the Tupi-Guarani family.
REFERENCES
Aikhenvald, Alexandra Y.
1995 Bare. Lincom Europe Materials: 100.
Allen, Benedict
1985 Mad white giant (London: Macmillan).
Anchieta, Joseph de
1595 Arte de grammatica da lingua mais usada na costa do Brasil (Coimbra:
Antonio Mariz). Reproductions published by Biblioteca Nacional, Rio
de Janeiro (1933) and Editora Anchieta, Säo Paulo (1946).
Berwick, Dennison
1992 Savages: the life and killing of the Yanomami (London: Hutchinson,
1992)
Derbyshire, Desmond C.
1977 "Word order universale and the existence of OVS languages," Linguis-
tic Inquiry 8.590-99.
1979 Hixkaryana. Lingua Descriptive Series, 1 (Amsterdam: North-Holland).
1985 Hixkaryana and linguistic typology (Dallas: Summer Institute of Lin-
guistics and University of Texas at Arlington).
1986 "Topic continuity and OVS order in Hixkaryana," Native South Ameri-
can discourse, edited by Joel Sherzer and Greg Urban (Berlin: Mouton
de Gruyter), 237-306.
1991 "Are Cariban languages moving away from or towards ergative sys-
tems?" Work Papers of the Summer Institute of Linguistics, University
of North Dakota Session, 25:1-29.
1994 "Clause subordination and nominalization in Tupi-Guaranian and
Cariban languages," Linguistica Tupi-Guarani/Caribe, edited by Mary
R. Wise (Revista Latinoamericana de Estudios Etnolingisticos VIII,
edited by Ignacio Prado Pastor), 179-98.
Descola, Philippe
1994 In the society of nature: a native ecology in Amazonia. Cambridge
studies in social and cultural anthropology; 93 (Cambridge: Cambridge
University Press).
Dietrich, W.
1990 More evidence for an internal classification of Tupi-Guarani languages
(Berlin: Gebr. Mann Verlag).
Dixon, Robert M. W.
1994 Ergativity (Cambridge: Cambridge University Press).
Introduction 17
Yde, J.
1965 Material culture of the Waiwai (Nationalmuseets Skrifter, Etnograflsk
Roekke, Copenhagen).
Part I
Grammatical Sketches
Outline of contents for grammatical sketches
INTRODUCTION
1 Word order
2 Parataxis
3 Ellipsis
4 Reflexives, reciprocals, unspecified arguments
5 Passives
6 Causatives
7 Comparatives, equatives
8 Coordination
9 Pragmatic and discourse characteristics
10 Interrogatives
11 Imperatives
12 Negation
13 Anaphora
14 Subordinate clauses
21 PARTICLES
22 PHONOLOGY
23 MORPHOLOGY
24 IDEOPHONES
NOTES
REFERENCES
APPENDIX: TEXT
WAI WAI
Robert E. Hawkins
Missäo Evang^lica da Amazonia
Introduction
Wai Wai is a language of the Carib family spoken in a village on the Mapuera river
of Para, Brazil, and in another village on the Jatapuzinho River of Roraima, Brazil,
by a few people living among the Hixkaryana people on the Nhamundä River in
Amazonas, Brazil, in another village near Gunn's Strip on the Upper Essequibo
River in Guyana, and by a group living among the Trio Indians in Southern Surinam.
At this writing there are about 1800 speakers of the language and the number is
increasing each year. Actually, there are no people who call themselves Wai Wai
alive at this time; the last member of the group who called themselves by this name
died 20 years ago or more. The people who are now called Wai Wai by the outside
world are remnants of the Mawayana, Hixkaryana, Katwena, Sherew, Karapaw
Yana, Cikyana, Tuuna Yana, and Parkwoto peoples. There were two Taruma men
who lived among them for some years, but they both died several years ago. Also,
one man of the Atrowari tribe lives among the Wai Wai, and there are two Trio
women who are married to Wai Wai men. All of these tribe remnants have their own
language or dialect, some of which are quite close to Wai Wai and some quite
different. The language of the true Wai Wai people was very close to that of the
Parkwoto people. The language of these two groups combined to become the lingua
franca of all these remnant groups, and nearly all who live among the Wai Wai
speak the Wai Wai language fluently. Some of them, however, still speak their own
language in their homes. It has been visitors from the outside world who have given
the name Wai Wai to the whole group who speak this language. The Wai Wai who
live in Guyana are learning English, those living in Surinam are learning Dutch, and
those living in Brazil are learning Portuguese. A good number of the Wai Wai young
men can do business in their respective trade languages, but there are very few who
speak any of these trade languages fluently. All of the Wai Wai, however, are being
acculturated at a fairly steady rate into the culture of the nations in which they live.
The younger people are naturally acculturating faster than the older ones. About
85% of the Wai Wai can read the Wai Wai language. Many of the young people
read and write their language fluently.
My brother, W. Neill Hawkins, did nearly all of the work on the phonology of the
Wai Wai language, as well as that of the description of the morphology of the nouns.
26 Hawkins
I did the greater part of the analysis of the verb morphology. I remained among the
Wai Wai after my brother left and have added to some of his descriptions of the
phonology and the morphology of the nouns. (See References at the end of this
paper.) I have done most of the work on the syntax of the language. The semantics
of a number of the particles was largely my work. It took many years to discover
the full meaning of some of them. I was aided in this study by Marjorie Crofts and
Margaret Sheffler of the Summer Institute of Linguistics.
1 Word order
The basic clause types of Wai Wai, classified according to their nuclear constitu-
ents, are: transitive (1), intransitive (2), copular (3), equative (4), and quotative (5).
All of these clauses have a predicate and a subject. After a participant is introduced,
the name of this person or thing is most often omitted, and a verb prefix is the only
indicator of the participant, whether he is the actor or recipient of the action. When
the subject prefix is third person it quite often is zero. The following examples
contain mostly nuclear constituents.
1.1 Transitive clauses. The transitive clause has three essential constituents: direct
object, a transitive verb, and subject. The normal position for a free form object is
immediately before the verb. Variations from the normal position are considerable
and are further discussed in sect. 9.2. There does not seem to be any normal position
for a free form subject. It may be either before or after the verb. In this sect, we will
only present independent transitive clauses. Dependent transitive clauses also occur
and are discussed in sect. 14.
The free form subject may be placed before the verb to highlight the subject.
Transitive verb person-marking prefixes show a split between subject and object
person markers, which constitute two distinct sets. Only one set actually occurs in
any given finite verb. There is a hierarchy of personal prefixes to transitive verbs,
with first and second person indicators being more overt than third person indicators.
There are no third person object prefixes, only first and second person object
prefixes. When the verb has a first or second person subject and a third person
object, only the subject prefix occurs, and the object has no indicator affixed to the
verb (9). When the verb has a first or second person object the object prefixes occur
and the subject, which again is always third person, has no indicator (11, 12). But
the third person subject or object is very clear in the mind of speakers even though
it is not overtly indicated. It is a significant absence, unless a third person free form
subject or object occurs (10). Thus it seems logical to me to indicate such with a
zero in the glosses of the prefixes. There is also a special portmanteau prefix form,
ki-, which indicates that the subject is first person and object is second person (13).
When the object-subject relation is reversed, the second person subject is indicated
by the normal subject prefix while the first person object is indicated by the first
person pronoun (14). See sect. 23.3 for paradigms of subject and object prefixes.
28 Hawkins
(9) WT0-ahsi-ya-si
lS-3O-hold-SF-INP
º am holding it.'
(13) K-enw-o.
lS + 2O-see-TP
º saw you.'
(14) Ow m-etap-e-si.
1PRO 2S-hit-SF-INP
'You are hitting me.'
In second person imperative transitive clauses a third person object is again either
zero or a free form pronoun. As to first person objects, there is a variation between
positive and negative imperative forms. With positive commands first person object
is always a free form. With negative commands it is usually a prefix, which is
attached to the complement of the imperative form of the copula, (sect. 1.3, ex (33),
and (16) below.)
(15) 0-Ahsi-ko.
3O-catch-2IMP
'(You) catch hold of it.'
There is one transitive verb stem that commonly breaks the rule that a direct
object is an essential part of the transitive clause, and only then when certain
postpositional phrases occur with this verb (17). It is the stem of the verb 'to see.'
Note in (18) that the object in the second sentence is indicated by zero.
Wai Wai 29
There is one type of transitive verb phrase indicating transitive action with an
unspecified personal object. It consists of the transitive verb stem nominalized by the
suffix -no and obligatorily followed by a form of the transitive verb -iri- meaning
'to put it,' 'to make it,Or 'to do it.' The verb stem with the suffix -no becomes the
direct object of the verb -iri-. The verb -iri- may be classed as an auxiliary verb
since it carries person of the subject, tense, and mode components of the action of
the nominalized verb which precedes it (19). Also when this verb is used as an
auxiliary verb derived to attributive or nominalized forms (21, 22) the general prefix
ci- is omitted (see sect. 23.4.5.1 (ii)). In example (20) the word ciiso is the main
verb, not an auxiliary verb.
la-wc-e-si.
IS-go-SF-INP
º am going to make a medicine dispensary (house).'
1.2 Intransitive clauses. The intransitive clause has two essential constituents, an
intransitive verb and a subject. The subject may be a free form plus subject prefix
or only a subject prefix.
30 Hawkins
(23) Êß-wMk-ya-sl
IS-sleep-SF-INP
º am going to sleep.'
1.3 Copular clauses. The copular clause has two essential elements, a form of the
copula and a subject. There are two kinds of copular clauses, clauses of existence
(26-28) and clauses of description (29-30). A copular clause of existence may include
a postposition phrase or an adverb phrase indicating location and also descriptive nouns
which modify the subject (27). But it does not include any adverbs describing attributes.
The subject of existence clauses may follow or precede the copula (26-27). In interroga-
tive clauses the normal order seems to be subject following the verb (26). Copular
clauses of description include adverbs and adverbial constructions which tend to occur
before the copula. A one word copular sentence may occur (25) but only in a response
expression. A free form subject occurs with the copula if that subject has not already
been introduced in the immediate context (26-27).
(25) N-0-a-y.
3S-be-SF-UNP
'It is.' (In response to the question, 'Is it there?')
The copula almost always occurs following verb stems (31) and noun stems (32)
having the negative suffix -hra. The negativized verb and noun stems are thus
adverbialized. They may carry object prefixes (33) or the detransitive prefix (34),
but they do not have subject or possessor prefixes or any suffixes of tense, mode or
possession. The associated copula carries all such indicators. Note in (32) that the
subject of the copula is clearly the possessor of the negativized noun 'throat.'
1.4 Equative clauses. The equative clause is a clause that usually has no finite verb
at all. The essential constituents of equative clauses in the present tense are a predicate
nominal plus a pronoun subject. Such clauses are used for indicating description,
identity, and pointing out. When the meaning is description or identity the predicate
noun occurs first and the subject pronoun follows (35-36). When the meaning is deictic
the subject pronoun occurs first followed by the predicate noun (37). Equative clauses
are never used in logical discourse nor in time line expressions. They are commonly
used in expressions of description and even more commonly when the descriptions are
vivid or emotional. In descriptions of distant past time, equative clauses are usually
followed by a form of the copula in the uninvolved mode of the past tense (40). For a
fuller description of the deictic pronouns see sect. 16.1.
1.5 Quotative clauses. Quotative clauses consist of any expression followed obli-
gatorily by a form of the verb ka/kas 'say', 'think', or 'do'. No word ever comes
between the quoted words and the verb ka/kas.
When there is a long quote repeated by another person the verb ka/kas may not
be repeated after each sentence, but only after the last sentence. (I have not analyzed
the morphemes of this rather long example, but the quotation clause is bracketed.)
each type of clause may occur in one sentence. The Wai Wai sometimes speak in
very long sentences.
In place of, or in addition to, peripheral clauses, phrases of all types may be
added to the nuclear clause of the sentence. These may be noun, adverb or postpo-
sition phrases, and often in any of these types of phrases from one to four particles
are added.
1.6.1 Modifying nominals. These are nouns (44), noun phrases (45), nominalized
adverb phrases (46). There are no adjectives in Wai Wai. Only nouns and noun
constructions modify nouns. Modifying nominals are common in all types of clauses.
They normally occur at or near the end of the sentence. There is apparently no limit
to how many modifying nominals can be added in coordination (48). Modifying
nominals are added for identifying (44, 45, 46), clarifying (47), or expanding pre-
vious portions of the sentence (48). The modifying nominals in the following
examples are bracketed. When such nominals occur with transitive verbs the nomi-
nals seem to refer only to the objects of the verbs (44, 48). At least, I have no record
of such referring to the subject of a transitive verb.
to-no komo].
at-NOMLZR COLL
'We all who live here have weapons.'
komo], [c-enepa-ne-m],
COLL ADVZR-steal-ADVZR-NOMZR
[c-es-eyi-so-m],
ADVZR-DETRANS-scold-ADVZR-NOMZR
[c-eti-xat-m-noka-x-mu
ADVZR-DETRANS-DESE) + NOMZR-POSR-REV-ADVZR-NOMZR
komo].
COLL
º judge people here, those who are bad, liars, thieves, those who
scold, those who come to hate each other.'
In copular clauses of existence (49) and in intransitive clauses (50) the modifying
nominal is related to the subject of the verb.
[o-mtapo-ta-rJ].
1-words-VBZR-POSN
"There is something to lead me astray, my words alas.'
On rare occasions the modifying nominal may occur before the main verb. In the
following example, which illustrates this, the two noun phrases occurring together
constitute a single direct object constituent, the second being a nominalized modifier
of the first.
1.6.2 Adverbiale. Adverbials that occur in copular clauses of description have been
described above in sect. 1.3, as they are a basic part of such clauses. Here we will
present adverbials that occur with verbs other than the copula. These include indica-
tors of purpose of motion (52), manner of actions (53), purpose of other actions (54),
postpositional phrases (55), and descriptions of any kind (56, 57). There are very few
underived adverbs in Wai Wai. There are, however, many adverbialized verbs
(52-54) and adverbialized nouns (56-57). See sect. 23.5 for a description of the
morphology of derived adverbs.
ah-rf pona.
eat-NOMZR lest
º will put the cassava bread in a shelter for fear the chickens might
eat it.'
See sect. 15.4 (i) for the ergative function of the postposition -(\v)ya 'to/by' when
it co-occurs with a nominalized transitive verb, as in (54) and (58).
When two or more adverbials are related to the same verb, the verb normally
comes in between them.
t-pono-n poko.
REFPOS-clothes-POSN about
'We thought about our covering, about our clothes.'
2 Parataxis
main clause rather than following it. (The brackets within the brackets indicate a
special two-word phrase.)
yi-hxi-so.
3-drag-PM
'This is what I would like to say to you, tomorrow please come drag
my canoe.'
Two clauses with a finite verb in each may be juxtaposed in one sentence if the
meanings are closely related. There is a slight intonation drop between the clauses,
but there is no distinguishable pause like that occurring between sentences. Only
clause groupings in which the second clause explains the first or identifies something
in it are referred to here. Two clauses of similar structure in which the second refers
to additional items or actions are discussed below under coordination in sect. 2.3.
çú-hcam-hoka-tho moso.
NOMZR-ignorant-REV-POSN + PAST 3PRO
'She makes them all the time. Sister taught her how.' (Lit.: She (is)
one who makes them all the time. She (is) one who was taught by sis-
ter.)
Equative clauses in past time are often followed by the uninvolved past form of
the copula. The equative clause itself is a complete clause. The copula in this case
doesn't constitute a separate clause. It merely carries the regular affixes indicating
person, tense and number.
2.3.1 Coordination of clauses. Coordinate clauses are somewhat rare in Wai Wai.
Lists of conditions or actions that are related to each other may be spoken with
pauses in between but having the intonation pattern of one sentence rather than
multiple sentences.
kaf-pe-ra ro mak
strong-ADVZR-NEG very much IS-be-SF-INP
'My head hurts, I am dizzy, I am just very weak.'
Coordination of clauses can be within a quotative clause and the verbs of the two
clauses may be different as in the following example. But as far as my information
goes, the clitics following the verbs, if any, are always identical in such clauses.
n-es-epefa-n-to ha na,
3S-DETRANS-injure-POSN-VBZR+TP RHY maybe
Coordination of noun phrases commonly occurs also. The noun phrases may be
objects of a finite verb (74) or subjects of such a verb (75), which in this case is the
copula.
mtapota-éÀ.
words-POSN
'You have heard the words of the church elders, the words of the
chief, and the words of the captains.'
One word may be repeated several times to show repetition or continuity of the
action:
Coordinate opposite questions are often indicated by the particle kail which
usually follows the second coordinate member (77). But in other cases, such coordi-
nate questions in the form of adverbial phrases occur without kau (78). In the
examples below the two coordinate phrases are bracketed.
n-0-a-y?
3S-be-SF-UNP
'How do you feel about it, is she one you will give, or is she one you
will not give?' (Said about a potential bride).
kaw-no po-na.
tall-NOMZR on-to
'She makes her nest in a tall tree.'
3 Ellipsis
In Wai Wai ellipsis is quite common in both response and non-response expres-
sions. Note that in the following two response expressions there is no subject or verb
in either.
Cewne kyam.
alone supposition
'Alone, I suppose.'
Anmi-ra.
catch-NEG
'Not catching any.'
The Wai Wai have the word nhnk, meaning 'yes,' but only on rare occasions is
it used alone as a response. Ellipsis of the main verb and other constituents might
be expected after this word, but in Wai Wai it is more likely that the word nhnk will
be elided and the main verb will be spoken either alone or with various other
constituents.
(84) Ml-paka?
2S-wake.up + TP
'Did you wake up?'
K-paka ha re.
IS-wake.up + TP RHY somewhat.
º woke up somewhat?'
(85) Mi-mok-o?
2S-come-TP
'Did you come?'
It is normal for a form of the copula to follow all negativized verb stems. Yet this
form of the copula may be elided from the second clause, in which case that clause
becomes a second adverbial phrase which clarifies the first adverbial phrase, Era wa.
y-eni-hra (nH0-a-xe).
GEN-see-NEG 3S-be-SF-COLL + INP
"That's how some people are, (they) never see daylight.'
When a begging attitude is expressed both the verb and other words may be
elided.
Reflexive, reciprocal and passive meanings may be expressed in Wai Wai by one
prefix to transitive verbs. Since in each use of this prefix the resulting verb form is
intransitive I call the prefix detransitivizer. It has several variants ef-/es-/ec-/ex-/ese-
Wai Wai 45
/e-. I illustrate in sects. 4.1, 4.2, and 5.1 the three different meanings derived from
this prefix. The particular meaning of each occurrence can usually be recovered
from the context. For further discussion of this prefix see sect. 23.2.5.
4.1 Reflexive
Reflexivity of possession is expressed by the prefix ti- to the possessed noun (93,
94). The same prefix indicates reflexivity of possessor of nominalized or adverbial-
ized intransitive verbs (95, 96, 98), and also on transitive verbs (99, 100). The same
prefix also occurs as the object of postpositions (101). In each of these cases the
prefix refers to third person or to the referents of the 1+3 pronoun amna (97). This
reflexive possessor prefix must be coreferential with the subject of the same clause
of which the possessed item is a part, or of the superordinate clause.
n-0-a-y oy-akno.
3S-be-SF-UNP IPOSR-brother
'My brother has strong feelings for what he has made.'
ke ti-im ya.
because REFPOS-father by
"The child was happy because his father helped him.'
ti-wya.
REFPOS-by/to
º heard she wants a canoe in which to cross the river.'
The clitic rma 'same' indicates something of an idea of reflexivity (see sect. 21.2
for other meanings). It has this meaning when referring to persons or objects.
Compare example (101) above with the following example where the subject is first
person.
4.2 Reciprocal
With a few postpositions the detransitivized forms indicate reflexive action, but
in such cases the resulting forms are obligatorily followed by the particle rma.
c-eh-ce-ri
1 + 2S-be-COLL-l + 2ÃÌÑ
'Let's not love ourselves only.'
5 Passives
becomes the subject of the clause and the doer of the action is not in view. Here I
give examples of this prefix with finite verbs (110, 111).
5.2 The adverbializing affixes. The adverbializing affixes t-/c- plus -so/-xi on
transitive verbs accompanied by the copula carry a passive-like meaning. They
indicate that the subject receives the action (114, 115). It apppears that the ergative
characteristics of the language are the source of these adverbializing constructions.
The same adverbializing construction occurs with intransitive verb stems, and the
subject is then the one who performs the action (112, 113).
yu-pun-thM okwe.
3-flesh-PAST alas.
'Its flesh is eaten away somewhat alas.'
5.3 The nominalized form of adverbialized verb steins. The nominalized form of
adverbialized verb stems described in sect. 5.2 also carries the idea of passivity. There
is, however, a semantic difference in the nominalized form compared with the adverbial
form, that is, the reference is always to a future action when it occurs on transitive verb
stems. The nominalizer is the suffix -m/-mu.
Wai Wai 49
5.4 The nominalizing suffix -xapu. The nominalizing suffix -xapu is also a pas-
sive-like form when occurring on transitive verb stems. It describes a referent as
having received the action of the verb stem. It indicates the present perfect tense,
that is, that the results of a past action continue to the time the speaker is speaking
(118, 119). Or it may indicate past perfect tense when followed by the past tense
noun suffix -nho/-nhiri (120). See also sects. 18.2(i) and 23.4.5.1(xi). This suffix
may be followed by the suffix -tho 'DEVALUED' (121).
aka-xapu-nhM rma.
dig.out-NOMZR + PERF-PAST right.away.
'The canoe split right after it had been dug out.'
(121) afma-xapu-tho
throw.away-PERF-DEV
One that has been thrown away'
5.5 -hni Nominalized negation. This suffix is the negative of the perfect tense.
With transitive verbs it indicates that the subject has not received the action up to
the present time (122, 123), or that the subject can never receive the action at all
(124). Again no indication is given about who has not done or cannot do the action.
This suffix is often reduced by morphophonemic processes to -n(i) (123-124). (See
sect. 22.6.)
50 Hawkins
6 Causatives
(i) hi Wai Wai intransitive verbs are made causative by the following suffixes: -re,
-meki, -nmeki, -ka, -nopu. These suffixes transitivize the verb and a direct object is
thus introduced. In the examples below I give the intransitive verb first then the
transitivized form below it for each verb stem.
(b) Ow m-ewre-mek-ya-si.
1PRO 2S-laugh-CAUS-SF-INP
'You make me laugh.'
(b) Ow mi-pana-ta-nrnek-ya-si.
1PRO 2S-ears-VSF-CAUS-SF-INP
'You are teaching me.'
A-pof-mam-nopi-ra 0-x-a-kne.
2O-large-VSF-CAUSE-NEG 3S-be-SF-UP
'He did not raise you.'
(ii) The suffix -porma indicates near negative causation, that is, that something
cannot quite be done or is prevented from being done. It is not commonly heard, but
it occurs with at least one intransitive verb and the resulting form is transitive (130).
This suffix also occurs with at least one transitive verb and with such forms the
negative causative agent is often not expressed (131).
(iii) Caused loss. The suffix -hka/-ka is a transitive verbalizer with nouns. The
meaning is that something or someone has caused the loss of a possession or
cessation of possession of an item. This suffix also occurs with verbs with the
meaning of cessation ofthat action. That function is discussed in sect. 18.2(vi).
52 Hawkins
6.2 With transitive verbs. Transitive verbs take the causative-indicating suffixes
-po and -mexpo. The causee is expressed by the indirect object postposition: ya when
preceded by a noun/pronoun (133b, 134b) or -wya with a person-marking prefix
(135b) (see sect. 17.2).
c-efiepa-ne-m.
ADVZR-steal-ADVZR-NOMZR
"The thief cannot be found out.' (He does not let himself be known.)
7 Comparatives, equatives
7.1 Comparison is often expressed by the relators yopo 'larger than,' 'more than,'
and xawyaka, 'smaller than,' 'less than'. The word yopo expresses comparison in
main clauses (139) and in comparative subordinate clauses (141). The word xawyaka
has only been observed in main clauses (140). The particle group ro maki 'very
much' and the particle nhe 'somewhat' are often used to reinforce the idea of
contrast in such sentences.
7.2 Equatives
(i) Equal size or strength may be expressed by the word ecenari 'equal to'.
(ii) The detransitivizing prefix is often used with the meaning of reciprocity to show
equal size or age, and following this prefixed word the suffix -re 'equal, same' is
often used to reinforce the meaning.
54 Hawkins
(iii) Various adverbs indicating size may be nominalized, then possessed, then
adverbialized again to show equality of size.
(iv) The relator wicaki 'size' is often used to express exactness of size.
(v) The relator vwara/wa/like' is used to express any kind of similarity whatsoever.
The meaning of this relator is often reinforced by various particles: rma, 'same,' xa
marha, 'exactly.'
8 Coordination
9.1 Topicality. There are a few things that I can say about topicality in Wai Wai,
though more study on this subject is needed, hi narrative discourse, once a participant
is introduced and the story continues about that one participant alone, he is not named
again for a considerable space, at times only at the end of the section concerning him.
With verbs the third person subject prefix and the number-indicating suffix seem to
serve to make it clear that the speaker is talking about the same person or topic. With
nouns, the same seems to be clear through the third person possessor prefix and
Wai Wai 55
possession suffix. Also at times, past tense indicators on nouns help to carry the thread
of topicality. But the 3rd person pronoun noro, referring to people or animals, is used
occasionally to make it clear that the participant previously referred to is still in view.
Where there is a postposition referring to a previously identified participant (148)
or a nominalized verb with a similar kind of referent (149), then noro is nearly
always used. The pronoun emphasizes the sameness of the participant.
When adverbs or verbs are fronted the object may be placed after the verb.
At times, when the adverb occurs before the verb, the object of the transitive verb
may even be fronted before the adverb, as it is in the following example, hi this
example, there are two coordinate objects, both of which are left-dislocated in
relation to the main clause. Such dislocated clauses are usually followed by a distinct
pause.
Postpositional phrases are most often at or near the end of the clause (162).
However, for emphasis (163), or for the sake of close relationship to the previous
clause (164), they may occur at the beginning of the clause.
10 Interrogatives
All interrogative expressions are marked by a mid or high intonation on the last
few syllables of the expression while declarative expressions have a falling intona-
tion on the last one or two syllables. Interrogative expressions may also be marked
by interrogative words in expression-initial position or by interrogative particles in
expression-medial or final position.
Responses to questions are most commonly brief one or two word expressions in
which the main word or words of the question are repeated with only personal prefixes
changed where necessary. There are the words nhnk 'yes', andpira, 'no'. The word for
'yes' is rather rare and seems to be used only when the speaker feels tired or uncom-
municative. The word for 'no' is somewhat more common. The most common way to
give a positive response to a question is to repeat the verb or other main word or phrase
of the question. Likewise, the most common way to give a negative response is to
negativize the main word of the question. These positive or negative brief responses are
sometimes followed by a fuller answer to give explanation.
10.1 Greetings and good-byes are usually one-word questions with one-word
responses, which word in each case is a finite verb.
10.2 Question words are ahce 'what,' ati 'which,' ahto 'where,' ahna 'to where,'
ahnixa 'from where,' ahtoxa 'through where,' ahcemaw 'when,' and onoke 'who.'
The word ahto is often nominalized to use in interrogative equative clauses (170).
es-ko.
be-2IMP
'Do not greet anyone you meet.'
10.3 Other questions are commonly expressed with a noun phrase (176) or post-
position phrase (177, 178) of which the noun interrogative ahce is the head word.
The copula is usually included in the question but rarely in the response. The
postposition, however, is never elided in such response expressions.
Interrogative expressions that inquire about something that has been said rarely
begin with ahce 'what' alone, but they begin with the postpositional phrase [ahce
wa] 'how.' The response may be quite long or short. A form of the verb ka/kas-
'say/do' obligatorily follows the quotation (179). The same phrase is used to inquire
how something was done. An explanation is usually the response (180).
10.4 Three particles. There are three particles that indicate interrogation or at least
some desire for a response. Kafi expresses interrogation of the opposite possibility
(181). Ma indicates surprise with a desire for a response (182). Na indicates possi-
bility (183). Sometimes it also indicates a desire for a response. This desire is
reinforced by the interrogative intonation.
a-karita-n?
2POSR-book-POSN
'Possibly you read that in your book, or do you not have a book?'
11.1.1 Second person imperative. See sect. 23.3.1 for a description of the mor-
phological variants of these imperative affixes.
Also in the second person imperative mode, a second person subject prefix occurs
with intransitive verb stems which are consonant-initial. This, however, is not one
of the regular subject indicators that occur with finite verbs. It is a- '2' of the
personal prefixes indicating possession. I will list a number of such verbs in the 2
imperative to display this:
11.1.3 Third person imperative sentences are quite common. The command is
given to the hearer concerning a third person or group. There is no second person
indicator in such commands even though they are made, of course, to a second
person. These commands seem to have as much force as second person commands.
collective if both of them participate in the action. But possibly as much as half the
time it is omitted in such forms.
There is a word, kayka, that means 'Let's go.' Often it has the added meaning
'You go first,' especially when going single file in a trail. But the meaning 'You go
first' is often lost (199). It is quite a common word in Wai Wai. It occurs both in
the singular (200) and collective (201) forms.
11.3 Polite requests are expressed by the particle pair cma re, plus a verb in the
non-past involvement mode.
11.4 The negative imperative is expressed by the negativized verb stem plus the
appropriate second person imperative form of the copula. I have no record of
negative imperative with first or third person.
Wai Wai 65
11.5 Obligation. There is a set of three affixes which occur on any verb stem and
indicate that the speaker feels there is an obligation to do an action or a value in doing
it. These affixes are ßß- + verb stem + -po + -re. The resulting form is an adverb and
may be followed by the copula in the third person singular, or it may be followed by
the verb of hearing (207) or of seeing (208). The adverb may also occur without being
followed by any verb at all (209). The copula is understood in such cases. Indication of
the doer of the action is never a part of this construction. The third person subject prefix
on the accompanying copula does not refer to the one who should do the action. The
same third person prefix occurs whether the one who seems obligated to do the action
is first, second, or third person. The subject prefix on the copula apparently refers to the
action itself, thus: 'It is good that such be done by someone,' or 'Such ought to be done
by someone.' According to my records, no free form subject or object ever occurs as a
part of an obligation clause. The subject or object is clear from the context only. Thus
in some of the examples below I have included the objects of the action in parentheses
(204, 205, 206). I have included two examples which have the context included to show
the object of the adverbialized verb stem (207), (209)
The negation of this expression often requires three separate words: (i) the
negativized verb stem of the action word, followed by (ii) the copula stem with the
obligation-indicating affix set, and (iii) a finite copula form having the third person
subject prefix and a non-collective suffix of time and mode (210, 211). For more on
the morphology of this set of affixes, see sect. 23.5.1.1 (iii).
C-enta-po-re w-enc-e-si.
ADVZR-hear-ought-ADVZR IS-hear-SF-INP
º am glad to hear your words. I think (hear) they are good.'
66 Hawkins
T-afma-po-re kfa?
ADVZR-throw.away-ought-ADVZR worthless
'What shall I do with this? Should it be thrown away?'
12 Negation
(See also sect. 11.4 for the construction of negative imperative expressions.)
12.2 Negation of adverbs and postpositions. There are only a few underived ad-
verbs that occur with the negative suffix -hra (see sect. 23.5.2(i)). A number of
adverbs are not negativized, but the negative meaning is indicated by the use of the
adverb of the contrary meaning (224, 225). A few adverbs and postpositions that
indicate location in space are negativized by the negative of the copula which
follows them (226, 227).
12.3 Negation of noun possession. Nouns which occur with possession indicators
other than -n are negativized by the suffix -hra to indicate negation of possession.
The suffix -m replaces the h of the suffix with noun stems which take the -n positive
possession suffix (231, 232), on noun stems which indicate categories of possession
(233, 234), and on the word kanawa 'canoe'(235). I assume that the -m is an
alternate form of possession indicator occurring only in these negative constructions.
(See sect. 23.4.5. l(x).) The suffix -hra occurs alone on all other nouns which take
possession suffixes. Again the resulting forms are adverbs and are usually followed
by the copula or verbs of motion. See sect. 22.6.5. for loss of /h/ from the negative
suffix.
The suffix -hni also occurs with nouns that may be possessed to indicate that the
referent is one who does not possess the item. The resulting form remains a noun.
It is a correlative of the suffixes discussed above in that the nouns which take the
suffix -m possession indicator before -ra also take the -m before suffix -rii. See
sects. 22.6.1.(i) and 22.6.5 for the morphophonemic changes in this suffix.
12.4 Negation of noun identity. The word pini is a negative of nouns which indi-
cates negation of identity. Further discussion of pini as a word is given in sect.
15.3.1.
12.5 The suffix -hto. There is a nominalizing suffix hto which indicates negation,
usually with strong feeling or a sense of permanent negation. This suffix occurs
mostly with the copula in postpositional phrases of cause (241). The verb moku 'to
come' has been heard on occasions with this same suffix (242). This suffix is always
followed by the possession suffix -ri. See also sect. 15.4(iii).
moku-hto-ri ke.
come-NEG + NOMZR-POSN because
'We cannot go because the plane is not coming.'
13 Anaphora
Anaphora are strategies in the language for referencing individuals and items
between phrases, clauses and sentences.
13.1 Three nondeictic third person pronouns occur in finite verb clauses to refer
back to something or someone mentioned earlier: ero 'it/that', noro 'he,' 'she', and
Wai Wai 71
nexamro 'them. ' Any of these pronouns may be used after a finite verb clause to
more clearly identify an item or action previously mentioned (243, 244). The phrase
ero yimaw refers to a time previously mentioned. It follows an adverbialized verb
stem indicating time (245).
The postposition phrase ero wara normally occurs at the end of a list of items
(246). It also often occurs at the end of a paragraph or discourse and refers to all
that has been said in the paragraph or discourse.
y-xe
3POSR-wanting IS-be-SF-INP
¢ knife, a spoon, a mirror, some fishhooks, fishline, that's what I
want. '
Any of these same pronouns are normally used with postpositions (247). Postpo-
sitions very rarely occur without an overt object either prefix or free form. If it is
not necessary to rename a participant with a postposition then noro or ero are used
as the object. The same pronouns occur with certain particles which indicate distinct -
iveness of the participant or an element of surprise concerning him or it (248). They
72 Hawkins
may also be used as direct objects and possessors (249) or as subjects of verbs (250),
and again the idea of distinctiveness of the referent stands out in most cases. The
antecedents in these examples are all in the first sentence, while the pronouns are in
the following sentences. This usage is normal.
(248) Rikomo mak mexe mikro. Noro yipu rma 0-tom-o esama yaw.
child just DISAG 3PRO 3PRO such REF 3S-go-TP path in
'He is just a child. Even so he went on the trail.'
y-akro-no-ma-xi ki-wc-e-s.
GEN-with-NOMZR-VBZR-PM IS-go-SF-INP
'Uncle is not well. All right, I will go help him.'
13.2 Deictic pronouns. There are three deictic pronouns referring to nearby ob-
jects, on, tan, moso. The first two refer to inanimate objects and the last refers
usually to people. However, in recent years moso is being used increasingly for
inanimate objects also. There are three deictic pronouns referring to distant objects
or persons miki, mikro, and mini. The first two refer to animate objects and the last
to inanimate objects. There are two collective deictic pronouns, moxam for near
animate object and mikyam for distant animate objects. There are no collective
deictic pronouns for inanimate objects. On may occur as the one-word object of a
postposition (251), or part of a compound object of a postposition (252), or the
object (253), or subject of a finite verb (254). It also may occur as the subject of an
equative clause (255).
Wai Wai 73
The deictic pronouns triik, mikro, and mikyam are almost never used anaphorically
as the subject or object of a finite verb. With the anaphoric meaning they are used
almost solely as the subject of equative clauses. When the pronoun is in the subject
position (after the predicate) the clause indicates identity (263b) or gives description
(259-261). When the pronoun occurs in the predicate position (before the subject)
in an equative clause its meaning is purely deictic (263a). However, the word mikro
may occur alone as an emotional expression, almost the same as an interjection
(262).
74 Hawkins
(262) Mikro.
3PRO
"There he is!'
13.3 The verb ka/kas 'say/do' is often used to refer back to a previously men-
tioned action.
13.4 Third person possessor prefixes, which also have an anaphoric function, are:
0-, yi-, i-, and ti-. 0- occurs when there is no free form possessor, regularly with
stem-initial verbs and, in some idiolects, with stems which begin with an unclustered
relaxed consonant ((265); and see sects. 22.2 and 22.3 for relaxed consonants and
consonant clusters.). Otherwise, one of the overt forms of the prefix occurs (266).
See sect. 23.4.5.1 for further details and examples.
Third person subject prefixes are also anaphoric. They always refer to a partici-
pant, place, or condition previously named (267). The third person reflexive personal
prefix is even more clearly anaphoric or cataphoric in that it refers to the third
person subject of the associated finite verb (268), or the subject of the associated
verb stem that has been derived to another word class (269). For other examples of
third person reflexive possessor see sect. 4.1.
ke.
because
'He did not go becase he was weak/sick.'
A third person prefix occurs with postpositions when there is no free form object.
This prefix may be zero (270) or yi- '3' (271). It is easily seen that such prefixes
refer to a previously named person or object.
13.5 Athematic focus. One of the three meanings of the particle ha (see sect.
21.2(ii)) is athematic focus, that is, it emphasizes or brings into special focus an item
which is not a part of the action or main theme of the discourse. This emphasized
part refers back to a previous word or phrase and explains or amplifies it. It does
not carry forward the action of the story (272). It should also be noted that ha with
this meaning always occurs at the end of the word group to which it refers. Thus
76 Hawkins
when any dangling phrase ends with this ha it means that that phrase refers to
something that precedes it rather than what follows. Thus it acts something like a
period or comma (273). With this meaning ha is not accented phonetically as it is
with its other meanings. See sects. 21.2(iii) and 21.3(ii) for the other two meanings
of this particle.
14 Subordinate clauses
o-wya].
1-by
º got a machete with which to underbrush a field.'
Wai Wai 77
mikyam.
3PRO + COLL
"They are more than can be numbered.'
At times verb stems are adverbialized first and then nominalized to form depend-
ent clauses, as in the following examples:
There are two other constructions which are built on verb stems to produce
adverbials indicating time or condition of an action. One of these is the suffix -ehe
'after', which is added to the otherwise unsuffixed verb stem. It mainly indicates that
the action of the verb stem to which it is affixed is prior to the action of the main
verb of the clause. It also indicates conditional action, but not so commonly as it
indicates prior action. The suffix -taw 'if/while' is added to the verb stem plus the
stem formative suffix, except with the copula. With the copula it is added to the
unsuffixed stem (289). The resulting form indicates that an action is contemporane-
ous with another action. It may also indicate conditional action.
The forms presented here are similar to postpositions, but I classify them as
adverbials since they show some difference from postpositions. When the verb stem
of derived adverbials is intransitive the prefix is the underlying subject of the action.
When the verb stem is transitive the same prefix is the underlying object of the
action, and the subject, if overt, occurs as a prefix to the postposition wya 'to/by'
(288) and (291), or as a free form noun followed by ya 'to/by' (292).
tawake w-0-a-si.
glad IS-be-SF-INP
'If the children sweep the village (grounds) I will be glad.'
w-enta-y.
IS-hear-IP
'Just as I came away I heard of his death.'
In the following examples the adverbial clauses do not modify the verb directly,
but are expansions of the adverb complements 'many,'and 'into it (water).' It is the
adverbial counterpart of the discontinuous modifying nominal construction. See sect.
15.3.2.
pin kaxi].
not.one because.of.being
"The dog does not say, "Give me meat," because he is not a person.'
14.3 Clauses of identification commonly occur in Wai Wai. These are subordinate
to the preceding main clause. Both the main clause and the subordinate clause have
finite verbs. But the subordinate clause functions as a modifier to a constituent of
the preceding main clause. See also sect. 13.1.
wii-k-e-kne] ero.
IS-say-SF-UP that
'Listen carefully to my words, the thing I said to the chief
The noun is defined as a word which may occur as the subject or direct object of
a verb, as the head of a postposition phrase or as the nucleus of the predicate of an
equative clause. A pronoun is a word that represents a noun, and thus pronouns are
included with the discussion of noun phrases. A noun phrase usually consists of one
or two nouns, or one noun and one pronoun. Such phrases commonly have one or
more particles. It is also quite common to have discontinuous noun phrases separated
by a finite verb, or in equative clauses they may be separated by the subject of the
clause, which is the second member of the clause, normally a pronoun. In such cases
the second member of the noun phrase is a modifying noun. It may further identify
the subject or modify the predicate.
15.1 Marking for case. Case markers indicate the relation of pronouns or nouns to
verbs. Free form subjects and objects of verbs are not morphologically marked. Also
word order does not seem to be a good criterion for determining subjects and objects
of verbs, since subjects seem to vary freely between pre-verb and post-verb position.
Direct objects also vary occasionally in the same way, though the more normal
position for these is before the verb.
Wai Wai 81
But case is indicated in several of the personal prefixes to verbs. There are two
sets of such prefixes, subject and object prefixes. Subject prefixes occur on finite
verbs (299-300). They always occur on such verbs even with a free form subject
(301). However, third person subject prefixes are 0- in some constructions (304).
Object prefixes on finite verbs exist for first and second persons only (302-303).
There are no overt forms for third person object prefixes to finite verbs (sect.
23.3.3). See also sect. 23.3.2 for the complete set of subject prefixes.
The indirect object postposition ya 'to', alternating with the bound form -wya
(sect. 17.2), may also indicate the causee of a transitive verb (305, 309—310) or the
addressee of any 'say' expression (306). It also occurs with the subject of nominal-
ized (307) or adverbialized (308) transitive verbs. This reflects the ergatively organ-
ized character of these derived forms. The free form of the postposition occurs when
the object is a free form (307). When no free form occurs, a prefix is attached to the
postposition (308).
komo ya.
COLL to/by
º hear the field cutters yelping.'
w-ek-ya-si.
IS-bring-SF-INP
'If I shoot a tapir I will bring you some meat.'
o-woxi-n ya.
lPOSR-in.law-POSN to/by
º will get my in-law to make a canoe for me.'
1. Forms derived from a transitive verb and preceded by a free form direct object
of the verb stem.
2. Forms derived from an intransitive verb and preceded by a free form subject.
3. A noun and its free form possessor.
4. A postposition and its free form object.
There is no pause whatever in normal speech between the two items of these
constructions. The order of such phrases is obligatory. The object, or subject, or
possessor is first, followed directly by the head word of the phrase. Only the
particles komo, 'collective' or ro, 'time' occur between the two words and that is
rare. I call this relationship Genitive.
Wai Wai 83
If the head word begins with a vowel the prefix y- 'GEN' is always added to it
(311, 312, 314a). If the head word begins with the third person prefix yi-, ti- or i-,
that prefix usually remains in the genitive relationship (313). The exceptions are
three verb stems which occur with the third person prefix ti- and begin with lol.
When the third person prefix is omitted from these verbs in the genitive relationship
the genitive prefix y- occurs (314 a, b.). See sect. 23.4.5.1 for a list of the ten verbs
that take the ti- prefix.
(i) Forms derived from a transitive verb stem and the genitive relationship to their
underlying direct objects are illustrated below:
(ii) Forms derived from an intransitive verb stem and the genitive relationship with
the underlying subject are illustrated below:
cf. eyehtopo
'his bathtub'
cf. yi-mtapo-ta-cho
3POSR-talk-VSF-CIRC
'his words'
(iii) Noun phrases of possession show the same close connection between the two
nouns. Here again we see the prefix y- appearing when the possessed noun stem
begins with a vowel (317). When it begins with a consonant, no prefix is attached
(318 a, b).
(b) yi-krapa-n
3-bow-POSN
'his bow'
(iv) Postpositional phrases show the same close connection with their objects.
15.3 Phrases of noun modifying noun. Again I include pronouns along with
nouns. There are no adjectives in Wai Wai. Only nouns and a few pronouns modify
nouns.
15.3.1 Continuous noun phrases. There are a few pronouns that occur immedi-
ately before nouns without intervening pause or word. The pronoun points out or
distinguishes the noun from other items (321, 322). These noun phrases are brack-
eted below.
Wai Wai 85
I postulate ñßçß 'NEG' as a noun since it may be separated from its head noun by
either of the particles komo (326), maki (327), or ro (328). Also it does not undergo
morphophonemic reduction to the form -hni as it would if it were a suffix But it
always has a noun or pronoun as its head word, and thus constitutes a continuous
noun or pronoun phrase. It indicates negation of the identity of the object or person.
It occurs commonly in equative clauses (323) and in pronoun phrases. In such
phrases it may occur as the subject of any verb (324) or as the object of a transitive
verb (325). See sect. 12.4 for more examples of this word.
pin kaxi.
NEG because.of.being
º don't like it because I'm not used to eating it.'
86 Hawkins
15.3.2 Discontinuous noun phrases. Nouns may also modify other nouns or noun
phrases in a discontinuous sequence. In clauses having finite verbs the head noun
and the modifying noun are often in a discontinuous sequence, being separated by
the verb, and possibly also by postpositional phrases or adverbs related to the verb.
Pause (indicated by comma) may also break the continuity of a noun phrase.
An example of such modifying nouns in a discontinuous sequence is seen in
possession indicators of certain nouns. With such nouns possession is not indicated
by affixes to the name of the item, but by generic nouns which name categories of
possession and take possessor prefixes. These generic nouns are most often in a
discontinuous sequence. Possession of domestic animals, items of food, drink and
shooting arms, is expressed in this way. The two nouns may occur in either order.
Nouns of explanation, expansion, or limitation are common in Wai Wai and are
used to modify either a subject or object in verbal clauses. Often these nouns are
derived from verb stems or adverb stems. At times verb stems are adverbialized and
then nominalized. Such derivational processes are fully described in sects. 15.4, 23.4
and 23.5, but I give some examples here. The modifying nouns may be juxtaposed
either after the verb (332) or before the verb (333). When the nouns are adjacent to
each other they are separated by pause, as indicated by the commas in the same two
examples. The modifying noun may modify the subject of a clause when the subject
follows the verb (332) or when it precedes the verb (334). Or it may modify the
object of a transitive verb (335). The modifying noun may function as a relative
clause (336). There may be 2 modifying nouns modifying the verb object (337). The
modifying noun may modify the possessor of a noun phrase (338). Occasionally the
normal modifying noun is dislocated to a position before the verb and even before
the interrogative subject (339). In this same example a quotative clause functions as
a second modifying nominal which modifies the first nominalized clause. Wai Wai
speakers may pile up modifying nouns quite a bit. Usually each one modifies the
first nominal, which is the head of the phrase.
oy-eh-topo-nho w-ahsi-pink-e-tik-e-si
IPOSR-be-NOMZR + CIRC-PAST 1 S-hold-REV-SF-finish-SF-INP
komo o-wya.
COLL 1-to
'Don't give this sort of your domestic animals to me.'
k-wakre-ne-nhM rma.
1 + 2-be.kind.to-AG-PAST even
'We despise even good people, even those who have been kind to us.'
k-poyino-nhM komo.
1 + 2POSR-fellow.tribesman-PAST COLL
'We heard of the death of some people, our own tribespeople.'
ka-cho-nho?
say-NOMZR + CIRC-PAST
'What is there that he cannot do, about which he says, Alas, apparently
I don't know how to do it, let it be?'
The pronouns ero or noro occur as the predicate of some equative clauses. Thus the
clause has two pronouns as its nuclei. Such a clause always refers back to the action of
the previous sentence. In such a construction a nominalized verb or modifier often
occurs after the pronoun subject and modifies the initial pronoun ero or noro.
wMhyaka-cho.
despise-NOMZR+ CIRC
'They say this,.. . That talk is despiteful of the chief
15.4 Nominalizations. Following I list all the suffixes to verb stems that occur to
indicate that the resulting forms are nouns. I assume that these suffixes indicate
nominalization, though all of them have an additional meaning. I include notes about
their occurrence with other classes of words. For the phonological variation of some
of these affixes see sect. 22.6.
action (sect. 23.4.5.1). If the verb stem is transitive the personal prefix refers to the
object of the action; with intransitive verb stems it refers to the subject of the action.
With transitive verbs the subject or doer of the action, when expressed, is expressed
by the ergative form, consisting of a personal prefix plus the postpositional suffix
-wya, or by a free form plus the postposition ya. In any of the above situations,
where the personal prefix can occur it may be replaced by a free form. When a free
form object of the action occurs before a transitive verb stem beginning with a
vowel the verb stem takes the genitive prefix y- (351). Forms having this suffix -ri
may occur as the direct object of the verbs 'to hear'(342) and 'to see'(343). They
occur most often as the object of postpositions as in (344) through (351). They also
rarely occur in an expression without a finite verb as something of an exclamation
(352).
ke yi-wya.
because 3-to/by
'We came fast because he helped us.'
(346) o-mok-ii ke
l-come-NOMZR+POSN because
'because of my coming.'
(350) 0-ah-ii ro ya
3-eat-NOMZR + POSN often to/because
'because of eating it often'
(ii) -ni- . . . -n/-tho/-thiri Object resulting from an action. This set of affixes occurs
only with transitive verbs and the resulting form refers to an item which receives the
action indicated by the verb stem. This is a possessed form which occurs with a
personal possession prefix or a free form possessor that refers to the underlying
subject of the action. The possession does not refer to the resulting item but to the
action done to that item. Forms with these affixes seem to occur in any syntactic
function where other nouns occur, but their most common use is as a modifier of
another noun, functioning like a relative clause (354, 355). In the examples below
they occur as the object of a postposition (353) and as modifiers of the predicate
nominal of an equative clause (354), and of the object of a transitive verb (355). The
suffix -Þ indicates present or future possession of the action (353). The variant forms
-tho/-thM indicate past tense possession and alternate as follows: -tho occurs when
the form has a first person singular prefix, and also when the third person possessor
of the action is indicated by a free form (355). The suffix -thiri occurs elsewhere
(354).
cf. Ah-ko.
(eat-2EVIP)
'Eat it.'
cf. Afma-ki.
(throw/drop-2IMP)
'Throw it.'
cf. wo-hra
(shoot-NEG)
'not shooting'
(iii) -hto Negation of the action. This suffix is rarely used except with the two verb
stems given below. I have no examples of it occurring with a transitive verb. It
nominalizes and negativizes the verb stem and is obligatorily followed by the
possession suffix -ri. It occurs with free form possessor indicators or with personal
prefix markers of possessors. Tense is not indicated by this suffix. See also sect.
12.5.
(iv) -topo/-cho Circumstances of the action. There is free variation of the two forms
of this suffix with most verb stems. The meaning of the suffix is very broad. It
includes instrument of the action (358, 360), place of the action (359), time of the
action (361), story of the action (362), and instructions about the action (363). Just
like suffix (i) above, this suffix co-occurs with personal possessor prefixes with
intransitive verb stems to indicate the subject of the verb stem (362) and with
transitive verb stems to indicate the object (363). It also occurs preceded by a free
form noun or pronoun and without a personal prefix to indicate the same possessor
functions. The genitive indicator y- also occurs with the same paramaters as it does
with the suffix -ri (361). The suffix -nhof-nhiri indicating past tense may occur
following this suffix (362).
y-ama-cho o-wya.
GEN-cut.down-NOMZR + CIRC 1-to/by
º want an axe with which to cut down a field.'
(v) -xapu Object that has received or performed action in the past with enduring
results. I label this suffix perfect aspect indicator (PERF). See sects. 5.4 and 18.2(i)
and (ii) for full discussion and examples of this suffix.
(vi) -yem Associate. This suffix occurs with the possessor prefixes and can occur
with the detransitivizing prefix (365, 368). It also occurs with two generic nouns
woosi (366) and kiin (367) to indicate an associate of the same sex. This suffix also
occurs with the past tense suffix -nho/-nhiri (368).
(365) aw-es-emani-yem
2POSR-DETRANS-play-NOMZR+ASS
'your playmate'
Wai Wai 93
(368) aw-es-emam-yemi-nhM
2POSR-DETRANS-play-NOMZR+ASS-PAST
'your former playmate'
(vii) -tamci Payment for work. This suffix is somewhat rare. It occurs only with
transitive verbs (369) or with the copula (370). The combination of verb stem plus
the suffix -topo ((iv) above) following the word epetho/epethiri 'payment' is more
common (371).
(viii) -xan Subject of recent action. This suffix is only used with a few verbs as far
as my records show.
(372) ewru-xan
be.born-NOMZR+ RECENT
'a newborn baby'
(ix) -ne Doer/agent of the action. This suffix occurs only with transitive verb stems.
It co-occurs with the personal prefixes which indicate the possessor as the
underlying object of the action (375). When a verb having this suffix is preceded by
a free form possessor, there is usually no prefix on the nominalized verb form (376).
But when one of the ten verbs having the prefix if- occurs with a free form possessor
this prefix occurs (377). When the free form possessor immediately precedes a
nominalized verb stem beginning with a vowel the verb stem takes the y- genitive
prefix (378). When the nominalized form occurs with the prefix yi-, this prefix may
either remain or be elided, if it is immediately preceded by a free form object (379
a,b). This suffix -ne also may be followed by the past tense suffix -nho/-nhiri to
indicate past action, that is, action which in the mind of the speaker occurred long
before or which has little effect in the present (380). These forms not only indicate
one who does the action, but they often carry the meaning of ability to do the action
(381).
(x) -no Unspecified personal doers or receivers of action. This suffix with
intransitive verbs indicates the underlying subject of the action (382), and with
transitive verbs indicates the underlying object of the action (383). It usually is used
of groups of people. Nominalized verbs with this suffix are nearly always followed
by a form of the verb ciri 'make.'
(xi) -hni Nominalized negation. This suffix again displays a difference in meaning
when occurring with intransitive verbs and with transitive verbs. In the former case
it means one who does not do the action, hi the latter case it means the one who
does not receive the action. This form occurs in any of the five clause types
presented in sect. 1. For the morphophonemic changes see sects. 22.6.1(i), and
22.6.5.
(xii) -Si Nominalization before the postposition xe. The verb stem has no overt form
of suffix in this construction. But the word xe functions as a postposition and thus
must be preceded by a noun or pronoun. Thus I postulate -j^nominalizing suffix. The
phrase including xe is collectivized, when appropriate, with the collective indicator
word so. The phrase is almost always followed by the copula. The verb stem also
occurs with personal possessor prefixes. If the verb stem is intransitive the prefix
refers to the one to do the desired action. If the verb stem is transitive the prefix
refers to the one to receive the desired action. The verb stem may occur with the
detransitivizing prefix also (391). The same construction occurs before the
nominalized form of xe, which is xati (392).
0-moh-cow-pe
3-come-COLL-3IMP
'Let only those come who want to come.'
For the derivation of nouns from adverbs and postpositions, see sects. 23.4.3 and
23.4.4.
16 Pronouns
16.1 Free pronoun forms. There is variation in pronouns to indicate number, and
animate and inanimate categories.
Wai Wai 97
ANIMATE INANIMATE
Person Non-collective Collective
1 owi
1+2 kiiwi Mwyam
1+3 amna
2 amoro amyamro
ANAPHORIC PRONOUNS
DEICTIC PRONOUNS
(394) Ow xa wi-tw-e-si.
1PRO DEF IS-shoot-SF-MP
º am the one who will shoot him.' (Others had tried to shoot the bird
and had missed.)
The 1-1-2 pronoun has a non-collective form kuwi and a collective form kswyam. The
non-collective form indicates that only the speaker and hearer are involved in an action
(395). The collective form indicates that there is more than one hearer (397). The non-
collective form may also function as an indefinite pronoun meaning, 'people in general,'
which, of course, would involve the speaker and the hearer(s) (396).
k-epirka-che.
1+2-fall-after
¢ person does not rise again after falling from very high.'
tan-to-no komo?
here-at-NOMZR COLL
'What shall all we who live here do?'
The pronoun amna indicates that the speaker and one or more persons other than
the hearer are involved in an action. This pronoun has no collective form. It may
occur with a transitive verb having a collective suffix, but only when amna functions
as the object and the subject is collective. Possibly the component of collectivity is
ruled out from use with this pronoun because the hearer is excluded by definition.
As the subject of finite verbs this pronoun most commonly occurs before the verb
(398) but may occur after it also (399). As indicator of the possessor of an object it
always occurs before the possessed noun (400).
The pronouns noro and ero refer back to a previously identified person or thing,
usually to one identified in a previous sentence (401, 402). They are usually non-de-
ictic (401, 402), but when the speaker is pointing out something very excitedly he
may use these pronouns (403). But even in such usage it is likely that the speaker is
referring to a person or animal which is very much in the mind of both speaker and
hearer because of earlier emphatic reference. These two pronouns are very common
in Wai Wai.
Wai Wai 99
The pronouns of the deictic list are also used as the subject or predicate of
equative clauses. In most such clauses the predicate occurs before the subject, which
is a pronoun (see sect. 2.2 for examples). When the order is reversed and the
pronoun is the predicate of the equative clause, it has an emphatic deictic value
(404-^05).
When the deictic pronouns are used as the subject of equative clauses, they also
indicate that a distinction is made between one near the speaker and one near the
listener or far from both. But the main idea of such clauses is emphatic description
of the subject.
There is an increasing mixing of the 2 near deictic pronouns moso and orii. Moso
is used quite often by some speakers for both animate and inanimate. But the norm
still seems to be the distinction given in the table above.
Oni and tan have the same meaning and in some cases may be used interchange-
ably. Oni is used more for tangible objects, and tan, besides being used for such
100 Hawkins
objects, is also used in the introduction of discourses to refer to the subject of the
discourse (409). Occasionally in formal speech both of them are used together (410).
Tan, besides being a pronoun, is also a locative adverbial meaning 'here.'
(408) On ha a-nah-ri.
3PRO IRES 2POSR-food-POSN
"This is your food.'
More study is needed on the difference between mini and moro. It is possible that the
former is used more often for objects (411) and the latter more often for concepts (412).
16.2 There are three question words which function like pronouns and thus may
be considered to be pronouns. They are onoke 'who?', ahce 'what?', and ati 'which?'
Question words usually occur as the first word of the sentence or clause of which
they are a part. They also occur as a complete response expression. They occur as
the predicate of equative clauses (413), as the object of postposition phrases (417),
as the object of transitive verbs (414), and as the subject of transitive verbs (415)
and intransitive verbs (416).
16.3 There are also pronoun phrases which function as indefinite pronouns.
There are two types of these phrases. These are illustrated below and each such
phrase is bracketed. The first type includes a question word and the particle na
'POTENTIAL' plus at times the particle so 'COLL' and/or a postposition. The
meanings of these phrases are 'anyone/whoever,' 'anything/whatever,' or 'any-
where/wherever,' and at least one of these constructions contains a negative word
which results in the meaning 'nothing' (422). Such an indefinite phrase can also
indicate uncertainty or lack of knowledge (418), (419), and (423). I bracket these
phrases to show that the separate words do not have separate meanings.
The suffix -no also indicates unspecified personal doers or receivers of action.. It
is discussed in sect. 15.4(x). The suffix -nano indicates the unspecified possessor of
a few nouns. It is discussed in sect. 23.4.5.1(xii).
16.4 Honorific. There are no gender or class distinctions indicated by Wai Wai
pronouns. There is one relationship distinction practised by some of the older people,
that is, men address their male relatives by marriage using the collective forms even
when they are speaking to one person. It is, apparently, an honorific device bom of
fear that the man's in-laws might take away his wife if he didn't treat them in this
special way.
16.5 Pronominal prefixes. Subject, object and possessor prefixes are described in
sects. 23.3.2; 23.3.3; and 23.4.5.1.
komo 'collective' and ro 'permanently/much' occur between the free form noun and
its postposition.
There are simple postpositions and those receiving suffixes to indicate static time
or place or motion to, from, or by a place. The second set in the table above
(yaw/hyaw 'in, with') has alternating stems. The first stems occur when there is a
free form object, and the second stems occur when there is a personal prefix. Most
of the forms listed here take any of the possessor prefixes, but the third set kwaw,
'in a liquid' only occurs with the third person non-reflexive prefixes. Dashes indicate
we have no record of a postposition indicating that meaning.
In the following set (428) the postposition yaw has an idiomatic meaning when
the free form noun or pronoun or when the prefix refers to a person. It means that
the one person is living with the other person.
(429) Tuuna kwaw nas oponon. 'My clothes are in the water.'
Yukwaka wafma. º threw it into the water.'
Kati kwaka cirko. 'Put it in the oil.'
104 Hawkins
The to set has several variations with complete stem changes. The form to 'by,
at' does not take personal prefixes. It does take the prefix ii- which serves to refer
to the place where a second or a third person is (iito 'in that place where you are or
he/it is.') The stem-changed form na 'to' takes the first person prefix oo- (pona 'to
where I am'), as well as the prefix ii- (Una 'to where you are or he/it is'). Neither
of these postpositions takes the second person prefix or the 1+2 person prefix. Also
they are never preceded by the 1 + 3 pronoun amna. The forms indicating motion
from a place take no prefix. Rather the stem changes to nixa when it occurs with a
free form noun object. The stem also changes to enexa when it occurs with no free
form object. It means 'from that same place,' and I consider this an adverb (sect.
20).
There are other postpositions which relate to place and space but which do not
take motion-indicating suffixes.
17.2 Non-space/time indicating postpositions. There are also words that function
as postpositions which indicate nothing of position in space or time at all. All of the
following postpositions may occur with the negative suffix -hra except (446) ya/
-wya. They all may occur with nominalizing suffixes also, these are quite irregular.
This form I usually write as a separate word because it occurs with prefixes (ii)
& (iii) and suffixes (iv) & (v). Where it is used very commonly with certain words
I have been in the habit of writing it as a suffix to those words (vi).
(442) poko This postposition has a positional meaning 'stuck to' (i), or
'holding on to'(ii). But it also has psychological meanings (iii),
(iv), (v).
(451) Yawaka xe wasi marari yamacho owya. º want an axe with which I
can cut a field (with which a
field may be cut by me).'
Wai Wai 107
17.3 Postpositions formed from nouns. There are suffixes added to noun stems to
derive postpositions and to give a more specific indication of place. Most of the
noun stems indicate parts of the body. For some of the stems no noun is known
(456), but they have the form of postpositions formed from nouns. Thus I assume
that in years past there were nouns that are no longer used, while the postpositions
formed from them are still used. See sect. 23.6.1.1 for fuller details of the morphol-
ogy.
There are a few nouns that take different suffixes to derive postpositions, hi the set
below these suffixes are in italics.
108 Hawkins
Postpositions also occur with the same prefix that we label as detransitivizer when
it occurs with verbs. For the allomorphs of this prefix see sect. 23.2.5. It cannot, of
course, function as detransitivizer with postpositions since they do not refer to
actions at all. Its meaning with postpositions is clearly reciprocal. When this prefix
occurs it may have a group of people as referent. Yet it never occurs with the
collective indicator so.
Postposition phrases may also include various particles. I have not made a study
of all the possible positions of the particles, but I list a few below. The particles or
particle phrases are bracketed.
At least one particle has been observed to occur between a postposition and the
collectivizer of its object, that is, the particle roro.
18.1 Tense, number and emotional involvement. Tense and number marked on
verbs are necessarily presented together as each suffix indicates both tense and
number. Also, tense and emotional involvement of the speaker are closely related in
Wai Wai. They are marked by suffixes. These suffixes are not affected by which
personal prefix occurs with the verb. I here present these two categories together, as
well as the component of number, which is marked by suffixes or by a particle. The
complete paradigm sets for number are listed in sect. 18.4.2.
18.1.1.1 Today past tense. The today past tense is used for any action that has
occurred on the day the speaker is speaking. It is quite rigid in its use. It does not
depend on the attitude of the speaker.
(473) M-aama-0?
2S-cut.down-TP + NCOLL
'Did you cut it down?'
(476) K-emahci-w
IS-run-TP+NCOLL
º ran.'
(478) M-ama-cow?
2S-cut.down-TP + COLL
'Did you all cut it down?'
Involved/uninvolved aspect of tense. The form of the other tenses, both past and
non-past, depends on the attitude of the speaker, that is, how emotionally involved
he is with what he is saying. Thus we label two categories, involved aspect and
uninvolved aspect. We give examples of them here along with the tenses associated
with them. The involved and uninvolved past tenses refer to any time in the past
prior to the day of the utterance.
18.1.1.3 Uninvolved past tense. There is a stem formative suffix which has four
allomorphs, -e/-ya/-wa/-a, and occurs in the uninvolved past tense, the uninvolved
and the involved non-past tenses, and in the adverbialized verb form indicating
conditional or contemporaneous action. For a further description of the alternation
of these allomorphs see sect. 23.2.1. The allomorphs -wa and -a have the meaning
of verbalization of nouns that results in verb stem formation. There are other suffixes
which have the same two meanings of verbalization of nouns plus verb stem
formation. (See sect. 18.2 (iv).) To distinguish between these two sets of allomorphs
I label the first set (-e/-ya) stem formative (SF). The other set of allomorphs (-wa/-a)
I label verbalizing stem formative (VSF). The variation in tense and number is
indicated by other suffixes.
18.1.2 Nonpast tenses. In Wai Wai, action performed in the present is marked
exactly the same way as action that is to be performed in the future. There is no
difference in the verb. Therefore, I label these forms as one tense and call them
nonpast tense (after Derbyshire 1985:96, who labels these forms as such for the
Hixkaryana language, which is related to Wai Wai.) Future action is often distin-
Wai Wai 113
guished from present action by future time words, such as amne 'later/ Both the
involved non-past suffixes and the uninvolved non-past suffixes always follow one
of the allomorphs of the verb stem formative suffix mentioned above.
18.1.2.2 Uninvolved nonpast tense. The uninvolved nonpast tense is always used
for questions in present/future time, and in phrases containing the particles indicating
speaker's attitude: ti, mi or na (see sect. 21.3). It is also commonly used when
speaking of distant places or persons. The suffixes indicating this tense always occur
following the stem formative suffix and carry the components of uninvolved nonpast
tense and number.
(i) Non-collective: -y/-0This tense indicator is -j^with every verb except the copula,
in which case it is -y.
18.2 Aspect
(i) The present perfect aspect of verbs indicates that an action occurred in the past
with effects that have endured. This aspect is indicated by the suffix -xapu. This
suffix nominalizes the verb stem. The resulting form is a non-possessed
nominalization which does not have a personal prefix. With transitive verb stems it
is the underlying receiver of the action that is the focus of the nominalization (496),
and with intransitive stems the underlying doer of the action is in focus (497). Thus
it is another construction that reflects the ergative characteristic of Wai Wai
nominalizations. This nominalized form often modifies the subject or object of the
verb, as in example (496). See sect. 5.4 for further discussion and examples of this
suffix.
(ii) The past perfect aspect of verbs indicates that the action occurred in the past with
effects that continued for a while but then ceased. This aspect is indicated by the
same suffix -xapu plus the suffix -nhM. See sect. 23.4.5. l(xi) for -nhiri 'PAST'.
ti-nah-ii y-apon-so.
REFPOS-food-POSN GEN-ask.for-PM
"This one comes repeatedly to ask for food.'
(iv) Ingressive action is indicated by the suffixes -\va, -ta, -mam, -pam, which are
described in sect.s 23.2.1 and 23.2.2 as verb stem formative suffixes which derive
verbs from nouns. The resulting forms are intransitive verbs. One or the other of
these suffixes remains in place throughout all tenses and derivations. The
meaning is 'to become an owner of,' 'to become a producer of or 'to experience
any change in circumstances.' The particle taki 'change of state' sometimes
accompanies such verbs and adds to the idea of ingression or change of state
(502).
(v) The verb stem -tihka may mean 'finish,' 'increasingly' or 'very much.' It occurs
alone as a transitive verb (506) or it may be compounded with another verb stem
(507). The compounded verb is either transitive (507) or intransitive (508),
according to the transitiveness or intransitiveness of the verb stem to which -tihka is
added.
Note: The c ' VSF' in (508) is what remains of the suffix -ta after the a is replaced
by the stem formative suffix -e, the lit being replaced by Id according to the
palatalization rule (sect. 22.6.3).
(vii) Loss. The suffix -ka occurs with nouns to derive verb stems, and it indicates
the loss of an inner state (514), loss of an item (515), or loss of a food (516) or
animal (517). It may be more exact to say that it indicates cessation of possession
of the item. The resulting forms are transitive verbs and thus in the active voice
imply that the loss had an outside cause. The same suffix often occurs in the
reflexive/passive type of construction, that is, with the detransitivizing prefix (516,
517). It is used where the cause is unknown or where one is speaking carefully to
avoid blaming someone.
(514) Ow m-ahwo-k-e-si.
1PRO 2S-joy-LOSS-SF-INP
'You are making me sad (taking away my joy).'
(517) K-et-oh-ka.
1 S-DETRANS-animal-LOSS + TP
º lost my pet.'
(i) Single conditions. There are two suffixes added to verb stems which are used in
conditional expressions: -taw 'while' or 'if,' and -ehe 'after' or 'if. The resulting
forms are adverbials. See sect. 14.2 for description of these forms and further
examples. Verb stems with these suffixes are very common.
(ii) Multiple conditions for any result are expressed differently from a single
condition. They do not occur with all the conditions having the suffix -taw 'if, but
with the stem formative suffix in a finite verb form for each condition. This finite
verb is in the uninvolved nonpast mode, and usually co-occurs with the particle na
'UNCER.' A summary of the conditions is then given using the copula with -taw
once only in the protasis, which is then followed by the apodosis.
wi-tw-e-si.
IS-shoot-SF-INP
'If I had my gun I would shoot it.'
aw-etapa-ce-ri oko.
2-hit-COLL-NOMZR + CONTRAF pain
'If you all had been there, it would have struck you.'
18.3.3 Debitive mode. This mode is expressed by verbs with the affix set tf-. ..
-po-re, which is described in sect. 11.5.
18.3.4 Potential mode. This mode is expressed by the uninvolved non-past tense.
The component of potentiality may be heightened by the addition of the particle na
'UNCER'. The brackets mark off an idiomatic phrase.
18.3.5 Potentially opposite mode. Expressions of this mode occur only in the past
tenses. The meaning is that something bad might have happened but did not happen
because of something protective that occurred. The particle cma ¼ÑÑ' not followed
by another particle, indicates this mode when the verb is in one of the past tenses.
120 Hawkins
18.3.6 Desiderative mode. One way of expressing this mode is with the particle
cma "WISH, OPPOSITE POSSIBILITY', which is fully described in sect. 21.3 (xv).
A second way this mode is expressed is by using a nominalized verb plus the
postposition xe 'wanting' (sect. 17.2), followed by the copula (527). The verb is
nominalized by a zero suffix. This nominalized verb and postposition constitute a
postposition phrase, which is functionally equivalent to a subordinate clause. Cma +
re may occur with the subordinate desiderative clause to augment its force (528).
18.4.1 Person. The persons of the subject and the direct object are marked in the
verb prefix. Each object prefix is a composite form that marks both third person
subject and first or second person object, or first person subject and second person
object. Collective number is marked by the verb suffix for the subject and by both
Wai Wai 121
the verb suffix and following particle so for the direct object. The 1 + 3 subject has
two markers: the 1 + 3 person free pronoun amna, and the third person subject prefix
to the verb. The 1 + 3 object is marked by the pronoun amna only. There is never a
collective indicator for the 1 + 3 person. The paradigm of the person-marking subject
prefixes in the finite verb is shown in sect. 23.3.2, and that of object prefixes in sect.
23.3.3.
18.4.2 Number
18.4.2.1 Collective and non-collective suffixes. These suffixes also indicate tense
and occur with verbs only. The suffixes listed here have been presented with
examples in sect. 18.1; here I present them together in 3 lists.
LIST 1
NON-COLLECTIVE COLLECTIVE
Involved nonpast tense -si -xe
Uninvolved nonpast tense -0/-y -tu
Today past tense -0/-w/-o -cow
Involved past tense -ye/-y/-e -ce
Uninvolved past tense -kne -tkene
The collective suffixes of List 1 always refer to the person of the subject when
they occur as the only marker of collective number. When the person of the object
is collective, the same set of suffixes occurs, with the addition of the particle so
'COLL'. (See 18.4.2.2 below and sect. 21.1(v)).Verbs with 1 + 2 subjects or objects
use the non-collective suffixes if only two people (speaker and hearer) are involved
(532, 534), and the collective suffixes if more than two are involved (533, 535).
(532) Ci-mi-ya-si
1 + 2S-tie-SF-INP + NCOLL
'We (you and I) will tie it.'
(533) Ci-mi-ya-xe
1 + 2S-tie-SF-INP + COLL
'We (all of us) will tie it.'
(534) K-een-a-s
1 + 2O-see-SF-INP + NCOLL
'He sees us (you and me)!'
(535) K-een-a-xe so
1 + 2O-see-SF-INP + COLL COLL
'He sees all of us.'
122 Hawkins
Here I list the first and third person imperative and hortative suffixes which
indicate person and collectivity or non-collectivity. These suffixes have been pre-
sented with examples in sects. 11.1 and 11.2.
LIST 2
NON-COLLECTIVE COLLECTIVE
1st person imperative -si
1+2 person imperative -ri -ceri
3rd person imperative -pe -cowpe
1st imperative of going -tan
'Let's go.' (kayka) -0 -tko
Here I list the second person imperative forms according to collectivity or non-
collectivity and according to motion or non-motion.
LISTS
NON-COLLECTIVE COLLECTIVE
2nd person imperative -ko/-ki -coko
2nd imperative of going -ta -tatko
2nd imperative of coming -tamko -tamcoko
The suffixes of List 3 co-occur in consonant-initial intransitive verbs with the prefix
a-, apparently indicating 2 person subject (536). Imperative of motion suffixes are
displayed in (537) and (538).
(b) A-wrata-coko
2S-cry-COLL
'You (COLL) cry.'
(b) Eh-tatko
bring-2IMPMOT + COLL
'You all go bring it.'
(b) En-tamcoko
see-2IMPMOT + COLL
'You all come see it.'
The collective particle so does not occur with any forms in List 2 (sect. 18.4.2.1).
18.5 Voice. This has been dealt with in sects. 4, 5 and 6 and 23.2.3-5.
18.6 Nonfinite verb forms. The only nonfinite verbal forms are the derived nomi-
nals (described in sect. 15.4) and derived adverbials (described in sect. 23.5.1).
18.7 Incorporation. Names of parts of the body are often incorporated into the
verb. Many of such compound forms have to do with suffering or some sort of
calamity (542, 543, 544).
(542) K-e-kana-katapa.
1 S-DETRANS-head-hit + TP
º hit my head.'
(543) K-e-hre-xwo.
1 S-DETRANS-lower.leg-break + TP
º broke my leg.'
124 Hawkins
(545) ft-es-ewan-kari-c-e-si
3S-DETRANS-chest-strong-VBZR-SF-DSTP
'He is getting up his courage (strengthening his chest).'
(546) Es-ew-y-ahru-ko.
DETRANS-eyes-GEN-close-2IMP
'Close your eyes.'
There is one verb stem that can be incorporated into any other verb stem to make
a compound verb, that is, the stem -tihka 'to finish doing it,' 'to become complete'.
This is treated in sect. 18.2(v).
There are a few other cases of compound forms, where the nucleus consists of
either two nouns (547), or a noun and a postposition (548). In these cases a verb can
be formed by the addition of a verbalizing suffix.
18.8 Auxiliary verb system. There are no true auxiliary verb stems in Wai Wai.
There is, however, one group of two verb stems in which the second, which is
always the copula, seems to function as an auxiliary verb. The first word of such a
group is a verb stem plus the negative suffix which derives an adverb (sect. 23.5.1.1
(ii)). The two words are closely related since the copula carries the affixes indicating
tense, mode, person and number, which are normally carried by the verb that
expresses the action. Also, when the copular form begins with a vowel, as in the
imperative mode, the final vowel of the negativized verb is obligatorily elided (550).
There are no adjectives or adjectival phrases in Wai Wai. Nouns modify nouns.
There are, however, two or three pronouns which can modify nouns. (See sect.
15.3.1.)
Adverbs derived from nouns and verbs are more numerous. A good sampling of
these is given in sect. 23.5.
20.1 Adverbs usually modify verbs, or function as the complement of the copula.
They may consist of one adverb alone, or they may occur with one or more particles.
Very often adverbs occur sentence initially (552, 553, 554).
126 Hawkins
komo poko.
COLL occupied.with
'Do your work with strength.' (Lit.: Be strong in doing your work.)
20.2 Adverbs derived from verbs often occur with a derived adverb to make an
adverb phrase. Such adverb phrases constitute subordinate clauses. They occur either
before the verb (558) or after it (559).
kaf-pe.
strong-ADVZR
'Eat now in order to paddle with strength.'
The suffix -aw follows the negative nominalizing suffix -pM to derive an adver-
bial form. The meaning is that an action occurs prior to another action. Such a form
occurs with either a preceding noun or pronoun (560) or a person-marking prefix
(561).
moku-hn-aw.
come -NOMZR+NEG -ADVZR+when
'He lived long ago, before we came.'
oy-ewru-n-aw ka.
1 -be.born-NOMZR+NEG-ADVZR+when TEMP
'They lived here before I was born.'
There is one postposition which takes the same form and appears to be in the
same class as that described in the previous paragraph. Yet the stem seems not to be
that of a verb but that of the noun meaning -ari- 'contents.' It is a little hard to make
that meaning fit its use with these affixes.
y-arf-n-aw.
GEN-contents-NOMZR+NEG-ADVZR + when
º was there in his absence (when he was not contained in that place).'
tko
NT IS-be-SF-INP
º am not a Wai Wai, but I am like a Wai Wai.'
20.4 Collective indicator. (See sects. 18.4.2.2 and 21.1(v).) The particle so indi-
cates collectivity with adverbs. Note that, in the following examples, collectivity is
indicated by the free form subject plus the collective particle homo (568), with the
tense-number suffix to the copula, and with the particle so after the adverb (568,
569). Distribution of items throughout all the members of a group is also indicated
by the same particle so following the adverb and following the indirect object (570).
Particles
21 Particles
Particles (except taa, see sect. 21.2(i)) follow words of any class other than the
ideophone. They never occur in isolation. I have examples of three particles that
occur occasionally with nominalizing suffixes: roro, maki, and ki. See sects. 21.1(ix)
and 21.2(viii and ix). In each case the focus of the nominalization is the adverb or
postposition which precedes the particle. Particles are not derived from any other
class of words. They may be divided into four classes according to their function:
modifying particles, discourse particles, involvement particles, and interjections.
21.1 Modifying particles. These particles are all closely related to the word that
precedes them. These also tend to occur alone in the phrase without any other
particle.
(i) pen is normally used only of persons. It means they are either dead, permanently
gone from the speaker, or injured.
oy-akno pen.
IPOSR-brother pity
'My poor brother caught a lot of rain.'
(ii) pen also means 'a large group' of any item. It is often used in exclamations with
no verb at all.
The word komo also is closely attached to the noun it modifies, as in the above
examples. But there are three particles, pen, pin, and ro that occasionally occur
between it and the noun komo modifies.
ro komo.
DISTRIBUTION COLL
'We will point out the loads for each one of us to carry.'
Wai Wai 131
(It should be noted in this example that the word komo does not refer to the
subject of the verb, as amna is never modified by a collective. Komo refers to the
multiple categories of items to be carried.)
(v) so 'COLLECTIVE' refers to objects of transitive verbs when those objects are
indicated by personal prefixes (584). It does not occur when the object of the verb
is a free form (585). It also occurs with objects of postpositions (586), or with
adverbs (587) to indicate collectivity of the referents.
(vii) ymo 'ENORMOUS, AWESOME' In the word okoymo 'anaconda snake' the
letters ymo are not a particle but a part of the noun. Yet the meaning of the particle
evidently stands out in that part of the word.
exi-taw ro.
be -ADVZR + while as . long, as
'Let's use the saw as long as we have it.'
mak.
just
'Why is he that way. It is just his nature.'
21.2 Discourse particles. These particles occur following any part of speech and
refer more to the whole clause or sentence than to the head word they follow.
(i) tea 'Assent,' 'discourse introducer'. This particle is always followed by pause. It
is the one particle that occurs expression initially and never occurs in any other
position.
ha.
RHY
'This man says he will tell what you have seen.'
(iv) taki 'thematic focus, change'. This particle is the opposite of the particle ha
'athematic focus'. Taki indicates thematic focus, that is, it focuses attention on the
next occurrence in the main line of action of narrative discourse. It seems not to be
used with this meaning in any other type of discourse. This particle also calls
attention to any change of state or action (606).
(v) ka/kapu 'now,' 'for a limited time'. The longer form seems to be decreasing in
use, but either form may be used.
(vi) cokori 'first in sequence'. An indication that this word is a particle and not an
adverb is that it never occurs expression initially. It often occurs in conjunction with
the particle ka described above.
(viii) maki 'limit of action or number','metaphor'. It can indicate that the preceding
noun or verb is a metaphor (614, 615). One of its meanings is 'just so much.' My
assumption is that etymologic ally the particle was simply ma and that the particle ki
(see (ix) below) was added later to form one word. It is one of the three particles
that occur with nominalizing suffixes (612). The nominalized form reverts to the
original stem, producing the form man.
(ix) ki 'exact measurement'. It can occur with a nominalizing suffix. For a clear
indication of the meaning compare examples (616 a,b).
(629) Ow xa wi-tw-e-si.
1PRO UN IS-shoot-SF-INP
º am the one to shoot it.'
(xv) hara 'in turn' (631), 'back again' (632), 'argument' (633)
(xvi) makirha 'in turn'. I do not know any difference between the meaning of this
particle and the first meaning of hara.
w-am-e-si marha.
lS-fell.trees-SF-ÃÍÑ also
'While I am there I will cut underbrush for a field, I will also fell the
trees.'
(xix) meero 'beyond expectations'. This particle may be translated 'even'. There is
a slight difference of meaning between this particle and the particle rma in its
meaning of 'contrary to expectations' (cf. (639) and (623)).
21.3 Particles indicating the attitude of the speaker. A noteworthy item of syn-
tax is that the particles na, ma, ti, and mi, when occurring with a nonpast verb only,
occur with the uninvolved forms of that verb.
(i) ti '3rd person responsibility'. The speaker uses this particle when he is reporting
the words of another. He may be reporting an action told by another person or he
may be reporting an opinion or a command of another person. He may be explaining
what another person has just said, but in such a case the ti is preceded by the verb
form kee 'he says.' I have been assured by Wai Wai speakers that the use of this
particle does not necessarily cast doubt on what they are reporting. Most of the
occurrences probably have the meaning of 'hearsay,' that is, mere reporting. But
when a person is carrying a message or delivering a command from another person,
the same particle is used with a somewhat different meaning. º am not responsible
for what I am reporting,' is the meaning that becomes clear at times (641).
Occasionally, when a person explains to another person what someone in the present
company has just said for all to hear, the verb 'to say' is used followed by ti. If it
were only hearsay reporting, that verb would be sufficient to indicate that. But the
ti is often added, with the clear sense of 'It is not I who said this.'
but it is also found in verbal clauses. At times this ha may not be used when a person
is narrating an event in an unemotional manner. But when emotion is aroused it very
often occurs. This particle ha occurs with more stress than either of the other two
particles ha (see sect. 21.2 (ii) and (iii)).
(iii) mi 'evidently, agreement'. The basic idea of this particle seems to be: I am led
to believe something because of the evidence or because of the words of another
person. It is also used in expressions of obedience to commands. This particle nearly
always cooccurs with the rhythm carrier particle ha, but the meaning of mi is
unchanged. By morphophonemic rules the vowel of the mi is lost, leaving only one
letter to the particle. We then join the two particles to avoid writing a one letter
word.
(vi) ke 'ignorance, instruction, correction'. The speaker may believe his hearer is
stupid or that he himself was not very wise. The instruction or correction one gives
another often includes this particle. These words of wisdom often carry with them a
note of some condescension in the speaker's voice. This particle often occurs in the
same clause with either the particle ma (651, 652), or the particle w/(653), or it may
occur alone without other particles (654).
(vii) men 'warning, urgency'. As with all warnings, this particle carries the urgency
for the listener to take careful note of what he is telling him. The speaker may be
warning of a danger or merely pressing his desire for some action upon the listener,
or he may be considering some danger to himself. The sequence of particles ha and
men, which is quite common, reinforces the meaning of first person responsibility
for an expression containing this ha. This is further supported by the fact that I have
never heard men in the same phrase with ti.
142 Hawkins
(viii) wa/wara 'intention,' 'plan,' 'readiness'. It should be noted that the two
allomorphs of this particle are homophonous with the two allomorphs of the
postposition meaning 'like' (sect. 17.2).
(ix) tko/matko 'new thought'. The particle may indicate a thought that is new to the
listener or to the speaker himself. The speaker may be even trying to recall
something he has heard in the past, that is, a new recollection of an old thought.
Sometimes the insertion of a new thought into a conversation between two or more
people may sound like a contrasting thought, and sometimes it is so. But the main
idea of the particle is not contrast but newness.
Ow xa matko ki-wc-e-si.
1PRO UN NT IS-go-SF-INP
'There is no one who will go. I will go (had not thought of going be-
fore).'
(xi) xe 'examine for proof. When the speaker is sure of something he has said, he
may use an expression with this particle. It means roughly, 'Go see for yourself that
I am right.'
ape-ko xe.
feel-2IMP for.proof
"The axe is very sharp. Feel it for proof.'
(xii) yka 'inquiry'. It indicates that the speaker does not know how a matter is and
seeks to find out. This particle most commonly occurs following the verb 'to see.'
(xiv) kyam 'supposition,' 'unimportance'. It indicates that what the speaker is saying
is said on the spur of the moment and unpremeditated, or that the results of an action
are small.
144 Hawkins
kyam mak.
unimportance only
º will build my house, just a small one.'
(xv) cma 'wish,' Opposite possibility'. When occurring without another particle in
the phrase, it indicates a wish for the opposite circumstance, which may be only
implied (669), or gratitude that the opposite circumstance did not occur. This particle
followed by the particle re 'frustrative' and a verb in the involved nonpast tense is
used to say 'please' (671). The idea is that if you please you will do what I say, but
if not it will be OK. This particle pair is never accompanied by an imperative form
of the verb. The particle cma, either alone or followed by re, and having the meaning
of Opposite possibility', can occur following a verb in any tense (670), (672).
(xviii) yu 'failure,' 'near failure'. The vowel of this particle is often dropped. When
it is, I write it as a suffix to the previous word to avoid a one-letter word. This
particle is often followed by the particle re 'frustrative'.
21.4 Particle pairs. Besides the particle sequences mentioned in paragraphs above,
there are certain particles that often occur in pairs and result in a special meaning
that might not be evident from the sum of the two meanings. I have already
mentioned the particles cma + re as being a special pair. A list of other pairs
follows:
xa hara - 'again'
(Alt. form of above pair xara)
ro maki - 'extremely'
xa marha = 'in the same way'
Illustrations of most of these pairs have appeared in the examples given above.
21.5 Interjections. Interjections are also particles in that they are not inflected. But
they are different from the preceding particles in that they may occur alone or
expression initially. In this way they are similar to ideophones. But interjections are
different from ideophones in that, although members of either class may be followed
by a form of the verb ka/kas 'do/say', following interjections this verb means 'say'
and following ideophones it often means 'do.' Also interjections express feelings,
whereas ideophones express actions.
(i) okre 'delight'. Intonation is an important part of this word to distinguish it from
the homophonous form described in the next paragraph. This word has a low tone
on the first syllable and a mid to high tone on the final syllable.
(ii) okre 'doubt'. This particle has low intonation on both syllables.
kahxi che-ka.
rapids among-to
'Alas, I dropped Ewka's gun into the rapids, alas.'
(v) oko/aka 'deep pain'. The word oko may occur alone or in a sentence. The word
aka which refers to the same kind of pain is only a one-word cry. It never occurs in
a sentence. The meaning of the two seems to be identical.
(vi) oco/aca 'burning pain,' 'brilliant light'. Here again the first word may occur
alone or in a sentence while the second only occurs alone. The meanings of these
two are approximately the same, although the first word here has the broader
Wai Wai 147
meaning indicating either burning pain or brilliant light. The second indicates only
a burning pain.
(viii) kn 'awe'
wi-nom okwe.
IS-leave + TP alas
Oh, I left my knife very far away alas.'
148 Hawkins
Phonology
22 Phonology
The phonemes are listed in sect. 22.1 and 22.2. The orthographic symbol used
throughout this paper is given first. This is followed by the corresponding standard
phonetic representation when this differs significantly from the orthographic symbol,
except for r and f, which do not correspond closely with any standard phonetic
symbol).
lot is a mid back vowel and is rounded. It has the central unrounded [Ë] allophone
following the alveopalatal consonants Id and /x/.
/a/ is a low front unrounded vowel. It has the less open and less low allophone
[a;] when it is short and when it is preceded by /y/ or Id and followed by /w/, /m/,
or /s/.
They are /t/, /s/, /x/, Id, /n/, /ft/, /r/, /f/, and /y/. I call these the Tense group, since
the articulation of each one is more tense than those of the next group.
The rest of the consonants make up the other group. They are /k/, /m/, /p/, /w/, and
/h/. I call these the Relaxed group. Generally speaking, relaxed consonants form
clusters with tense consonants and with other relaxed consonants much more than
tense consonants do with other tense consonants.
/f/ is formed with the blade of the tongue and is an alveopalatal with popped lateral
release. It clusters in word-medial position before all the relaxed consonants except
/h/ and after all of the same consonants except /p/.
150 Hawkins
lyl is a high front semivowel, the same as a quick /i/ in Portuguese. It clusters in
word-medial position before any relaxed consonant and after all the same consonants
except /p/. It does not cluster with any of the tense consonants.
/h/ is a central unarticulated aspiration. It clusters before all the other consonants and
after the consonants /t/, /n/, /c/, /r/ and /m/. It does not occur word initially except
in free variation with zero before vowel-initial words and in the particle ha.
22.3 Syllables. There are five syllable patterns in Wai Wai, V, CV, CVC, CCV,
and CCVC. Examples follow in the first syllable of each of the following words:
Wai Wai 151
The syllable division presented here is based on the phonemes which occur in
word-final position. All vowels occur in word-final position and also the consonants
/m, /n/, /w/ and /y/. No other consonants occur word finally. Thus the phonemes that
occur word finally can be expected to occur in syllable-final position also. All other
consonants are put with the vowel following them.
There are only four triple consonant clusters in Wai Wai. They are -nth-, -mth-,
-mch-, and -yhy-. It can be seen that they all include /h/ and none of them occur at word
boundaries. In slow speech all three of these clusters have unstable vowels after the first
consonant (included in parentheses). In normal speech those vowels are not heard.
There are nine consonant clusters involving /k/ that occur word initially:
There are three consonant clusters beginning with /x/ that occur word initally:
There are several other consonant clusters occurring word initially. Of these the
first two (in 696) do not occur following pause. The third only occurs following
pause in the ideophone cmow 'kiss'.
152 Hawkins
(700) written
[mawaaxa] man's name Mawaxa
[okoomo] 'wasp' okomo
[yemeeme] 'mama' yememe
[tamaatama] 'butterfly' tamatama
[kiriicici] 'glass beads' kiricici
[akaanakatapa] 'He hit you on the head.' akanakatapa
Some three syllable words, however, that have a basic CV repeating pattern occur
with length on the first vowel. We then write that vowel as a double vowel.
But there are at least two postpositions of two short syllables each which take
prefixes producing a VCVCV pattern and the vowel of the second syllable becomes
long as is normal, and is thus not written.
The word xati One who desires' may receive a verbalizing suffix and verb
suffixes in addition to the personal prefix. In such forms the personal prefix remains
long.
154 Hawkins
(705) Oo-xati-mnok-e-si.
10-one. who.desires-CESS-SF-IN?
'He will stop loving me.'
When two-syllable words which have no long vowels are followed by certain
particles, the last syllable of the first word becomes long. Phonologically, they act
as one word in which the second syllable is long according to the normal pattern.
22.5 Word stress. The word stress on two-syllable words in Wai Wai normally
occurs on the last syllable of the word.
But if such words are followed by certain particles (I am not sure if the rule holds
for all particles or not), not only does length develop on the last syllable of the first
word, but the stress moves from the first word to the particle. The first word then
has no stress at all.
In finite verbs and in words derived from verb stems the stress occurs usually on
the penultimate or final syllable. First, we give examples with stress on the penulti-
mate:
Next, we give examples of verbal words with stress on the final syllable. More
study is needed to define the rules of such stress.
22.6.1 Vowel loss. For convenience of description I have categorized the phonemes
of Wai Wai as follows:
(i) Unstable vowels are lost in word-final position after the nasal consonants /n/
(735) and /m/ (736) unless the word is followed by a word beginning with a
consonant cluster (737).
Wai Wai 157
(ii) Unstable vowels are also lost from morpheme-final position when occurring
between /k/ and a morpheme beginning with a single continuant consonant (738,
739) other than /s/ (740), unless that consonant is followed by another unstable
vowel (741, 742). Unstable vowels in morpheme-final position are never lost before
a consonant cluster (743).
(iii) Unstable vowels are lost from a morpheme when it occurs before a vowel-initial
morpheme.
(746) Ti- + erema + -ce + -ii = Ceremacert 'Let's all sit down.'
(iv) Unstable vowels are also lost from word-medial syllables at morpheme
boundaries under any of the following types of environments:
(a) They are lost between a relaxed consonant and a tense consonant in either
order. See examples below; also (747-748) above.
(c) They are lost when /n/, /m/ or /p/ is followed by /t/, /s/, Id, or /y/.
(760) Eyepu + -so > Eyepso > Eyehso Eyehso kiwcesi. º am going to
take a bath.' (For change of/p/ to /h/ see sect. 22.6.4.)
(d) They are lost between unclustered /r/, /t/, /s/ and /x/, and /y/. (For loss of /y/
and for change of/r/ to /f/ see sect. 22.6.9.)
(e) They are lost from the utterance-final position when they are unstressed and
when they follow an unclustered nasal consonant.
Wai Wai 159
(f) Unclustered III, which is normally stable, is lost from a stem-final position
between an alveopalatal consonant and lyl or /k/.
(767) Kitaci + -ya = Kitaciya > Kitacya > Kitaca > Kitahca
º might get lost.' (For insertion of /h/ see sect. 22.6.9.)
I have not labeled Id and lol unstable because they are rarely lost. Yet they are
lost in certain situations described in the next two paragraphs.
(v) Word-final /e/ is lost from the suffix -pe before the suffix -no, and it is also lost
from the word onoke before words beginning with a vowel or with an unclustered
consonant other than /t/, Id, /r/, or /x/. It is also lost from the involved past tense
singular suffix -ye when it is preceded in the word by more than two syllables or by
a consonant cluster.
(769) panape + -no > panahno 'shallow one, thin one' (Id is lost and /p/
changes to /h/. (See sect. 22.6.4.)
(770) kawawape -I- -no > kawawano 'sour one' (Id is lost and /p/ changes
to /h/ and then is lost. (See sect. 22.6.5 for loss of /h/.)
(772) Onoke krapan on? 'Whose bow is this?' (Id not lost)
(775) Wi-mi-ye.
IS-tie-IP
º tied it.'
160 Hawkins
(776) W-afma-y.
IS-throw-IP
º threw it.'
(777) ft-erema-y.
3S-sit-IP
'He sat down.'
(vi) Word-final /o/ is lost from the word kofpo when the suffix -no is added, and is
also lost from the word-final position in the four other words mentioned below when
they are followed by the particle ha.
22.6.3 Palatalization. The consonants /t/ and /n/ are replaced by Id and /Þ/ respec-
tively when they occur before /i/, /e/, or /y/.
the IM before that is replaced by /i/ by vowel harmony. Then the /t/ be-
fore that is replaced by Id by palatalization.)
22.6.4 Reduction of /k/ and /p/ to /h/. The consonant /k/ is replaced by /h/ before
a morpheme beginning with /t/, /k/, /s/, or Id (790-793). The consonant /p/ is usually
replaced by /h/ when it becomes part of a consonant cluster (794—796).
22.6.5 Loss of /h/. When /h/ occurs at morpheme boundaries the /h/ is usually lost
if it is preceded in the word by more than two syllables (797), by a long vowel
(798), by a consonant cluster (799-802), or by the letter /m/ (804). /h/ is also lost
from stem-medial positions when it is preceded in the word by a long vowel (805).
(798) Eemipu + -ko = Eemipuko > Eemipko > Eemihko > Eemiko
'Patch it.'
(805) Wi- + ii + tihka = Wiiitihka > Wiitihka > Wiitika º finished mak-
ing it.'
22.6.6 Loss of prefix tf-. The prefix ii-. marking adverbialization of a verb (sect.
23.5.1.1) or noun (sect. 23.5.1.2) is lost before word-initial /k/ and before the stems
cemaro,. 'lie,' and sahsa 'saw', and the resulting forms have zero prefix. But the
prefix ti- meaning third person reflexive possessor (sect. 4.1) is not lost in a similar
environment (811-813).
22.6.7 Loss of the word-final syllable η from verb stems. Verb stems that have
ri as their final syllable lose that syllable before a suffix that begins with an
unclustered tense consonant, except before the tense consonant lyl (814—816).
22.6.8 Change of /u/ and l\l to their respective semivowels. The vowels /u/ and
/i/ remain vowels when they are stressed, and when they are followed by a word
beginning with a consonant cluster. Unstressed, and occurring after other vowels,
they are changed to semivowels. The stress patterns vary, as is shown below, and
that affects the form of these two vowels. The vowel Í remains the same before /t/
(820). These vowels are changed to semivowels in other positions.
22.6.9 Loss of /y/ and insertion of /h/. When a morpheme beginning with lyl oc-
curs following Id, /n/, /r/, or /s/ the lyl is lost. In such cases the /r/ and /n/ are
replaced by the palatalized forms /f/ (825) and /nV (826). When a long vowel should
normally occur before Id the vowel remains short and /h/ is inserted before the Id
(828-829).
(829) enati + -ye > Nenatye > Nenacye > Nenace > Nenahce.
'It ran out, finished.'
22.6.10 Replacing of /u/ by l\l. Verb stems which end with the syllable pu occur
with the /u/ replaced by N when followed by the allomorph -ra of the suffix -hra
'NEGATIVE + ADVERBIALIZER,' and the allomorph -n of the suffix -hni 'NEGA-
TIVE + NOMINALIZER.'
Morphology
23 Morphology
Some of the morphemes have been described elsewhere in this paper and some
lists have been given. I will here seek to discuss the morphemes that have not been
previously discussed or that have only been partially discussed.
23.1 Verb stem variations. There are a few verbs that have variant stems beyond
the range of morphological rules. We list these here.
Eto-ko.
to-2EMP
'Go.'
There are several verbs which have stem variation depending on transitiveness or
intransitiveness:
166 Hawkins
There is one verb stem that may be either transitive or intransitive according to
the context (847). Another verb stem may be either positive or negative according
to the context (848).
The stem of the copula is, of course, the most varied stem in the language. I list
its variant forms here:
Wai Wai 167
23.2.1 Stem formative suffix. The meaning of this suffix and its variation according
to tense or derivation has already been presented in sect. 18.1.1.3. I discuss here the
alternation of its allomorphs more fully. The allomorphs are -e/-ya/-wa/-a. The first two
allomorphs are non-derivational, that is, only stem formative. The final two allomorphs
are both stem formative and derivational, deriving nouns into verb stems.
(i) Allomorph -e occurs on verb stems that end with /a/, /o/ or /e/ and it replaces the
/a/, /o/ or /e/.
(b) W-aam-e?
1 S-cut.down-SF + UNP
'Shall I cut it down?'
(b) W-enep-e-si.
IS-send-SF-INP
º will send him.'
(ii) Allomorph -a occurs with a limited number of noun stems which occur with the
possession suffix -n. I label the suffix -a as the stem formative suffix, but it also
168 Hawkins
seems to function as the derivational suffix deriving the nouns into verbs. To
demonstrate this I give the underived noun form with its possession indicator
following each verbal example.
(iii) The copula also takes the allomorph -a as a stem formative (857). hi such cases
the stem of the copula is zero or the one phoneme /x/. This suffix, which is
derivational with all other verbs, is not derivational with the copula. When the
copula occurs with the adverbializing suffix -taw 'if/when,' the stem formative suffix
which is normal for this construction does not occur (858).
(iv) Allomorph -wa occurs mostly on noun stems. It serves the double function of
deriving a verb stem from a noun stem, as well as being a stem formative for a set
of tense suffixes. It also usually has the meaning of ingressive action, or change of
state (sect. 18.2(iv)), but in (859) that meaning is difficult to see. Allomorph -wa
occurs following the possession suffix -n with nouns which take that suffix (859,
860). Noun stems that occur with other possession suffixes take the suffix -wa
without any possession indicator before it (861, 862). The examples marked (b) are
the underived nouns showing their possession indicating suffixes.
(b) o-warawa-n
1 POSR-trading.partner-POSN
'my trading partner'
(b) aw-epera-n
2POSR-sickness-POSN
'your sickness'
(b) oy-ew-ru
IPOSR-eyes-POSN
'my eyes'
(v) Allomorph -ya occurs with all other stems (863-866). In (865) the /y/ of this
suffix does not appear because it coalesces with the /r/ of the stem to become /f/
according to the rule in sect. 22.6.9. This allomorph does not have the components
of ingressive action or derivation. But it occurs following certain derivational
suffixes (864, 866). It is thus an outer layer stem formative suffix.
(863) K-mok-ya-si.
IS-come-SF-INP
º will come.'
(864) 0-Pof-mam-ya-si.
3S-large-INGR+VSF-SF-INP
'He is growing.'
170 Hawkins
(865) W-aaf-a-kne.
IS-cany-SF-UP
º carried it.'
(866) Ki-hya-pam-ya-si.
1 S-shame-INGR+VSF-SF-INP
º am ashamed.'
23.2.2 Derivational stem formative suffixes. There are other derivational suffixes
besides -wa discussed in the previous paragraph. These discussed below derive other
word classes into intransitive verbs.
(i) The two verbalizing suffixes mentioned above (864, 866) derive verbs from
nouns (867, 868), or from adverbs (870-871). They are similar in form. They are the
suffixes -mam and -pam. They also have the meaning of ingressive action related to
the noun root (sect. 18.2(iv)). It can be translated 'become' or any similar word.
(867) Ki-tit-mam.
IS-still-VSF + INGR + TP
º became still (made no answer).'
(Tit seems to be a noun in the word tit-pe (stillness- ADVZR) 'calm
(wind)'
(868) Ku-wun-pam.
1 S-laziness-VSF + INGR + TP
º got lazy.'
(869) 0-Cey-pam.
3S-hardness?-VSF + INGR + TP
'It got hard.'
(870) 0-Pof-mam-ye.
3S-large-VSF + INGR-IP
'It grew large.'
(871) 0-Cerere-pam-ya-kne.
3S-brittle-VSF 4- INGR-SF-UP
'It became brittle.'
(ii) Several verbs that occur with the stem formative suffix -a in certain forms occur
with -n, which seems to be a possession indicator before that stem formative suffix.
But in other forms of the same verb root there is considerable variation in what seem
to be merely verbalizing stem formatives. There are three such forms: -m, -nm, or
mm.
Wai Wai 171
(iii) Verb stems that occur with the stem formative suffix -wa in certain forms (sect.
23.2.1(iv)) occur with the verb stem formative suffix -ta in all other forms. Most of
these suffixes are added to nouns, but in one case the stem formative suffixes are
added to variants of the noun stems, as in the verb 'to talk.' In the noun iapoia/talk'
these verbalizing and stem formative suffixes replace the last syllable of the noun
(874). Nouns which take the possession suffix -n also take it before the suffix -ta as
well as before the suffix -wa (875). Nouns which occur with other
possession-indicating suffixes drop the possession indicator before taking the
ingressive verbalizing suffixes (874, 876). Some verbs occurring with these suffixes
have the meaning of ingressive action (876). But in others this meaning is hard to
find, as in the verb 'talk' displayed below (874).
(iv) A limited number of verb stems occur with stem formative suffix set -wa and
-timi instead of the set -wa and -ta discussed in the previous paragraph. All these
variant forms have the double function of stem formative and verb derivational
suffix indicating ingression. The suffix -timi when preceded and followed by a vowel
is shortened to -tm.
There is one verb that occurs with the suffix set -wa and -im with the same
components of meaning and same positions as the sets described above. Note that in
the first example below with the suffix -wa the /w/ of the suffix is elided following
the /w/ of the verb stem.
23.2.3 Other stem formative suffixes to nouns. There are five verb formative
suffixes occurring with nouns which seem to have a causative meaning. The result-
ing verb stems are all transitive.
(ii) The suffix -hto (for loss of initial h in (882), see sect.22.6.5)
(i) Allomorph es- occurs with stems beginning with /e/ followed by a consonant
other than /x/ or Id.
(890) K-es-ehce-m-e-si.
1 S-DETRANS-medicine-VBZR-SF-INP
º am putting medicine on myself
Wai Wai 175
(891) K-ex-exewmik-ya-sl
1 S-DETRANS-choke-SF-INP
º am choking myself. '
(892) Ec-eceto-ko.
DETRANS-hold:steady-2IMP
'Steady yourself.'
(iv) Allomorph et- occurs with stems beginning with /a/, lot, /i/, or with a consonant
followed by l\l or /u/, or with a stem that takes the prefix t- '3' or 'GENL' (sect.
(894) ft-et-on-e-si.
3S-DETRANS-eat-SF-INP
'It gets worn (is eaten).'
(895) ft-et-wMma.
3S-DETRANS-ruin + TP
'It was ruined. '
(896) ft-et-puru.
3S-DETRANS-roast + TP
'It is (became) roasted.'
(899) N-ese-wakr-e-si.
3S-DETRANS-be.kind.to-SF-INP
'He is being kind to himself (buying lots of goods).'
(900) N-e-hkoto.
3S-DETRANS-cut.in.two + TT
'It got cut in two.'
(901) N-e-macaka.
3S-DETRANS-push.over + TP
'It fell (got pushed) over.'
23.3.1.1 Second person imperative affixes. These are listed in sect. 18.4.2.1.
They are discussed in sect. 11. Two items of morphophonemic variation remain to
be mentioned: the second person static imperative suffix -ki occurs with stems that
end with /a/ (902). The form -ko occurs with all other stems. Intransitive verbs that
begin with a consonant take the second person subject prefix a- (907) except the
verb stem to 'go' (908) which takes an alternate stem in the second person impera-
tive forms. Some consonant-initial transitive verb stems take the GENERAL PRE-
FIX, functioning as a third person object marker (905). (See sect. 23.4.5.1 (ii) for
the variant forms and the phonological conditioning factors of the GENERAL
PREFIX.) Other consonant-initial transitive stems (e.g., 904) and all vowel-initial
transitive stems (903) have no prefix, but reference to third person object is still part
of the meaning of the transitive stems.
(902) Erema-ki.
sit-2EMP
'Be seated.'
(903) 0-en-ko
3O-see-2IMP
'Look at it.'
(904) 0-mam-to-co-ko.
3O-circumference-VBZR-COLL-2IMP
'Surround him.'
Wai Wai 177
(905) Õß-hkoto-ko.
3O-cut.in.two-2IMP
'Cut it in two.'
(907) A-mo-ko
2IMP-come-2IMP
'Come.'
23.3.1.2 First and third person imperative affixes. Subject prefixes occur on
first and third person imperative verbs, and they are the same as on first and third
person finite verbs (909, 910). The collective-indicating suffix on verbs having a
third person subject is the same as for collective-indicating suffixes on finite verbs
(911). I have no record of object prefixes occurring on first or third person impera-
tive verbs. There are different first and third person imperative-marking suffixes for
such verbs, and the prefix and suffix always agree in person. (See sect. 18.4.2.1 (List
2) for suffixes and sects. 11.1.2 and 11.1.3 for other examples.)
Hortative (1 + 2S) suffixes are given in sect. 18.4.2.1 (List 2) and are described
and illustrated in sect. 11.2.
23.3.2 Verb subject prefixes have been referred to several times. Below I present
a table of these prefixes, followed by a description of the morphologically deter-
mined alternation. For morphophonemic alternation see sect. 22.6.1(iii), 22.6.2 and
22.6.3.
178 Hawkins
2 m(i)- m(i)-
3 ç(ß)-/0- ç(ß)/0-
1+2 t(l)-/tit(i)- t(i)/tit(i)-
1+3 (amna + n(i)/0-) (amna + n(i)/0-)
The transitive set of subject prefixes given above occurs only when the object is
third person. For all other object persons the transitive set of object prefixes occurs
(see sect. 23.3.3 below).
(i) The above two columns of prefixes are indentical except that the 1 person prefix
for transitive verbs is wi-/0- (912), while that for intransitive verbs is ld-/0- (913).
In this group of examples I illustrate all the subject prefixes except the 1 + 2 prefixes
which are described in (ii) below.
(912) W-aaf-a-si.
IS-carry-SF-INP
º will carry it. '
(913) K-erem-e-sL
IS-sit-SF-INP
º will sit down.'
(914) Mi-hkoto?
2S-cut.in.two + TP
'Did you cut it in two?'
(915) M-epirk-e-si.
2S-fall-SF-INP
'You will fall.'
(916) N-on-e-si.
3S-eat(meat)-SF-INP
'He is eating it (meat).'
(917) Ni-wmk-o.
3S-sleep-TP
'He went to sleep.'
Wai Wai 179
(ii) Of the two forms of the 1+2 person prefix, tit(i)- occurs before stems that begin
with a single consonant (926), and before all stems that in other forms occur with
the prefix tf-GENL (928) and (929). (See sect. 23.4.5.l(ii) for verbs which take the
tf-GENL prefix; and sect. 22.6.3 for palatalization of/t/ to Id.) The form t(i)- º +2'
occurs before all other stems.
(924) C-imi-ya-si.
l + 2S-tie-SF-INP
'We are tying it.'
(925) Ti-hc-e?
l + 2S-go-SF-l-UNP
'Shall we go?'
(927) Ôú-mtapo-ta-ce-rf
l + 2S-talk-VBZR-COLL-HORT
'Let us talk.'
(928) Tit-on-e-si.
1 + 2S-eat(meat)-SF-INP
'We are eating it.'
(930) Tit-wo-rl
l+2S-shoot-HORT
Let's shoot it.'
(932) Titi-raka-ce-rl
1 + 2S-split-COLL-HORT
'Let's split it.'
(933) Cici-f-a-si.
1 + 2S-put/fix-SF-INP
'We are fixing it.'
23.3.3 Verb object prefixes occur with transitive verbs. Almost the same set of
prefixes occurs as possessor prefixes with nouns, and with possessed nominalized
and some adverbialized forms that are derived from verbs (sects. 23.4.5.1 and
23.5.1). The set is also used to indicate person of the object of postpositions (sect.
17). The set consists of:
1 oy- o-
2 aw- a-
3 0- 0-; also the general prefix yi-, i-, ti-
in certain constructions
1+2 k- ki-
1+3 — —
Wai Wai 181
The above set of object prefixes occurs only when the subject is third person or
1+3 person. When the subject is first person and the object second person, the
prefix is k-M- (the same form as 1 + 2O when the subject is third person, see table
above). The free form amna is the only marker of 1 + 3 object.
(937) isT^-etap-e-si.
3S-30-hit-SF-INP
'He will hit him.'
(938) K-een-a-si.
lS + 2O-see-SF-INP
º see you.'
In the case of the second person subject acting on the first person object the
object is indicated by the appropriate free form pronoun: owi T, and knwi º +2',
or kiwyam '1+2 COLLECTIVE'.
The set of prefixes which I have called the "general prefix" functioning as a third
person object marker occurs on transitive imperative verbs when the phonological
conditions are met for those prefixes (944-945). Otherwise there is no overt form of
prefix for third person object. A third person possessor prefix that is identical to this
general prefix occurs quite commonly with non-derived nouns and with nominalized
verb stems. For a discussion of this see sect. 23.4.5.1.
Consonant-initial intransitive imperative verbs have the subject prefix a- '2' (the
same form as the second person object prefix described above).
23.4.1 Noun stem variation. There are at least six nouns that have variant stems
beyond the range of the morphological rules.
23.4.2 Noun stem formation from verb stems. There are 12 processes of nomi-
nalization which are fully described in sect. 15.4.
(i) The suffix ~mu/-m derives nouns from adverbialized verb stems and from many
adverbialized noun stems. Forms with this suffix refer to persons or items that have
the attribute indicated (953, 954), or that are doing or receiving the action indicated.
If the root of the word is an intransitive verb, the referent is doing the action (955).
If the root is transitive, the referent is receiving the action (956). The longer variant
of the suffix occurs when preceded in the word by a consonant (955), or when
followed by a word beginning with a consonant cluster (957). The shorter variant
occurs elsewhere. Verb stems that are nominalized with this suffix also may occur
with the past tense indicator -nhiri 'PAST', which in this construction never occurs
with the -nho variant (958). (See sect. 23.4.5. l(xi).)
(953) ti-kpo-re-m
ADVZR-sweet-ADVZR-NOMZR
'a sweet one'
(954) ti-swa-ye-m
ADVZR-blue-ADVZR-NOMZR
'a blue one'
(955) c-erema-x-mu
ADVZR-sit-ADVZR-NOMZR
One who is seated'
184 Hawkins
(956) t-aa-so-m
ADVZR-take-ADVZR-NOMZR
One that is to be taken'
(958) c-enepa-n-e-mu-nhM
ADVZR-steal-POSN-ADVZR-NOMZR-PAST
One who has been a thief
(ii) The suffix -no derives nouns from various underived adverbs. This suffix has no
variant forms and is suffixed to the otherwise unaffixed stem. See sect. 22.6.1(v) for
the phonological changes in the syllable pe when followed by the suffix -no. It
should be noted that this is a different suffix from the suffix -no which nominalizes
verbs with the meaning of unspecified object. See sect. 15.4(x) for a discussion of
that suffix.
(iv) There are a few adverbial words that are based on noun forms and thus do not
take nominalizing suffixes. The basic noun form may be adverbialized with a suffix
Wai Wai 185
(966) or with the adverbializing postposition me (967-968) (for the postposition me,
see sect. 17.2).
(v) There are three adverbs which take irregular nominalizing suffixes:
(i) Postpositions that end with lol or /w/ are nominalized by the suffix -no, as with
some underived adverbs (sect. 23.4.3(ii)).
(975) yi-ret-wo-no
3-upstream-at-NOMZR
'the one that is upstream'
186 Hawkins
(976) a-wya-n
2-to-NOMZR
One that belongs to you'
Possessor Prefixes
1 Person oy-/o-
1 + 2 Person k(i)-
2 Person aw-/a-
3 Person 0-/i-/yi-/ti- (see (ii) below)
3 Person reflexive t(i)-
Free form possessor + vowel-initial possessed item: y- 'GEN'
(i) Possessors are indicated by prefixes listed above in the following examples:
kT-mxik-ii k-mariya-n
l+2POSR-child-POSN l+2POSR-knife-POSN
Our child' Our knife'
The third person 0- variant occurs with stem-initial vowels (989). Examples and
discussion of third person variants /-, yi- and ti- are given in (ii) below.
Transitive verb stems that have been derived into nouns take the same possessor
prefixes to express the underlying object (990) or, in the construction described in
sect. 15.4(ii), the underlying subject (353, 354). Intransitive stems take the same set
of prefixes to express the underlying subject (992). These prefixes include the third
188 Hawkins
person reflexive prefix (991). With the exception of the one transitive verb nominali-
zation noted above, underlying subjects of nouns derived from transitive verb stems
are expressed with the postposition form -wya 'by' (993). See also sect. 15.4.
(ii) The third person prefix forms /-, yi- and ti- have a more general usage, when no
third person is involved. (See below for the discussion of the GENERAL PREFIX
and the phonological conditioning factors which determine the use of these forms.)
As third person possessor marker, they occur with underived possessed nouns only
when there is no free form possessor (994-997).
This prefix also occurs on derived possessed nouns (1001) and with nominalized
postpositions when they are not preceded by a free form possessor (1000). With
Wai Wai 189
chew-no me.
among-NOMZR ADVZR
º lived among them, among the white people.'
xe w-0-a-si i-yo-hto-topo.
DESID lS-be-SF-ÃÍÑ 3-tooth-VBZR-NOMZR + CIRC
'My knife is dull. I want a file with which to sharpen it.'
t-ow-topo.
3-take.out-CIRC
º have a burrowing flea (in my foot). Bring me a pin with which to
take it out.'
The non-third person uses of yi- and i-, and the phonological configurations of the
stems that constrain their use, are now described. In these usages the forms are
glossed 'GENL'. There are ten verbs (listed below) in which the prefix ti- replaces
yi-/i-, in both the third person and non-third person usages. This set of 'GENL'
prefixes co-occurs with the nominalizing suffixes -xapu and -hni on transitive stems
of the configuration mentioned above.
(a) as a third person marker, with transitive stems in the second person imperative
mode:
(b) as a third person marker, with transitive stems in the adverbial negative
construction:
Wai Wai 191
(c) as the GENERAL prefix, with intransitive stems in the adverbial negative
construction:
(d) as the GENERAL prefix, with intransitive stems in the purpose of motion
construction:
The configurations of the stems that take this prefix are as follows: yi- occurs
before stems that begin with a consonant cluster (1016), or that begin with a single
consonant followed by /i/ or /u/ (1017, 1018), or that begin with a tense consonant
(1019). i- occurs with stems that begin with /y/ (1020). ti- occurs with the 10 verb
stems listed below (1021). The same parameters hold for both underived noun stems
and derived noun stems. See sect. 22.6 for phonological variation of this prefix.
(Following are the 10 verb stems that take the prefix t(i)- '3' or 'GENL': (see
sect. 22.6.3 for the palatization rule.)
The same prefix y- is also added to derived transitive verb stems which begin
with a vowel when they are preceded by a free form possessor that refers to the
underlying object. In the case of derived intransitive verb stems that begin with a
vowel this prefix refers to the underlying free form subject. See sect. 15.2 for
examples.
There is another position in which -y- occurs as an infix, and that is where two
noun stems, the first of which refers to a body part, are incorporated into one word
and the second stem begins with a vowel. This use of -y- seems similar to the
genitive prefix described above. Therefore, I give it the same label. The resulting
form may be simply a compound noun (1029), or it may be verbalized (1031, 1032),
renominalized from the verb stem (1030), or adverbialized (1033).
(1030) ew-y-ahwo-re-cho
eye-GEN-happiness-CAUS-CIRC
'a means of comforting someone'
(1032) Es-ew-y-ahru-ko.
DETRANS-eyes-GEN-close-2IMP
'Close your eyes.'
194 Hawkins
(1033) ti-pimi-y-ewax-ke
ADVZR-neck-GEN-stiflhess-ADVZR
'having a stiff neck'
(iv) With many nouns present possession is indicated by the suffix -rf, -ru. I list just
a few examples of this suffix here:
(viii) A large group of nouns takes the suffix -rii to indicate possession. Any
possessible loan word taken into the language takes this suffix to indicate possession.
(ix) Negation of possession and identity is described in sects. 12.3 and 12.4.
In the examples in (1040), the iy- prefix appears to be another variant of the yi-
'GENL' prefix. In all the cases I have observed it occurs before stem-initial ï or u.
196 Hawkins
(xi) -nho/-nhin 'Past tense indicator on nominalized verbs'. This suffix indicates past
tense on most nominalized verbs, on all nominalized postpositions, and on all
nominalized adverbs. It occurs with five of the derived forms described in sect. 15.4:
-xapu, -yem, -ne, -hrii, and -topof-cho. There are two meanings for this past tense
suffix on nouns derived from verbs. It indicates change of action or state (1045,
1046), or it indicates action performed in the distant past (1041-1044). There is
variation among speakers as to what one may feel is the distant past. There is
morphologically conditioned variation between the different forms of this suffix as
follows: -nho occurs when the word has a first person singular prefix (1041) or a
free form possessor (1042), -nhiri occurs in all other cases.
(1041) oy-es-emani-yemi-nho
IPOSR-DETRANS-play-ASS-PAST
'the one I formerly played with.'
The suffix -nho/-nhiri also indicates past possession on a few nouns which
describe a relationship. It indicates that the relationship has ceased.
A past state of existence of underived nouns can also be indicated by this same
suffix.
The suffix -thof-thin occurs on the majority of possessible nouns to indicate past
possession. Nouns that take the present possession suffix -ri lose this suffix before
this past possession indicator. Nouns that take the possession indicator -rii retain this
suffix before the past possession indicator. For this reason I do not label the syllable
ri of the allomorph -thin as a distinct possession indicator. The morphologically
conditioned variation of this suffix is exactly the same as for the suffix -nho/-nhiri
described above.
suffix -no to indicate the same thing. I have no record of either of these suffixes
being used in the past tense.
(i) Devaluation and endearment. The suffix -tho occurs with unprefixed noun stems
and indicates low estimated value of the item if the item is impersonal (1059). Old
age is implied in many cases. If the noun refers to a person, especially to a relative,
the same suffix indicates endearment (1060).
(ii) Diminutive indicator. The suffix -ci occurs on nouns and indicates that the item
is small. A noun with this suffix may be used as an apellation for a person. Such a
noun may be followed by the adverbializer me as a description of a person or item.
This suffix has a counterpart, the particle ciki, which has the same meaning plus the
component of endearment. The suffix and the particle do not occur together in the
same phrase.
(i) The affix set ti-. . . -so/-xi averbializes verb stems. The suffix variant -xi
normally occurs following stem-final /a/ (1062) and -so follows all other stems
(1063). There are a few stems which occur with -so following the phoneme /a/
(1064). The underlying stem of such verbs ends with pu, which is lost through the
morphophonemic processes of vowel loss, conversion of /p/ to /h/ and loss of /h/.
See sects. 22.6.1, 22.6.4, and 22.6.5.
A variation of the above forms occurs when the verb root begins with Id, /s/, /t/,
or /k/. In such cases the prefix is elided but the suffixes are unchanged. The
adverbialized forms rarely occur, the renominalized forms occurring much more
often. Therefore, I list only the renominalized forms below. The nominalizing
suffixes are discussed in sect. 23.4.3(i). The same elision of prefix rule holds for all
the adverbialized forms described below, whether formed from noun roots or verb
roots.
(ii) The suffix -hra is the negative adverbializer of verb stems (1070, 1071). Before
the imperative of the copula the /a/ of the suffix is regularly lost (1072). See sect.
12.1, and also 22.6.5 for loss of initial /h/ when the suffix follows certain
configurations.
(iii) A set of three affixes, ti-. . . -po + -re, indicates obligation to do an action, or
that the speaker feels it would be good to do an action. (See also sect. 11.5.) (Cf.
the same set of affixes with noun stems, sect. 23.5.1.2(xv)). These affixes occur on
either transitive (1073) or intransitive (1074) verb stems. The resulting form is an
adverbial and may be followed by the copula, in the third person singular form, or
the verb eni 'see' (1073) or the verb enta 'hear'. The copula varies in tense
according to the time of the obligation and the mood of the speaker. The
nominalization of the adverbialized forms are heard only occasionally (1075).
Another negativized form of this adverbial includes the full set of affixes, both
the prefix and the two suffixes, and the negative adverbializer suffix is then added.
The meaning includes personal feelings, as the above constructions do. I have
recorded this on only the two verbs given below, and in each case the adverbial is
followed by the positive finite form of the same verb.
(iv) The suffix -poro indicates that an action is about to be done. It should be noted
that this form does not include the adverbializer prefix ti- which most derived
adverbs take. I have no record of this suffix with transitive verbs.
(v) The affix set ti-. .. -ro indicates that the action occurred repeatedly or
continually during the time another action was occurring. I have only heard this
adverbial used as the adjunct of verbs of coming or going. It is clear that the prefix
is the adverbializer and the suffix carries lexical meaning, that is, time of the action.
We may say that the suffix also carries the component of adverbialization to
conform to other adverbializing affix sets.
(vi) The suffix -so/-xi indicates action that is the purpose or goal of going or coming.
The allomorph -xi occurs following stems ending with /a/ (1087). The allomorph -so
occurs with all other stems. Verb stems having the purpose of motion suffix occur
either before or after the head verb (1088), but they most commonly occur before
the head verb. The adverbializing prefix never occurs with these forms. Possessor
and genitive prefixes occur with transitive stems, and the general prefix with
intransitive stems (1087, 1090, 1091).
Wai Wai 203
(vii) The suffix -tome/-chome indicates that the action is the purpose or goal of
another action. There is free variation between the variants of this suffix following
many verb stems (1093), but with a few verb stems, including the copula, there is
no variation (1092, 1095). Possessor-marking prefixes occur with verb stems having
this suffix (see sect. 23.4.5.1). See sect. 15.4(iv) for the similar forms -cho/-topo
'CIRC' which function as nominalizing suffixes.
eh-tome.
be-PURP
º will put my book under cover so it won't get wet.'
(viii) The suffixes -taw and -ehe. Adverbials formed by these suffixes are fully
discussed in sect. 14.2.
Adverbialization:
Nominalization
(ii) The affix set / / - . . . -ye/-y. The allomorph -ye occurs when it is preceded in the
word by only two CV syllables and is not preceded by a word-medial consonant
cluster, or when the word is followed by a word beginning with a consonant cluster,
or by the nominalizing suffix -m.
Adverbialization:
Nominalization:
(iii) The affix set ti-... -n + -ye is a variation of the previous affix set. I interpret
the suffix -n as the possession indicator (see sect. 23.2.1(ii)). The /y/ of the following
suffix coalesces with the /-n/ to form /-n/, and thus the /y/ does not appear. However,
the adverbialized form only rarely occurs. The forms that are nominalized from the
adverbialized forms are very much more common. Therefore, I list only those in the
examples below:
(1103) t-maya-n-e-m
ADVZR-wildness-POSN-ADVZR-NOMZR
One that is wild' (No noun root known for this word.)
(iv) The affix set ti-. .. -ke has the same meaning as the previous sets.
Adverbialization:
Nominalization:
(1107) c-epethi-ke-m
ADVZR-payment-ADVZR-NOMZR
'a costly one'
t-ar-ke-m
ADVZR-pregnant-ADVZR-NOMZR
'a pregnant woman'
(v) The affix set ti-. .. -pe has the same meaning as the previous sets. I have only
two adverbs of this set recorded.
Adverbialization:
(vi) The suffix -pe occurs without a prefix on several stems. They are apparently not
noun stems, since they take nominalizing suffixes. But I have included them along
with noun stems, thinking possibly they were either nouns or particles at some time
in history. See sect. 22.6.4-5 for the reduction of /p/ to /h/ and loss of /h/ in ex.
(1111).
Adverbialization:
Nominalization:
(vii) The suffix -me also occurs without an accompanying prefix. Compare the
postposition me 'ADVZR' (sect. 17.2). I have postulated this syllable -me as a suffix
on a few nouns and one pronoun because it occurs very often with them and it is
pronounced without any intervening pause between so that it seems to be one word.
Adverbialization:
Nominalization:
(viii) The affix set ti-. .. -xe also adverbializes noun stems.
Adverbialization:
Nominalization:
(ix) The affix set ti-. .. -we also adverbializes noun stems. I know of no pure nouns
with which to compare some of these stems. But there is a negativized form and a
verb form in which the basic stem functions like a noun stem by virtue of the
derivational suffixes that co-occur (see the first two examples in (1116)).
208 Hawkins
Adverbialization:
Nominalization:
(x) The suffix -ka\v adverbializes a few nouns, mostly parts of the body. (Cf. the
adverb kaw 'tall, high, long', sect. 20.) I postulate these examples as one compound
word rather than two separate words because in the third example in (1118) the
suffix of the noun is lost when followed by the adverb kaw.
Adverbialization:
Nominalization:
(xi) The affix set /-. .. -kaw adverbializes stems which are based on noun stems. The
meaning is 'stretched out/up.'
Adverbialization:
Nominalization:
(1123) t-pana-caf-kaw-no
ADVZR-ears-?-long-NOMZR
One having long ears'
(1124) ti-h-kaw-no
ADVZR-head-high-NOMZR
One having his head raised'
(xii) The suffix -pene also derives adverbs from nouns. It indicates that a person or
animal possesses an item that is very large, or possesses a large quantity of an item.
It does not occur with the adverbializing prefix if-, but it does take the general prefix
yi- when the phonemic structure is right (sect. 23.4.5.1(ii)). Note that the
nominalizing suffix requires a change in the final vowel of the adverbializing suffix.
Adverbialization:
Nominalization:
(xiii) The suffix -ciciy/-ciciye indicates that a body part, or part of any item, is
prominent. These forms also occur without the u- adverbializing prefix. The longer
variant occurs before the nominalizing suffix.
Adverbialization:
Nominalization:
(xv) The affix set ti-. .. -po + -re indicates that an item or person that one
possesses is good. (Cf. the same set of adverbializing suffixes with verb stems, sect.
23.5.1.1(iii). This form is often nominalized by the suffix -mu/-m.
(xvi) The negative of the above forms is formed by the suffix set -po + -ra for the
adverbialized forms and by the suffix set -po + -n for the nominalized forms. It
should be noted that the possession suffix -n also occurs preceding these suffixes
with the set of nouns that take that suffix in the positive forms (1 1 36, 11 37). Nouns
that take the possession indicator -rf do not occur with a possession suffix before
this suffix (1138).
(1136) Pono-n-po-ra
clothes-POSN-good-NEG + ADVZR 1 S-be-SF-ÃÍÑ
º have clothes that are not good.'
o-pono-n
(IPOSR-clothes-POSN)
'my clothes'
o-kanawa-ri
(lPOSR<anoe-POSN)
'my canoe'
(i) The suffix -hra occurs with the stem of a few underived adverbs and indicates
negation. The word class is not changed.
The negativizing indicator -hra also occurs with derived adverbs and is added to
the basic noun stem of the derived adverb except in the word for 'strong' (1142).
212 Hawkins
(ii) Collectivity of adverbs is indicated by the particle so. I have labelled this
indicator as a particle rather than a suffix because it occurs with transitive verbs,
with adverbs, and with postpositions.
23.6 Postposition morphology. There are suffixes added to noun stems to derive
postpositions and to give a specific indication of place. Most of the noun stems
indicate parts of the body. I can find no way to predict which suffixes will occur
following which stem. See sect. 17.3 for syntactic functions of these postpositions. I
will list the noun here first for easy comparison:
(i) Nouns which take the suffixes -w, 'at, on' -ka, 'to' -ri 'along' and -y 'from':
(ii) Other nouns take the suffix -ta 'LOG' followed by the suffixes -w, -ka and -y. It
seems to me that the suffix -ta indicates location in general. It does not refer to
location in time when suffixed to nouns, as it does when suffixed to verb stems.
(iii) There are a few nouns which take the suffix -a 'LOG' following the nominalizer
-n to indicate location in general. The -a is then followed by one of the same three
specific location suffixes -w, -ka and -y. These derived forms, however, are adverbs
rather than postpositions, since they do not co-occur with prefixes or noun objects.
I list them here because of their similarity to postpositions.
Ideophones
24 Ideophones
Ideophones are vivid words. They are used to make stories and reports more
interesting. They are usually uninflected and often occur in isolation (1168-1169).
When occurring in isolation they are usually said with strong emphasis, almost like
an interjection. They are the only class of words in Wai Wai which are never
followed by the other classes of particles. Ideophones are used to indicate or
describe actions. They may be followed by the corresponding verb form (1161-
1163). Sometimes there is reduplication of a syllable within the ideophone (1169).
Ideophones often function as the object of the verb ka/kas- 'say, do' and thus are like
quotations (1157, 1164, 1166). They may be followed by the postposition wa/wara
which normally follows nouns (1159). Many of them are onomatopoeic, but others
have nothing to do with the sound of an action (1159, 1161, 1162, 1165, etc.).
Ideophones are usually expression initial. I have recorded 75 or 80 of these words.
There are probably a good many more. I list a few of them here.
(1168) Powa.
full
'It got full.'
216 Hawkins
(1169) Kupruruniru.
to.go.in.a.line
'They went in a line.'
There are a few ideophones which depart from a normal phonological rule of the
other word classes, that is, that the phoneme /t/ never occurs before the phoneme III.
(1178) Tantiri.
break (slender object)
'It broke.' (the stick)
Wai Wai 217
There are a few ideophones that take verbalizing suffixes such as those used to
verbalize nouns. This might tend to indicate that ideophones are basically nouns. But
I have only discovered three such verbalized ideophones. Thus I feel it best to leave
ideophones as a separate word class.
Abbreviations
References
Derbyshire, Desmond C.
1985 Hixkaryana and linguistic typology, Summer Institute of Linguistics
Publications in Linguistics 76 (Dallas: Summer Institute of Linguistics
and University of Texas at Arlington).
Hawkins, W. Neill
1952 A fonologia da lingua Uaiuai, Boletim 157, Emografia e Tupi-Guarani
25 (Säo Paulo: Universidade de Säo Paulo).
1962 A morfologia do substantivo na lingua Uaiuai, Publica^oes Avulsas 21,
Museu Nacional (Rio de Janeiro: Universidade do Brasil).
Hawkins, W. Neill, and Robert E. Hawkins
1953 "Verb inflections in Waiwai (Carib)," International Journal of Ameri-
can Linguistics 19:201-211.
Wai Wai 221
APPENDIX
Told by Forosha
0-c-e-tkene.
3S-go-SF-UP + COLL
'All right, this is what I will say to you my brother, while you were
here people went to the forest.'
noro pen.
3PRO deceased
'This is how he was when he came from the forest, after he came, he
dawned (passed the night), he dawned, then he got sick (like a corpse).'
enexa-nT-nho rma.
from.there-NOMZR-PAST PROX
'Afterwards he was weak (sick) alas right after he came back from
there.'
amna.
1 + 3PRO
'"I sure am weak my brothers," he said to us. "Alas," we said.'
okwe.
alas
'We lost our helper alas.'
Introduction
COLOMBIA V
÷·
San Carlos
de Rio Negro
VENEZUELA
Santa Rosa de
Amanadona
Sao Gabriel
da Cachoeira
Tariana
East-Tucanoan
Warekena 227
beginning of the last century. Towards the beginning of the twentieth century
the majority of Warekena had moved to Venezuela, where they switched to a
different language - Baniwa of Guainia. In spite of having changed their lan-
guage, these Warekena preserved their ethnic name and the story of their
origins from Xie. Some of them moved back to their "proto-home" on the Xie
river in the early 1920s.
The only previous work on Baniwa of Guainia (which displays minor dialec-
tal differences from Warekena of Xie) is: (a) short word lists in Cr6vaux et al.
(1882), Grasserie (1892), Civrieux and Lichy (1950); (b) a short grammatical
sketch in Grasserie (1892) and a little morphological data in Mosonyi (1968).
1 Constituent order
AVO:
(1) wa-hä waji yutfia-ha ema
then-PAUS jaguar kill-PAUS tapir
'Then the jaguar killed the tapir.'
230 Aikhenvald
VS0:
(2) jupe-he Jiani-pe
many-PAUS child-PL
'Children are many.'
SaV:
(3) peya nu-yalitua wiyua
one Isg-brother die
One of my brothers dies.'
SioV:
(4) nu-yue mawaJi
Isg-for hungry
º am hungry (Lit.: For me-hungry).'
Demonstrative adverb ale 'so, thus' is always placed before the predicate,
e.g., (8). It behaves as an independent proclitic (see sect. 22.4.1) and may form
a single phonological word with the following predicate.
(8) ale-wayata
thus-speak
'Thus he spoke.'
Warekena 231
Usually a clause contains only one oblique argument or indirect object. In the
rare examples of two oblique arguments in one clause the locative one, as in
(10), or the temporal one, as in (11), always follows the other oblique.
Left dislocation of arguments into the pre-predicate position is discussed in sect. 9.1.
2 Parataxis
nu-miJiuta ijiwa
Isg-hide from.him
saw Yamadu (evil spirit), I hid from him.'
3 Ellipsis
ya-mia-be-pia-wa malayu
NEG-PERF-can-NEG-NONACC deer
'He (deer) fainted, he could no more, the deer could (run) no more.'
234 Aikhenvald
Juduna wa ni-ja-tj"i-wa
come.down then 3pl-go-REP-NONACC
'They (the abandoned children) went until they arrived in the jungle.
The (male) child climbed on the tree to see where is the hill. He climbed
and saw nothing. Then he came down, he came down, they went again.'
3.2 Ellipsis and pivot restrictions. Warekena has a mixed pivot (Dixon 1994)
which combines the properties of an S/O and an S/A type. The Equi-NP deletion
in clause coordination (or predicate coordination: see sect. 8) tends to operate in
terms of an S/O pivot, if the two coreferential constituents are 3rd person singular
(both either feminine, or non-feminine), or both plural. There are no pivot restric-
tions for the 1st and 2nd person subject, as in (26). This can be accounted for by
the fact that a 1st or 2nd subject cannot be omitted, in the sense that it is
Warekena 235
(i) purposive clauses (sect. 14.2), temporal clauses (sect. 14.4), locative clauses
(sect. 14.5);
(ii) serial verb constructions operate in terms of the same subject constraint
kalana tepa-ba-mia-lehe
caranlpalm hard-AUG-PERF-ADJ + PAUS
'Caranä palm is not good to make hammock; (it) is too hard.'
Impersonal is often used in serial constructions of the modal type (sect. 18.8).
kalana-tua-li bitsa-ha
carana-made.of-ADJ hammock-PAUS
One cannot make a hammock of carana-palm.'
In serial constructions of modal type, if ba- is used on the first predicate, Ipl wa-
can be used on the second one, since the impersonal marking is often understood
as close to the first person, as is the case in many other Maipuran languages.
Impersonal marking is often used with the verb eda 'perceive, see/hear,'
similarly to a negative existential (see sect. 12.3):
wala-paJu pe-pitj"i
cook-PURP 2sg + eat-OBJ.FOC
' "Good (said the rabbit), I shall go with you to your house, to cook your
food (as someone to cook your food)" (said the rabbit to the jaguar).'
Intransitive verbs which refer to physical states usually do not take cross-ref-
erencing affixes, and can only be used impersonally:
5 Passives
6 Causatives
mawa-h aji
go.out-PAUS fire
"Then they (Yanomami) let it (a corpse) stay one night (in the fire), then all
is burnt, then they let the fire go out.'
Both orders seem to be in free variation, and are rather rare, since Warekena,
similarly to other Northern Maipuran languages, avoids expressing overtly two
arguments of a transitive verb in one clause. The underlying A is very fre-
quently omitted (51, 52).
(53) wa Ja we ni-yuluta-wa
then go leave 3pl-lie-NONACC
'He (deer) was going to leave them (turtles) to lie.'
Syntactic causatives can be formed with Sa verbs (45, 48), S0 verbs (47), and
transitive verbs (49), unlike morphological causatives which cannot be formed
on transitive verbs. The main semantic difference between syntactic and mor-
phological causatives of intransitive verbs is that the morphological causative
implies direct causation, and the syntactic causative implies indirect causation:
The verbless locative clauses can have the meaning of 'appearance' (63):
Warekena 243
wa Jimapie-pe-mi ema
then bone-PL-PEJ tapir
There is no food, there are tapir's bones.'
Equational verbless clauses, also used for pointing, consist of two juxtaposed
NPs. The first of them occupies the subject slot, and the aspect clitics can occur
with it if the clause has a non-present reference. This is illustrated with the
following examples:
Attributive clauses have an S0 verb in the predicate slot. S0 verbs are itali-
cized in (69):
244 Aikhenvald
Attributive clauses with the So predicate papeya 'similar, another' are used
to express similarity:
(a) possessive verb deka 'to have' is used when possession is in focus:
kulua-palu namali
suck-PURP people
'He (the Evil spirit) has a tail, he has a tongue to suck people.'
(c) double subject construction, when the possessor constituent is the topic of
the discourse. There are two types of double subject construction, illustrated
below. The one contains an attributive type clause, with an So predicate (75).
The other contains an existential (verbless) clause (76).
8 Coordination
A:
(90) wa-ha waji weya yutjia-ha ema
then-PAUS jaguar want kill-PAUS tapir
'Then the jaguar wanted to kill the tapir.'
S:
(91) wa: pala-mia-ha ema
then run-PERF-PAUS tapir
'Then the tapir was running.'
O:
(92) wa: pala-mia ema pauta-ha napitu kulimalu
then run-PERF tapir step-PAUS back turtle
ilike-na-mia-wa
be.buried-REFL-PERF-NONACC
'Then the tapir was running, he stepped on the turtle's back, and (it) was
buried.'
(94) comes from the story similar to the one about Hansel and Gretel. A snake
had come up to the boy (already a grown-up man); man is the topic and left-dislo-
cated:
ya:me, ni-Ja-wa
far 3pl-go-NONACC
'Then they(the old man and his children) went by the road, by the road,
they went, far into the jungle, far, they went.'
wa(-hä) is used to start a narrative, introducing the theme of the future narrative:
wa(-hä) can be used to mark the important result of an action, as in (100), which
follows the description of how the two dogs and a man with a knife were
attacking the snake (from the same story as (99)).
ya-eda-pia-na
NEG-see-NEG-lsg
am afraid of him (the evil spirit) until when he has gone, he did not
see me.'
When wa(-hä) is used to resume a completed stretch of the text, it can appear
in a sentence-final position. This occurs frequently in repetition.
kulimalu wa-hä
turtle then-PAUS
'Turtle arrived to where (he) killed the tapir, turtle arrived to where
(he) killed the tapir, so the turtle arrived.'
Ja-wa ja-mia-tfi-wa
go-NONACC go-PERF-REP-NONACC
'He (the jaguar) woke up, and came down (the tree); he started going,
he went again.'
aliwa-mia-wi wa wa-Ja-mia-wa
dawn-PERF-lpl then lpl-go-PERF-NONACC
'Then we slept until it dawned, it dawned on us, then we went.'
In (117) the repetition of ema-tfi-wa and ema is used to indicate the prolonged
action. The pausal form is used on the last occurrence ema 'he cried', to sum
up what was happening.
9.4 Cleft construction, left and right dislocation. Warekena uses a number of
different devices to mark discourse prominence. Among them is the positioning
of verbal aspectual markers, which depends on the focus of the narrative (see
sect. 18.8.1), and cross-referencing, which is used to signal the cleft construc-
tion (also see sect. 18.4.2). The most frequently used devices are left dislocation
and cleft construction. Left dislocation of a constituent is used to attract the
hearer's attention. The left dislocated constituent may take a pausal form and is
followed by a pause:
A left dislocated constituent can be repeated in the clause, and then the left
dislocation is used as a kind of a stronger reminder, as is the case with tsinu
'dog(s)' in (119).
ni-api-mia-ha tsinu
3pl-take-PERF-PAUS dog
'Then they (the abandoned children) went again. The dog(s), they took
the dogs.'
Warekena has a cleft construction characterized by: (i) left dislocation of the
subject; (ii) -Ji 'relative' marker on the verb; (iii) suppression of A/Sa cross-refer-
encing prefixes on the verb. A cleft construction is used to mark subject (A/Sa) in
a contrastive focus. Independent personal pronouns most often occur in this func-
tion. No other constituent can be inserted between the subject and the verb, as
shown in (120), where the cross-referencing prefix of 2nd person sg is absent from
the verb -tse- 'to know', since the subject piya 'you' is topicalized and fronted. The
left-dislocated subject can attract the aspectual clitics, as is the case in (120).
(127) follows (126). It shows the optional character of the pausal form:
The neutral sentence, where namaJi 'people' is part of the main predication,
would be:
9.5.1 Emphasis. The clitic -Ju is used to mark a very strong emphasis. It can
occur on any word class.
-Ju on a verb may mean intensifying the action expressed by the verb. It can
be used with an S0 verb, as in (133); an Sa verb as in (134); a transitive verb as
in (135). In (134) 'emphatic' means shall indeed go'. This example comes
from the story about the abandoned children; the sister expressed her definite
decision to go and see the witch.
yajapua
night
'Here they went to talk until it was late at night.'
258 Aikhenvald
waya-ne Diutsu
speak-POSS God
'He (Jesus Christ) told them the word of God, really much, how we live
here in the world, all what they really knew, they knew of the word of
God, they knew the word of God, some (of them), some men really knew
the word of God, some women really knew the word of God.'
-Ju can occur on a nominal modifier, as in (138), (139), and it can co-occur
with -Ji 'relativizer/adjectivizer', in which case it precedes -Ji:
wa-wenita-wa
Ipl-buy-NONACC
'Then they (rich white people) sell (things) to us, it is expensive, they
sell to us expensively, and we do not have much money to buy very
expensive (things), we cannot buy very expensive (things).'
In repetition, forms with -Ju and without it can follow each other:
-Ju can be used on a locative constituent. (145) comes from a text given by
the speaker in response to my insistent questions about whether his family
originated from the Xie, or they had come from Venezuela. The use of -Ju bears
a distinct nuance of contrast:
The following example illustrates the clitic character of -Ju, since it under-
goes attraction to the negative proclitic ya- (see sect. 22.4.1):
9.5.2 Cleft constructions. -Ju can be used on a clefted constituent. This kind of
cleft means emphasis 'just and only' (unlike cleft with relative forms in -Ji, which
implies contrastive focus). The cross-referencing prefixes are then suppressed from
the predicate, but the relative -Ji marking on the verb is not used:
Warekena 261
10 Interrogatives
Warekena has no special segmental marker for polar (yes/no) questions. They
are marked with a special rising intonation, and the verb occupies the
sentence-initial position:
(150) aJiwa-pi?
a wake-2 sg
'Are you awake?' (a morning greeting)
In content questions, the question word (interrogative pronoun; see sect. 16)
occupies the sentence-initial position. Relative -Ji marks the predicate when the
question constituent is the subject (//>' 'what, who', damaM 'who, which one'),
or the object (ifi 'what'), in the same way that it does for subject or object
relativization (sect. 14.3).
The relative form -Ji can occur with the predicate of a complement clause
which contains an oblique question to the subject constituent:
weluami Diutsu
1 pi + father God
'Then the people said to him (Jesus Christ), "Good, teach us, we want
to know where our father God lives".'
Warekena 263
11 Imperatives
(162) manuma-pi
be. silent-2sg
'Be silent!'
(163) pi-manuta-wa
2sg-stop-NONACC
'Stop (start stopping)!'
Second person imperatives are the most frequent ones; the examples of third
person imperatives are shown below. The predicate is in the clause-initial
position. It can take an aspectual clitic, e.g., -yaJu 'yet'.
A serial construction with a motion verb -fa 'go' can be used in a cohortative
sense:
12 Negation
In equative clauses (sect. 7), the negation ya-. . . -pia occurs on the predicate:
In syntactic causatives and serial verb constructions, there is only one nega-
tion, which goes on the first component and negates the whole verbal phrase,
e.g., (177) (syntactic causative), (178) (serial verb construction). This is one of
the criteria for analyzing serial verb constructions as one predicate (sect. 8.8.1).
yutjia-li mawaya
kill-REL snake
'It was not he who killed a snake, another man killed the snake from me,
it was not the soldier who killed a snake.'
The sentential negation (sect. 12.5) can be used for more emphasis, cf. (180)
from the same story:
benejl mawaya
NEG + what snake
º stayed (there), I got tired, I did not see any snake.'
bena-iji nu-yue-he
NEG-what lsg-to-PAUS
º did not have either manioc bread or flour, I had nothing.'
Warekena also has a negative prefix ma, which is not productive, and is
attested in the following cases: two S0 verbs ma-kale 'be out of breath, breath-
less', ma-numa (NEG-mouth) 'to be silent, mouthless', and one Sa verb ma-kuta
'disappear' (cf. Tariana -kuda 'appear, put on light', ma-kuda 'disappear, switch
off light').
268 Aikhenvald
Possibly, this is the same prefix which appears in the preventive (warning)
mood marker ma-tse 'lest', discussed in sect. 18.3.
wa yu-ma-hä ya-hä
then 3sgf-say-PAUS NEG-PAUS
' "Did the snake not come?" (asked the father). Then she said, "No (it
did not)." '
In emphatic commands, negative proclitic ya- can be used on the negated verb:
13 Anaphora
The most frequent form of anaphora is the omission of a noun phrase, which is
retrievable from the context and the pivot restrictions (sect. 3). The A/Sa and
O/So cross-referencing markers have anaphoric functions.
Demonstrative and third person pronouns also can have anaphoric functions:
270 Aikhenvald
As is shown in sect. 18.4.2, when the A/Sa constituent is focused and omitted
from the surface structure, the cross-referencing prefix /- '3sgnf is used
anaphorically.
14 Subordinate clauses
The A constituent of the main clause may follow the complement clause.
272 Aikhenvald
munduwe
world + LOG
Our knowledge is nothing here in the world.'
The coreferential constituent can be the S of the main clause and the S of the
relative clause. This type is comparatively rare. In (215) the shared constituent
is the Sio of the main clause and the Sa of the relative clause. The predicate -yue
'to, for; to belong' is of an Si0 type.
The coreferential constituent can be the Ï of the main clause and the S of the
relative clause. It is S0 in (216) and (217):
The coreferential constituent is the Ï of the main clause and the Sa of the
embedded relative clause in (218) and (219):
ni-yalitua-mia-h [wiyua-mia-li]
3pl-brother-PERF-PAUS die-PERF-REL
'Then they make much fire, they throw him into the fire, their brother
who died.'
Relative clauses may also be used headlessly (see sect. 15.4 on the headless
use of modifiers). In negative clauses, -Ji is attracted to the negative proclitic
ya-, in agreement with the clitic character of -Ji.
276 Aikhenvald
(229) ya-li-be-pia
NEG-REL-can-NEG
'a weakling, a coward (lit.: the one who cannot (do things)'
Headless uses of -Ji relative forms are also considered in sect. 18.6. (231) is
frequently used at the end of stories.
In this case the subordinate clause may be introduced by the locative subor-
dinate conjunction waJi 'where':
In a few cases, the predicate of a relative clause of this type has a case
marker, as in (234). It is not a locative subordinate clause, since the predicate
of locative subordinate clauses (see sect. 14.5) is not marked with -Ji.
Warekena 277
mina-liana pani-Ji
body-MASC house-NPOSS
'Then we went away from there on the boat (and with) a paddle bor-
rowed from the owner of the house.'
nuluami Diutsu
lsg + father God
'All we see is the creation of my father, God.'
wani ni-yuleta-mia-wa
here 3pl-return-PERF-NONACC
'Having killed one tapir and one deer, they returned here.'
nu-yue-li] ya-mia-yutjl-pia-na
lsg-to-REL NEG-PERF-strong-NEG-lsg
'Having no manioc bread, no flour, nothing, I am not strong.'
The predicate of a converbal clause can have an aspect marker -wa, which
follows -Ji:
Warekena 279
Converbal clauses may be used to emphasize the action expressed by the verb
in the main clause:
ya-yue-pia pina-Ji
NEG-to-NEG cure-NOM
'Whatever another person may say to me, there is no antidote (against
snake bite).'
14.4 Temporal clauses. The predicate of a temporal clause is marked with the
clitic -Ji-bena (see sect. 18.6.2). Coreferential deletion operates in terms of an
S/A pivot. The order of subordinate and main clause depends on the meaning of
the predicate of the subordinate verb. When the subordinate clause is postposed
to the main, the action of the subordinate clause precedes the action of the main
280 Aikhenvald
ukapi-wene-he
room-EL-PAUS
'They (the menstruating girls) eat all, smoked meat, flour, manioc bread,
they eat all, as soon as/after they come out of their room (ritual seclu-
sion).'
yanumami-nawi
Yanomami-PL
'So they did (performed) long ago, after Yanomami people had died.'
jiani-pe
child-PL
' "I am digging (a hole) far away, it will go out far away, for me to
escape from the jaguar, after/because I killed her children," (said the
rabbit).'
Warekena 281
ni-yeleta-li-bena-ha nelima-pe
3pl-arrive-REL-when-PAUS cousin-PL
'They went to tear manioc to make manioc bread to eat, to make manioc
bread, manioc flour, tapioca to eat, smoked meat, after the cousins
come.'
The -Ji-bena construction is rarely used when the complement clause has an
inanimate S, and the main clause has an animate A/S. Then parataxis is used,
as in (256). An example of -Ji-bena is (257).
14.5 Locative clauses. Locative subordinate clauses are expressed by the sub-
ordinating marker waJi 'where' (which etymologically may go back to \va
'presentative' and -Ji ' relativized). They can refer to either location or direc-
tion. Locative subordinate clauses follow the main clause, and the head is
usually omitted from the main clause. The constituent order is waJi V Sa (259,
260), waJi VS0 (261), waJi V Ï (262, 263), and waJi VOA (264). In coreferen-
tial deletion, an S/A pivot operates (263).
corroborated by the fact that they are found only in repetition, as in (265). The
left dislocated A constituent is italicized.
In one case a coreferent locative constituent atapi 'tree' was not omitted from
the main clause:
Wali may take a perlative case-marker -wa, as in (267). This example is also
unusual because the predicate of the locative clause precedes the subordinator
wali.
yu-yawaluta wali-wa-ha
3sgf-go.round where -PERL -PAUS
'Then they (the abandoned children) pushed her (the witch) where she
was going around (the pan with boiling water).'
14.6 Terminative clauses. Terminative clauses are marked with the loan sub-
ordinate conjunction ate, te 'until' (used with a temporal or spatial meaning).
They usually follow the main clause. Coreferential deletion in terminative
clauses functions in terms of an S/O pivot (268-271).
O:
(272) wani ni-mutjita-mehe ni-mutjlta-mehe
here 3 pi -bite -PERT + PAUS 3pl-bite-PERF + PAUS
S0:
(273) kunehu ipeta-ha waj! ate ya-mia-yutji-pie waji
rabbit beat-PAUS jaguar until NEG-PERF-strong-NEG jaguar
'The rabbit beat the jaguar until the jaguar had no more strength'.
amuji
sun
'Then he stayed until the sun was high.'
Sa:
(275) kunehu ipeta-mia ipeta ipeta ipeta ate dalina-mia waji
rabbit beat-PERF beat beat beat until faint-PERF jaguar
'The rabbit beat the jaguar until the jaguar fainted. '
wakwena-mia-wa wakwena-mia-wa
untie-PERF-NONACC untie-PERF-NONACC
'The monkey bit it (the rope) until the rope tore, it (the rope) got untied.'
te, ate 'until, up to' can also be used as a preposition, possibly under Portu-
guese influence:
anetua-li
good-ADJ
'So I am evangelical, until now I live well.'
Ate and waJi can co-occur forming a compound conjunction 'until where'.
14.7 Oblique questions. Oblique questions (also sect. 16.5) are introduced
with the help of interrogative pronouns. They usually follow the main clause,
like complement clauses. The predicate has no special marker.
15.1 Marking for case. Similarly to the vast majority of Maipuran languages,
Warekena does not have any case marking on core constituents (i.e., A, S or
in terms of Dixon 1994), but has a number of cases used to mark peripheral
constituents. This is also characteristic for other Northern Maipuran languages
of the region, e.g., Bare, which is structurally very close to Warekena, and
Baniwa of Igana. Peripheral case markers usually occur on the head of posses-
sive, adpositional or adjectival NPs. Warekena has the following peripheral
cases.
Locative is marked with the suffix -iwe. Its locative meaning is illustrated
with (289, 290). (289) shows an opposition between elative and locative.
pani-Ji wenejewe-he
house-NPOSS Xie +LOG-PAUS
'Pedro arrived from Venezuela (and) made a house on the Xie river.'
There is, possibly, another locative affix, -tfi/fi (on the variability of tf and J
in Warekena see sect. 22) On', which has restricted occurrence . It occurs with
body parts, e.g., nawa 'my leg', nawa-fi On my leg'; i-tfipaJa 'his foot', i-
tfipaJa-fi On his foot'; certain nouns referring to location, e.g., uJupe 'foot' (of
a mountain), fawape 'jungle', aJipeJi 'cerrado; dense jungle', and interrogative
da-tfi 'where?'. Possibly, this affix is also present in the adposition mina-fi On'
(290), cf. mina 'body'. These nouns, illustrated in (296)-(299), usually do not
take locative case suffix -iwe; tawape-ffiwe in (220) is an exception. Directional
case marker -waba and perlative/transformative case marker -wa can go either
on the noun stem, as in (297), or on the noun stem which contains -tfi, as in
(299). The occurrence of different vowels in two otherwise identical forms
containing -tfi + -wa in (299) is explained in sect. 22.4.5.
aJipe-li-tJ"i-wa-ha ya:me-lu
thick.jungle-NOM-LOC?-PERL-PAUS far-EMPH
The difference between directional case and locative which combines direc-
tional and locative meaning, as in (291), is that the latter means 'go towards a
location and stay there', whereas the former means 'go towards a location'.
Elative is marked with suffix -wene. It means both 'from', as in (302), and
'from out of, as in (303).
290 Aikhenvald
anepili-wene-he
right-EL-PAUS
'Then they tore him (the bad soldier) apart, one horse took one leg from
the left, another horse took another leg from the right.'
Occasionally the vowel a of the noun root becomes e before -wene 'elative',
as in numewene 'from the mouth', cf. numa 'mouth'. This phonological process
(see sect. 22.5) could indicate that the underlying form for the elative suffix is
-iwene. However, the form numa-wene 'from the mouth' and such forms as
wa-wene 'from there', da-wene 'from where?' indicate that -wene is the under-
lying form. The variation of numewene and numa-wene may be due to an
analogy with locative -iwe, cf. numewe (numa-iwe) 'in the mouth'.
awipemi-waba-ha wabupi
headwaters-DIR-PAUS spring
'The jaguar ran by the shore of the spring to the headwaters of the
spring.'
Warekena 291
-Wa may also mean 'until' with a terminative verb, or a verb marked with
perfective aspect.
pani-wa ya:japua
mid-PERL night
'Then they sat down to turn the smoked meat until midnight.'
(c) Other meanings of -wa are: 'for' (310, 311); and 'as, in order to acquire a
quality of (312-314).
nu-pata-palu puli
Isg-get-PURP game
'Then I killed the stingray for my bait to get game.'
-Wa case is not used to mark essive constituents; the unmarked form of the
noun is used :
panifi 'house' is used with the locative case marker when it has a more
specific meaning, as in (317), where panifiwe means 'at home'.
Nouns unmarked for case can be used in the corresponding meaning if the
NP has already been marked with a case in the preceding clause (318), or if the
meaning is clear from the context e.g., i-fuumiawa 'his wife' in (319).
ni-tapa-ha mana
3pl-walk-PAUS near
'Then they (pigs) arrived near him (the jaguar), the pigs arrived, they
walked near (him).'
Occasionally, nouns which form their possessed forms with -Je, also have a
form in -te, and the semantics of the opposition of the two is similar to that of
-ne vs -te; e.g., nu-muJupa-Je 'my boat, my boat right here', pi-muJupa-te 'your
boat over there'. This is rare, and it looks like a regularization, probably to be
interpreted as a language death phenomenon8.
The semantics of possessive suffixes in Warekena may shed some light upon
the semantics of noun grouping in accordance with the possessive suffixes they
combine with in Maipuran languages in general (see Payne 1991:378).
Possessive NPs in Warekena are formed by juxtaposition of the components
and display the constituent order Possessed-Possessor which is highly unusual
for Maipuran languages. This is the unmarked order in 'part-whole' construc-
tions, as shown in (320)-(323) . The inherent gender and number of the posses-
sor are marked on the possessed noun, as illustrated in (322):
When a genitive construction bears the meaning of 'made out of, the inverse
order Possessor-Possessed is preferred, as illustrated in (324, 325) below.
Usually an adjectival NP is used to convey the meaning 'made out of, and
the compound suffix -tua-Ji is used on the noun:
In conversations, numa can also be used in the genitive constructions with the
order Possessed-Possessor:
Modifiers can refer to any part of the genitive construction. They are adja-
cent to the constituent to which they refer.
15.2.2 Gender. Warekena, like most other Maipuran languages, has two gen-
ders, feminine and non-feminine (or masculine). Gender is marked in cross-ref-
erencing prefixes, enclitics and independent pronouns of the third person
singular (see Table 1 in sect. 15.2.1 and Table 3 in sect. 15.3.4). The gender
opposition in singular demonstratives is maintained only in the Anamoim dia-
lect. There are no gender distinctions in plural. For animate beings, masculine
and feminine genders relate to sex distinctions. Nouns with an inanimate refer-
ent are treated as belonging to masculine gender. Thus, masculine gender can
be considered as unmarked.
The opposition of two genders is maintained in a few nouns, and the deriva-
tional affix -yawa or -wa is used to mark feminine; e.g., ete-ne Old man',
298 Aikhenvald
Plural suffix -pe can have collective meaning, e.g., fiani 'child', fiani-pe
'children, family'; fabine 'family, household', fabine-pe 'all the members of a
household'.
If the noun contains either of the following two homophonous suffixes, the
plural suffix -pe precedes -mi:
-mi 'human' (non-productive): ena-mi 'man, person', pi. 'men, persons' ena-
pe-mi, nuJua-mi 'my father', my father's brother', pi. nuJua-pe-mi 'my father's
family; my father's brothers';
-mi 'pejorative' (productive): fimapie-mi 'a thrown away or left over bone',
pi. fimapie-pe-mi 'thrown away bones' (345).
Warekena 301
(ii) Plural marker -ne is used with nouns with animate non-human referents,
e.g., tfinu 'dog', tfinu-ne 'dogs', and a few other nouns, e.g., akayu 'year',
akayu-ne 'years' (250).
(b) Emphatic plural is marked with -nawi, used with countables (human as well
as non-human) and uncountables. It denotes 'very many indeed, a whole group
of, as in (346), (347) or 'many members of different groups', as in (348), and
can be used with animate and inanimate nouns. -Nawi may be related to Baniwa
and Tariana nawiki 'people'.
(c) Collective plural is marked with -natfi ' collective inanimate', e.g., ipa-
na$i 'a lot of stones, a set of stones'; minapi 'banana tree', minapi-natfi 'a bush
of banana trees'. -Natfi can also have a locative meaning, e.g., weni 'water',
weni-natfi 'much water, a place with much water'; witfi 'wind', witfi-natfi 'a
windy place' (see sect. 15.4.1).
-peJi is used to mark collective plural on nouns with an animate non-human
referent:
Warekena has double plural marking, when the plurality of a referent has to
be particularly emphasized, as seen in the forms for 'women' and 'men' (in
italics) in (350).
302 Aikhenvald
One of the speakers gave the following 'hierarchy' with respect to the
meaning of simple plural, emphatic plural and double plural:
When a plural referent has once been introduced, the plural marker is not
repeated on further occurrences of the same referent:
A few nouns, e.g., namaJi 'people, person', and names of tribes, usually do
not take a plural marker (352).
A noun marked with -nawi triggers plural agreement on the verb when it is
in A/Sa function (353). The plural marker -nawi is used if the plurality of
referents has to be stressed. NamaJi-nawi means 'very many people', as in
(353), or 'very many different people, peoples', as in (354).
Warekena 303
y anumami -na wi
Yanomami-PL
'So they behaved long ago, when the Yanomami people died.'
Plural agreement with quantifier fupe 'many' is more frequent with nouns
with human referents; cf. (356) with (357) and (358).
Plural agreement is optional with the quantifier payaJu 'all' (see sect.
15.3.2). If it is introduced once, it is not repeated (93, 350). Plural is generally
not used with the interrogative quantifier ipeJi 'how many, how much':
(360) ipeli
ipcj.i pi-tani-hi
jji-iaiii-iii
how.many 2sg-child-PAUS
'How many are your children?'
a-tfiabu-ni 'big-bellied' (see sect. 15.4.1, on the prefix a-}. A few adjectives
contain a non-productive suffix -mi and no other adjectivizer; they also do not
distinguish between attributive (365) and predicative (364) use. These adjec-
tives are: awaJapefia-mi 'poor', Uua-mi 'big', puJapia-mi 'thin'.
wa-wenita-wa
Ipl-buy-NONACC
'We cannot buy expensive (things).'
An NP can contain more than one adjectival modifier. Usually both follow
the head noun, and only the last one is marked with the adjectivizer -li.
Dimension adjectives tend to be placed closer to the head (374).
Examples of this type are rather rare. All the speakers seemed to prefer to
repeat the head noun (375).
ate yaliwa
until now
'Now I am evangelical until now, I live as a good (man), as a good
(man) I live, well I live up to now.'
Nouns can be occasionally used as modifiers. Then they follow the head. The
only nouns which have been attested in this function are the nouns with refer-
ence to gender: neyawa 'woman' and enami 'man', mainly used with kinship
nouns. I suspect this is an innovation, and these nouns are used as modifiers
when the actual gender form of a corresponding kinship term has been lost, e.g.,
nu-matuimihe enami 'father-in-law', nu-matuimihe neyawa 'mother-in-law'; nu-
tani enami 'a son' (lit.: child-man), nutani neyawa 'daughter' (lit. child-woman).
The modifier follows the head noun, and if it refers to the possessee of a
genitive NP, it follows the whole NP: tani nuku enami 'nephew' (child uncle
man, i.e., a male child of an uncle).
futfi 'much, large quantity' (also an adjective: 'big') is used with uncountable
nouns; e.g., futfi 'much', as in futfi dineiru 'much money', futfi afi 'much fire',
futfi weni 'much water':
fupe ' many, much' is used with countables, as in fupe namaJi 'many people':
ipeM 'how much, how many' can have a rhetorical use meaning 'every' with
countables and uncountables.
PayaJu 'all' also belongs to quantifiers. It can be used with countable nouns,
as in (387), and uncountables, as in (388). In the latter case and in (389) it is
used in the sense of 'whole'.
PayaJu differs from other quantifiers in that it can take cross-referencing O/S0
clitics in a non-predicative function, as shown in (390). Other quantifiers take
cross-referencing clitics only when used as predicates and treated as stative verbs
(see sect. 18.4.1). PayaJu is used with cross-referencing clitics in the following
cases, which agree with general rules of the use of cross-referencing clitics in O/So
function outlined in sect. 18.4.2. A cross-referencing clitic is never used in repeti-
tion (392). Clitics can refer to A (390), Ï (391) or S (392). In (390), the head noun
has an animate referent and is omitted (-ni '3ñà refers to A).
payailu ni-yulua-pajia-ha
all 3pl-fall-FUT-PAUS
'They will fall all (of them), all the trees, all will fall.'
payalu
all
'Then the white people take piagaba, cipo-vine - everything.'
PayaJu is also used as the head of a relative clause (see sect. 14.3):
A quantifier may follow the head noun if the latter undergoes right disloca-
tion as a part of an antitopic, for clarification purposes.
tsuludawa-ne Jupe
soldier-PL many
'Then this white man had many men, many soldiers.'
Adjectival roots such as isui, futfi without an adjectivizer can be used adver-
bially (see sect. 15.3.1 (examples 376, 377) and sect. 20), and sometimes it is
difficult to distinguish between a quantifier and an adverb:
It can be used with countable and uncountable nouns; this use may be a
caique from Portuguese (403).
Table 3. Demonstratives
Anaphoric Spatial
just previously near distant
mentioned mentioned
sg pi sg pi
Special feminine forms of demonstratives and the plural forms naita, anita are
used only in elicitation or when, in conversation, the speaker has been reminded of
them. Masculine (non-feminine) is used for feminine as the unmarked form in texts
and spontaneous speech: eni neyawa 'this woman', eta neyawa 'that woman'; but
in elicitation feminine forms are used: ayupaJu neyawa 'this woman', ayuta neyawa
'that woman'. This agrees with a general tendency of losing gender agreement in
head-modifier constructions in Warekena.
E, e-he is an anaphoric pronoun which is used to indicate a referent just
mentioned in the text and to emphasize its identity. The referent can be animate
or inanimate. This pronoun can be used both as a modifier and headlessly.
(407) ya-mia-e-pia-ha
NEG-PERF-he-NEG-PAUS
'It is not this one (not the same pen).'
When demonstrative pronouns eni, eta, and eya are cliticized (see sect.
22.4.1), they lose the first syllable.
Eya is an anaphoric pronoun which means 'the one mentioned in the previous
text'. It is generally used after a long stretch of a text has passed without
mentioning this particular referent, as a way of 'reintroducing' the referent. In
(408), the white man has been introduced earlier, and eya is used to draw
attention to him again.
(410) illustrates the opposition between eya and e. Eya is used in the first
sentence because the evil spirit has not been mentioned in the immediately
preceding text.
yajene-pia e-he
NEG +angry-NEG DEM-PAUS
'He is like this, this Awakaruna (an evil spirit). He is not angry.'
Warekena 315
Spatial demonstratives are eni 'proximate: the one just here (frequently ac-
companied by pointing)' and eta 'distal, that one over there', often accompanied
by the adverb yame 'far away'. Both are independent clitics (see sect. 22.4.1).
Their use as clitics (enclitics or proclitics) or as independent phonological
words depends on the emphasis they receive in discourse. They are also used as
anaphoric pronouns in texts (411, 412). Eni is used similarly to a definite article
and is also used to mark the topic of the discourse, in the meaning 'this very
one just mentioned, the one we are talking about'. The article-like uses of eni
are illustrated with (416) and (417).
eni-hi Diutsu
DEM.PR-PAUS God
'Then the bad man (just mentioned) ran from them, he ran from the
people far away, he did not want to see this God.'
malayu
deer
'Then the deer died, the deer died.'
When the turtle becomes the main participant, it is marked with the proxi-
mate demonstrative:
neni-hi apostulu-nawi
DEM.PR.PL-PAUS apostles-PL
'Jesus Christ spoke to them all, to them who were learning from him, all
of them, the apostles.'
Two demonstratives, the spatial eni and the anaphoric e, can co-occur in the
same phonological word:
Eta can also be used anaphoric ally, and then it means 'that one, already
mentioned (distantly)':
uleta-pitj"i
swallow-OBJ.FOC
¢ wave was big, the snake was coming, he arrived near that woman to
swallow her.'
Jutfi-li yue
big-ADJ to
'Then that man has a big knife.'
318 Aikhenvald
Eta can also be used to mark the head of a relative clause ("the one that").
In (424), eta is encliticized to the preposition minafi On' (see sect. 22.4.1).
Demonstratives tend to precede the head noun in the NP. In NPs which
contain a demonstrative and another modifier the following rules apply. The
demonstrative precedes a numeral (427, 431).
neJima-pe kwatru-ni
cousin-PL four-3PL
'Then the cousins went by the road, the four cousins.'
They are not cliticized if they require a special emphasis, as in eni kansau
'this song' in (250) (which follows (429) in the narrative), or in very slow,
paused speech, as in (431).
15.4 Nominalizations
Suffix -Ji is used to form instrument and object nominalizations: miwata 'play',
miwata-Ji 'a plaything'; neta 'pray', neta-Ji 'prayer' (cf. neta-fi 'praying'); mutUeta
'to steal', mutUeta-Ji 'stolen object'; wiyua 'die', wiyua-M 'poison'; fina 'tell',
fina-Ji 'story'. Nominalizations in -fi and -Ji may have a close meaning, e.g., awata
'believe', awata-fi, awata-Ji 'religion, belief. An unproductive suffix used for
object nominalizations is -uJe , found in tsawituJe-fi 'bow', nu-isawituJe 'my bow',
cf. tsawita 'shoot arrow'. Locative nominalizations are formed with the suffix -tfi:
waJuta 'plant', waJuta-tfi 'plantation' (also see note 5).
Agent nominalizations are formed with the suffix -ina (see sect. 22.5 on
phonological processes at morpheme boundaries): miwata 'play', miwatena
'player'; waJa 'cook', waJena 'cook'; tfina 'tell', tfinena 'story teller'. A few
agent nominalizations are formed with a 0 suffix: puteta 'fish', puteta 'fisher-
man'.
Unlike other Northern Maipuran languages, nominalizations in Warekena are
alienably possessed. They are used as heads of NPs:
16 Pronoun system
16.1 Independent personal pronouns. Personal pronouns of the 1st and 2nd
person and 3rd person plural pronouns in Warekena are formed with the help of
an emphatic suffix -ya attached to the cross-referencing prefix: nu-ya T, pi-ya
'you sg', wa-ya 'we', ni-ya 'you pi, they' (see sect. 18.4.1 on verbal cross-ref-
erencing). The use of the same form for the 2nd and 3rd person plural is an
innovation of Warekena. Third person singular pronouns are cognate to demon-
strative pronouns and contain the suffix -paJu: e-paJu 'he', ayu-paJu 'she'.
Since the pronominal constituents are expressed by cross-referencing on the
verb, independent personal pronouns are rarely used, and only under the follow-
ing conditions:
(a) When the constituent expressed by a personal pronoun is in a contrastive
focus, as in (437) (in italics), or defied, as in (438) (see sect. 9.4 for cleft
constructions).
wani-M pani-Jiwe
here-PAUS house-NPOSS+LOC
'He (the toad) called them (his cousins) to go, "You hunt, to kill game
for us, I shall stay at home".'
(b) when the participant expressed with the personal pronoun is one of several
that are enumerated:
Numeral One' in this use can take plural marker -pe, and then it means
'some':
Indefinite pronoun peya-pe is usually proposed to the head noun, but can also
be postposed, in repetition:
Pronoun imaJu 'the other one of a set of similar objects' is illustrated below:
the preservative wa 'then': wani 'here', wata 'there'. Allative adverbs contain
the directional suffix -ba (cf. nominal allative -waba): wa-ba 'to here' (also:
winf), wata-ba 'to there'. We hypothesize that the nominal directional marking
-waba goes back to double case marking (-wa 'perlative' 4- -ba 'directional').
Elative adverbs contain the elative suffix -\vene: wa-wene 'from here, from
there'.
The adposition is the head of the NP. Person, number and gender are
cross-referenced on the adposition with prefixes (see Table 1 in sect. 15.2.1).
Most adpositions can be used both as prepositions and postpositions.
Prepositions in Warekena are statistically more frequent than postpositions in
elicitation, conversation and texts. The placement of adpositions depends on
several factors, which will be listed below. Adpositions are used to express
peripheral roles, such as: (a) instrumental, comitative: epi 'with (comitative)',
ima 'with' (comitative, instrumental); and (b) spatial: minafi On, on top of,
piatu 'in front', epuna 'after (space)'. The only exception is yue 'for, to' which
is used to mark a core constituent. It is used to mark possessive relationship in
possessive sentences (see sect. 7.2), and to mark an Sj0 constituent, functioning
as a dative subject (see sect. 18.4.1):
weni iwe
water LOG
'Now the glass is dried, there is no water in it.'
tawape minaj!
jungle on
'They went until they arrived in the jungle.'
Epi and ima 'comitative' are mainly used as prepositions. Epi means 'to-
gether with, with the full participation of.
Warekena 329
Ima means 'with, with the help of, with a partial participation of:
In (468), ima means 'with', but without the implication of 'together with, in
the same house':
Ima tends to have comitative meaning when used as a preposition (470, 471),
and instrumental meaning when used as a postposition (472, 473).
The apparent closeness in meaning and almost interchangeability of ima and epi
in Warekena is hard to account for. It may be partly due to language death
phenomena which provoke a higher degree of variability, especially since the main
contact languages, Portuguese and Nheengatu, mark comitative and instrumental in
the same way. Warekena epi has cognates such as Tariana api 'with: comitative',
Bare abi 'with: comitative, (rarely) instrumental'. Warekena ima is cognate to Bare
ima 'instrumental, (rarely) comitative'. Bare displays a similar tendency to Ware-
kena in that abi and ima show a large degree of interchangeability in texts. This
isogloss is interesting for two reasons: Bare and Warekena share a number of
structural features and cultural isoglosses, possibly due to prolonged contacts. Both
languages have been in the same kind of contact situation, i.e., under strong
pressure of Nheengatu and Portuguese, for a longish period of time. This isogloss
may either show a convergent phenomenon in two genetically close languages
under the same kind of contact pressure of other languages and in the process of
language death, or be the result of a shared structural feature.
Another postposition used with instrumental meaning is iyu:
Warekena 331
Ifiwa 'from, away from' has the primary meaning 'away from' implying move-
ment away, abandoning (478). It can have a malefactive nuance (479, 480).
The main difference between elative case -wene and preposition ifiwa in their
spatial uses is that the basic meaning of the former is 'from inside of, and of
the latter 'away from'.
Petfi On' is used only as a postposition:
The use of minafi On, over' with the verb tfina 'to tell' is also influenced by
Portuguese contar sobre 'tell about (lit.: tell on)' (429).
Most adpositions have a nominal origin. They have special spatial forms,
similar to locative cases formed on nouns. Unlike nouns, adpositions do not
have a locative case formed in -iwe (however, some adpositions contain -iwe as
a fossilized marker). Directional case is expressed by the suffix -waba, e.g.,
nu-piatuaba 'towards in front of me (direction)', nu-piatu 'in front of me (loca-
tion)'. Piatu 'front' can be used as a noun. Elative case is formed by the
addition of -wene to the adposition:
Body parts are often used as spatial adpositions. They are preposed to the
argument, in agreement with the constituent order in possessive NPs:
(490) is the second sentence of the same story. Epi 'with' is preposed to its
argument, fianipe 'children', and it has the cross-referencing prefix:
A number of optional enclitics are used for tense reference when it has to be
stressed. Criteria for clitics are given in sect. 22.4.1. The tense enclitics are used
when the tense reference has a particular importance, for example, in sequencing
of events, as in (496), or it has to be underlined that the action refers to a particular
period, as in (495). These clitics are: -yaJe 'remote past', -pafia 'future'.
-YaJe 'remote past'. This clitic does not co-occur with other aspect markers (see
sect. 18.2), and can only be used once in a clause (495, 496).
yulua-pia-ha
fall-NEG-PAUS
' "Now the world will start ending, all the trees will fall, all the trees,
all will fall, only one tree, only one tree, the one near which (lit.: at the
foot of which) I am standing, will not fall," (said the rabbit to the
jaguar).'
-pafia follows -wa 'non-accomplished', if the two co-occur in the same word:
18.2 Aspect. Warekena has eight aspectual enclitics and suffixes, all optional.
They are: -mia 'perfective', -wa 'nonaccomplished, continuative', -nia 'change
of state', -peta 'intensive or immediate action', -yaJu 'yet, more', -dekana
'going back and forth', -tfiJi 'habitual', -tfi 'repetitive'. Aspectual meanings can
also be expressed by serial verb constructions (see sect. 18.8.1).
ja:bine-pe yuwaba-ha
family-PL towards-PAUS
'He (the turtle) called his family, then his family came to him (as the
result of his calling).'
-mia is often used with this meaning in embedded clauses which contain the
conjunction ate 'until':
malayu
deer
'Then they were late, they slept, they slept until they woke up, they
woke up, (then) the deer spoke.'
ya-mia-yutfi-piehe malayu
NEG-PERF-strong-NEG + PAUS deer
'The deer ran, he died, he fainted, the deer, the deer had no more
strength'.
-mia means something like 'already' in (506), or 'any more' if used with
negation (507).
338 Aikhenvald
(c) -mia is used to mark first plural imperative which refers to an action that has
already started, or has been planned in advance. (508) comes from the story
about a running competition between a cunning turtle and a naive deer. The deer
is urging the turtle to start the competition which has already been planned.
(d) with the verb of perception eda 'perceive, see/hear,' -mia means 'a sudden
result, unexpectedly':
-Mia is not used with the verb of perception eda 'perceive/see/hear, under-
stand' if there is no implication of 'unexpectedness' (512). It is also not used
with verbs of speaking, such as "say", "speak", "answer" (see 504, 506, 510).
Examples such as (507) show the clitic character of -mia; it always undergoes
attraction to the negative proclitic ya-.
18.2.2 -wa 'non-accomplished action', -wa is a suffix, since it does not un-
dergo attraction to the negative proclitic ya (521). It follows the negative -pia
in (22), (23). It is used with (ya-) in the following meanings:
-wa in this inchoative meaning is very often used with motion verbs. It can
also co-occur with -pafia 'future'.
yu-Jia-wa
3sgf-stay-NONACC
Then they (abandoned children) saw a woman, she was weeping.'
The inchoative and continuative meanings are not always easy to distinguish,
especially when -wa is used with motion verbs. In repetition, the inchoative
meaning is more linked to the beginning of a discourse, and the continuative to
the following sentences of discourse, as can be illustrated with (519):
(b) -wa can also occur on the predicate of a dependent complement clause:
(c) -wa is used to express immediate action, and even immediate future. (521)
shows the contrast between -wa used to express immediate action in a serial
verb construction, and the predicate without -wa, which indicates a permanent
state of affairs.
ni-Jia-palu-wa
3pl-live-PURP-NONACC
'Then she (the witch) gave them (the abandoned children) a house to
live in.'
-wa can mean unexpected, unplanned action when used with active (Sa) or
transitive verbs (but not with the verb of perception).
-wa can occur with either component of a serial verb construction, but never
with both; see sect. 18.8.1 for criteria for serial verb constructions and how they
differ from a series of predicates. In (526), -wa occurs with/j'a 'to sit', and gives
that verb an inchoative meaning (also (528)). In (527), -wa occurs with the verb
of movement fa 'go', and it gives it an inchoative meaning. -Wa specifies the
action of either component of a serial verb construction independently, and so
it belongs to nuclear, and not to core, serialization (see Foley and Olson 1985).
-mia 'perfective, resultative' and -wa can co-occur, in either order: -mia-wa
and -wa-mia, with the following semantic difference: -mia-wa means 'a continu-
ative action the result of which is important; gradually attaining the result', as
is illustrated by (529-530); -wa-mia means 'the result of a started action' (531).
ni-kawyu-ta-palu ni-du:le
3pl-smoke-CAUS-PURP 3pl-stock
'So they settled down (sat and stayed there) to smoke the stock.'
The typical ending of most stories shows an alternation of -mia-wa and -wa:
ni-ja-mia-wa ni-yuleta-mia-wa
3pl-go-PERF-NONACC 3pl-return-PERF-NONACC
ni-yuluta-mia-wa kulimalu-nawi
3pl-lie-PERF-NONACC turtle-PL
'So they (turtles) started and finished scattering themselves, they re-
turned to the foot of a hill, then the turtles finished lying down.'
(534) illustrates the semantic contrast between -mia-wa, -mia and -wa. -Mia-
wa means 'a continuative action which has happened already', -mia means
'resultative, accomplished action or state', and -wa is used in the sense of
unaccomplished, continual action. The aspectually unmarked form ya-nupa-pia-
hä 'NEG-come-NEG-PAUS' is used in repetition.
wa yele-mia waji
then tired-PERF jaguar
'Then the rabbit was already going far away, then the jaguar was waiting
(for him), waiting on the lake shore until the jaguar has become tired,
the rabbit did not come/was not coming, then the jaguar has become
tired.'
tenepu
road
'So he spread them until they were finishing, the turtles were finishing,
until where the road was coming to an end.'
ni-yeJeta-wa ni-pipi-nia-mia-ha
3pl-arrive-NONACC 3pl-be.lost-INCH-PERF-PAUS
'So they (the abandoned children) did not know where to go, they have
become lost.'
18.2.4 -peta 'intensive or immediate action'. These two meanings of -peta are
illustrated with (432), (540)-(542).
Warekena 345
nu-Juma-ha
1 sg-suffocate-PAUS
'My mother smokes, she smokes very much, (so that) I get suffocated.'
18.2.5 -yaJu 'yet, more'. This meaning of -yaJu is illustrated with (543-546).
kalibe
manioc, liquid
'The woman is working, she is drinking manioc liquid going back and
forth.'
18.2.7 -tfiJi 'habitual'. Both -tfiJi and -dekana have not been attested with
negation, so it is difficult to decide whether they are clitics or affixes. -$Ui
precedes both -mia and -wa in a verbal form. This, however, is not a sufficient
piece of evidence in favour of or against the affixal status of this morpheme,
since -Ji 'relativizer', apparently, an enclitic, also precedes -wa (see sect.
18.2.2), and follows -mia.
Repetitive -tfi precedes -wa 'nonaccomplished action' (551) and can precede
or follow -mia (552). In the case of a sequence of aspect markers -mia-wa 'a
continuative action the result of which is important, gradually attaining the
result' (see above), -tfi goes after -mia and before -wa, as illustrated in (552)
and (553).
aliwa-tfi-mia wa wa-ja-mia-tji-wa
dawn-REP-PERF then lpl-go-PERF-REP-NONACC
'We slept just like that (i.e., without any water to drink), until it had
dawned again, then we went on again.'
-tfi means 'further on' in (554)-(556). With this meaning it follows both -mia
and -wa in (554).
ne-mia maJayu
3pl +eat-PERF deer
They (the turtles) ate more and more (of the dead deer), they ate the
deer.'
-tfi does not co-occur with negation. Its clitic character is confirmed by the
relative freedom of placement with respect to other aspectual markers. In serial
constructions -tfi can occur on either component, specifying the action of each
component independently; this can be seen by comparing (553) and (556). Thus,
-tfi belongs to nuclear serialization.
yajapua
night
'They (the jaguar and the rabbit) went on talking a lot, until late at
night.'
(566) cannot be interpreted as 'very many people can hear the voice of birds'.
ya-be-pia-hä mutfita
NEG-can-NEG-PAUS bite
'Then the turtle gnawed tapir's bones, the turtle cannot bite.'
Reduplicated intransitive verbs can take the causative suffix -ta in a few
cases; e.g., paJa 'run', paJa-Ja 'run far away', paJa-Ja-ta 'make run far away'.
Reduplication of the causative morpheme is more productive. (573) comes from
a story about evil spirits. The evil spirit Awakaruna is described, the reduplica-
tion of the causative kune-ta-ta (fear-CAUS-RED) 'he frightens very much'
being used to emphasize how frightening he is:
kune-ta-ta-ma-ba-lu
fear-C AUS -RED-DEL-AUG -EMPH
'He (Awakaruna) shouts to frighten people. He does not kill people, only
frightens them, he frightens them very much.'
ni-mai-na-wa ni-yutfia-na-wa
3pl-fight-REC-NONACC 3pl-kill-REC-NONACC
They (non-evangelicals) drink whisky, they faint of drunkenness, they
quarrel and kill each other.'
S 0 i=A2; (584) illustrates 81 = 82 and (585) illustrates Si=A2. There are no pivot
restrictions, i.e., any possibility of coreferential deletion is grammatical.
Ai=A 2 :
Sol =A 2
Si=S 2
(584) wa-hä Jesu Kritu yapa-wa
then-PAUS Jesus Christ enter-NONACC
pane-wa na wale-Ji
middle-PERL village-NPOSS
Then Jesus Christ entered into a house, into a house, to go in the middle
of the village/
Si=A 2
waya-ne Diutsu
say-POSS God
'Then he went to tell, to tell this word of God.'
It may be used with the meaning of 'so that', when the result of the action is
emphasized, as illustrated below:
-PaJu is a clitic, since it is attracted to the negative proclitic ya- (588) (see
sect. 22.4.1).
ya-palu nu-wiyua-pihe
NEG-PURP Isg-die-NEC + PAUS
'Tie me here on the tree, so that I do not die.'
-PaJu does not combine with -mia 'perfective'. It combines with -\va, which
may follow or precede it. It precedes -paJu when it has an inchoative meaning
(515, 589). Otherwise it follows -paJu (522). -PaJu may mark the predicate of
a complement clause (590).
Warekena 355
-PaJu can be used similarly to -buJe Oblique focus' (see sect. 18.5.3.2 on the
differences):
pi-Jiani-pe pe-mia-ni
2sg-child-PL 2sg + eat-PERF-3pl
'The rabbit answered (the jaguar), "You have indeed eaten your chil-
dren." '
In serial verb constructions -paJu can be marked only once. It occurs with
the first component and refers to the whole construction. Thus, -paJu belongs to
core verb serialization.
mala-mia-ha wabupi
shallow-PERF-PAUS spring
'The jaguar ran by the shore of the spring, to the headwaters of the
spring, for him to go and wait for the tapir, (since) the spring is shal-
low.'
We means 'to leave' when -paJu is used in a complement clause, and 'to let'
in syntactic causatives (where -paJu is never used; see sect. 6):
Matse can be used to warn of bad consequences of an action which are either
self-evident, or outlined afterwards.
matse pe-na
WARN 2sg + eat-lsg
'Then the monkey answered (the jaguar), "I shall untie you (from the
tree), and you will eat me, beware lest you eat me".'
(i) Obligatorily transitive verbs always have two arguments; if one of the two
is omitted it can be retrieved from the context, following the pivot restrictions
as shown in sect. 3. Examples of obligatorily transitive verbs are: eda 'see/hear,
understand, look after', mutfi-ta 'bite', kuJua 'drink', teJuka 'cut', e 'eat', pie
'find', pata 'catch', tfina 'tell', peta 'hit', weya 'want', daJuna 'tie', aJita 'tie
thouroughly', wakwa 'untie', deka 'have'; yanata 'write', ma 'do, say', we 'let,
leave', wepa 'leave, throw', yanita 'take', yutfia 'kill, hunt', yata 'ask', wanufa
Order', tse 'understand, know', wapeta 'attend to', muta 'call', ayuda 'help',
paJa-ka 'weed',yw/a 'bite (mosquito)',yubuta 'peck'', fimeta 'smell', beni 'like',
beda 'try' (food), eJita 'pick', tutu 'embrace, hug a child who is sitting in one's
lap', tfitfyuta 'hug', waJamatfia 'save, guard', djaJu 'hold a child who is standing
in one's lap', pite, a:\vale 'bury', djeweni 'lie, tell lies', tfiwa 'slice', dewe
Owe', wenita 'buy', wende 'sell'; awata 'believe', neta 'pray', mitfiuia 'hide',
yuwita 'blow (the fire), isiata 'join'.
Transitive verbs of class (i) are higher on the transitivity hierarchy (Hopper
and Thompson 1980) than transitive verbs of other classes.
Some of these verbs preferentially take sentential complements, e.g., beni
'like', tse 'know', we 'let, leave'. This class also includes a number of loan
verbs, e.g., ayuda, ayuza 'help' (Portuguese ajudar, Spanish ayudar), deve
Owe' (Portuguese dever, Spanish deber), wende 'sell' (Portuguese, Spanish
vender). These verbs are obligatorily transitive in Portuguese and Spanish.
The class of obligatorily transitive verbs can be considered an open class,
since it accepts new members - loan verbs.
(iii) S = O ambitransitives are a big class. They include: eta 'burn', paka 'break
in two pieces', bafa-ka 'cut, split, tear (e.g., a thread)', beJika 'tear' (into small
pieces, e.g., paper, cloth), yu(w)aka 'tear after having folded', alia 'break ' (a
pen, a stick), aita Open, split open, spread (washing)', ika 'close', enina 'be
born, give birth', tawina 'grow, bring up'; eputa 'wake up'; atseta 'learn, teach',
kudunia 'melt', bukuita 'light' (fire), makuta 'disappear, \ose\frita 'fry', waJa
'cook', kawyuta 'smoke (meat, fish)', tfikya 'crisp', yata 'take, stick', daJuna
'drag (a canoe)', 'crawl (a snake)', faita 'scare, stun, surprise', meta Obey,
permit', mebuta 'be surprised, surprise', puJyuta 'remember, think', yuwaka
'lean', wafata 'mix', mafini 'damage, spoil', inapa 'finish, end'.
Intransitive verbs also fall into three classes: Sa verbs, S0 verbs, and Sj0
verbs.
(iv) In Sa type verbs Sa person markers are prefixes. The majority of the Sa
verbs can be causativized by adding the causativizer -ta. The causatives of Sa
verbs are never S = Ï ambitransitives; they can be either obligatorily transitive
or A = S ambitransitives. All motion verbs belong to the Sa type, some verbs
denoting typical activities (paddle, work, play), and some verbs denoting states,
e.g., die, faint, and physical conditions involving actions, e.g., sneeze, cough.
Examples of Sa verbs are: fa 'go', patata 'work', u\va 'climb', yuJua 'fall',
Warekena 361
tsapia 'jump', bitfika 'leave, go out', fia 'sit, live, stay', yapa 'enter', miwata
'play', paJa 'run', yuJeta 'return', ema 'cry', mita 'fly', yenita 'laugh', nupa
'arrive', yaya 'weep (children)', wiyua 'die', daJina 'faint', awa 'bathe', fipata
'rot', atfita 'sneeze', fuma 'cough', fuweta 'spit', tehi 'blow (nose)', yaJa
'swell', pauta 'step', yawaJuta 'go round', pipi 'be lost', guweka 'bark', bitfika
'go out', apita 'hang', yeJeta 'arrive', tfiJuna 'crawl', tama 'dance', yupukwa
'get used', daJa 'snore', biyada 'escape', fuma 'drown', tapana 'yawn', tfikuta
'hiccough', weta 'turn into', naka 'get up', yama 'float, go downstream', kadika
'go upstream', yata 'put to shore' (canoe), yanana 'surround', mufa 'become
wet', kutsa 'become humid', ifita 'have a name', tsuJuita 'crouch', kefuka
'cross', tfima 'sleep', kerika 'smash', kuJuka 'have holes', inaka 'change,
move', muta 'stretch', muJu 'crush', teya 'vomit', unita 'swim', mebuta 'be tied
(of a dog)', mawa 'go out (fire)'.
A thematic formative -ka can be distinguished in the following Sa verbs:
bitfika 'go out', inaka 'change, move', kerika 'smash'; kuJuka 'have holes',
guweka 'bark'. The existence of this formative is corroborated by the redupli-
cation of the kind guwe-we-ka 'bark a lot'. However, this is subject to some
variation, e.g., bitfika-ka 'he went out a lot'. A few Sa type verbs are denominal
formations which contain a suffix -ta, e.g., ifita 'to be named, to have a name'
(cf. ifina 'name'), and possibly unita 'to swim' (cf. \veni 'water'). The verb ifita
is, however, problematic as far as its transitivity is concerned; it is closer to
copula verbs, since it necessarily takes a copula complement:
(v) Verbs of the S0 type in Warekena can denote state, e.g., uJa 'be dirty',
anuana ' be sick',/e/e 'be dry' or refer to natural phenomena, like aJiwa 'dawn,
to get up early\fibuma 'grow dark, spend the night'. They can form causatives
which are S = O ambitransitives. S0 type verbs do not take cross-referencing
prefixes. Cross-referencing enclitics are used to mark a pronominalized S0, or a
right-dislocated S0 expressed by an NP (see sect. 18.4.2). Examples of S0 type
verbs are: dabana 'first', yeJe 'tired', uJa 'dirty', anetua 'good, beautiful',
atulapi 'full (food), pelu 'full', anepi 'right', u:Ja 'soft', tepa 'hard'; colours:
feJe 'black', aJi 'white', ayune 'blue', amu 'rosy, red', te\va 'yellow', puJe
'green'; mufa 'damp',/e/e iAiy",fibu 'spend a night, grow dark', yawa 'be late',
akune 'be afraid', tfitfi 'rotten, ill-smelling', djewi 'deaf, ma-puJi-paJu 'blind',
abu:da 'bald', niwe 'high, lofty', futfi 'big, fat', puJapiami 'thin', anuana 'ill,
sick', takuwa 'feverish', awi 'hurt', yaJu 'sickly', kaJebu 'quick', fume
'survive, remain alive', katsaJi 'married', aJema, aJe 'be this way', maJa
362 Aikhenvald
(617) aliwa-wi
get.up.early-lpl.So
'Good morning (Lit.: we got up early).'
(vi) Sio verbs are a small class, and they denote physical state110". The Si10
constituent is complemented by the postposition yue 'to, for'. Si0 verbs cannot
be causativized. Any Sio verb can be used as an S0 type verb; then it can be
Warekena 363
causativized. The Sj0 verbs include: spate 'cold', mawaJi 'hungry', ata 'warm,
hot', akawi 'bubble'; names of illnesses: matutfi 'bad cold', maJeia 'malaria'.
A predicate used as Sj0 denotes a temporary state, or condition, as in (620).
The same predicate used as S0 denotes a permanent, or prolonged state, as in
(621).
(621) mawali-na
hungry-Isg
º am hungry (permanently, or over a continuous time-span).'
(622) refers to the abandoned children who became hungry after wandering
in the woods for a long time:
(623) refers to the jaguar who has been suffering from hunger for a long time
and could stand it no more:
There are a few verbs that can be used as both S0 and Sa, but with a semantic
difference. An example of this sort is S0: puJa-pia-mi 'be thin', puJa 'become
thin, lose weight'; cf. the causative of the latter:
Warekena uses cross-referencing prefixes for marking A and Sa, and cross-
referencing enclitics for marking and S0. cross-referencing prefixes are also
used for marking the pronominal arguments of adpositions and pronominal
possessor in possessive constructions, cross-referencing prefixes and enclitics
are given in Table 1 in sect. 15.2.1.
Cross-referencing prefixes are obligatory in predicate-argument construc-
tions, and they encode a pronominal argument, A (625) or Sa (626), and mark
the agreement with a full NP: A (627), Sa (628).
(626) yu-wiya-mia
3sgf-die-PERF
'She died.'
i-beni jia-wa
3sgnf-like live-NONACC
One (evil spirit), Awakaruna, he likes to live in the direction of a hill.'
(631) comes from the story about evil spirits. The referent of i- '3sgnf, an
evil spirit, is in focus.
Warekena 365
i-yue-he piji
3sgnf-for-PAUS tail
'When he nibbles (at cattle), big are his nails, his side is striped; it is a
big stripe, his tail is striped, he has a tail.'
Similarly to Baniwa, enclitics are not used to mark agreement with an O/So
NP constituent (634):
Cross-referencing enclitics can co-occur with a full NP, when it is used for
clarification (130), (636); see sect. 9.4.
366 Aikhenvald
18.5.1 Causative. Morphological causatives are formed with the help of -ta suf-
fix, added to the stem. Causatives can usually be formed only on intransitive verbs
of Sa and S0 types (with a few exceptions to be discussed below). The difference
between morphological and syntactic causatives is analyzed in sect. 6.
Causatives of Sa type can be transitive (637-639) or A = S ambitransitives
(640-643). They are never S = O ambitransitives. Examples of causatives of Sa
verbs are:
yuJeta 'return', yuJeta-ta 'return something';
mita 'fly', mita-ta 'make fly away';
yenita 'laugh', yenita-ta 'make laugh';
biffika 'go out', bitfika-ta 'make go out, drive out';
biyada 'escape', biyada-ta 'make escape';
paJa 'run', paJa-ta 'make run';
yapa 'enter', yapa-ta 'make enter';
fuma 'drown, get drowned', fuma-ta 'drown, strangle';
weta 'turn into something', weta-ta 'turn somebody into something';
naka 'get up', naka-ta 'lift';
yama 'float, go downstream', yama-ta 'put to float, disembark, take
downstream'.
Some causatives have an idiosyncratic meaning: e.g., wiyua 'die', wiyua-ta
'make die with the help of sorcery, poison, etc.' (unlike yutfia 'kill'); yuJua
'fall', yuJua-ta 'miscarry (a baby)'; teya 'vomit', teya-ta 'make sick'; uwa
'climb', uwa-ta 'lift, put on an elevated surface; raise (prices)'; nupa 'arrive',
nupa-ta 'bring'.
The following motion verbs of the Sa type have no causative: fa 'go', yeJeta
'go, arrive', pauta 'step', yawaJufa 'go round', tama 'dance'. The following Sa
verbs, which denote physical characteristics, have no causative: daJa 'snore',
atfita 'sneeze', fuma 'cough', fuweta 'spit', tehi 'blow (nose)', yaJia 'swell',
guweka 'bark', tapana 'yawn', tfikuta 'hiccough'.
The use of causatives formed on Sa verbs is illustrated below.
wa-yama-ta-wa pani-Ji-waba-hä
Ipl-go.downstream-CAUS-NONACC house-NPOSS-DIR-PAUS
'Then we arrived there, and we headed downstream, towards home.'
dabana 'be first', dabana-ta 'make first, begin'; fefe 'be dry', fefe-ta 'dry,
make dry'; akune 'be afraid', kune-ta 'scare'; fuJuJu 'be in a hurry', fuJuJui-ta
368 Aikhenvald
'make hurry'; amena 'be sharp', amena-ta 'sharpen'; pa:fi 'be ashamed', pafi-ta
'make ashamed, blame, punish'; uJa 'be dirty', uJa-ta 'make dirty'; kuyuJe 'be
sad', kuyuJe-ta 'make sad'; afuJu 'be glad', afuJu-ta 'make glad'; uJuJiwi
'appear', uJuJiwi-ta 'make appear'.
Some causatives from S0 verbs are S = O ambitransitives, as can be illus-
trated with (644) and (645). The -ta causative form of the verb dabana 'be first'
is the only way of saying 'begin'.
Examples of fefe-ta used intransitively are given in (459), (589). The same
verb is used transitively in (646), a continuation of (459).
Some -ta causative forms of S0 verbs are used only transitively, e.g., kune-ta
'scare' (573). S = O ambitransitives can sometimes have slightly idiosyncratic
meanings:
Usually, transitive verbs cannot form causatives; e.g., waJa 'boil, cook',
*wala-ta 'cook'. There are a few exceptions to this rule. The obligatorily
transitive verb kuJua 'drink' (649) has a morphological causative kuJua-ta 'to
make drink' (650). Note that the underlying A becomes O, and Ï becomes an
oblique, contradicting Comrie's generalizations about the treatment of argu-
ments in causatives (Comrie 1981:169).
Warekena 369
However, the syntactic causative nu-ta kuJua-ha º gave (him) to drink' seems
more natural (see sect.6). The causative kuJua-fa is very restricted; for instance, the
speakers did not accept sentences like *nu-kuJua-ta-pi weni (ima) º make you
drink water'. This causative formation may be considered a kind of relic in the
language. It should be noted that all the other Northern Maipuran languages do not
form causatives on transitive verbs, the only exception being the verb 'to drink';
e.g., Bare -dia 'drink', -dia-sa- 'to make drink/drunk', Baniwa of I$ana -133,
Tariana -iJa 'drink', Baniwa of I^ana -i^eta, Tariana -Ueta 'make drink, suckle (a
baby)'. Warekena has one more case of causative formed on a transitive verb: guJa
'peel', guJeta 'make peel' (attested once in spontaneous speech).
In one case, the suffix -ta was used with an A = S ambitransitive in an
idiosyncratic meaning: yubua 'dig', yubua-ta 'stumble (one's foot)'.
There is a small degree of variation between S = O ambitransitives and Sa
verbs with corresponding causatives for a few verbs. The verb enina 'give birth;
be born' was continuously used as an S = Ï ambitransitive by the speaker. When
we were discussing the formation of causatives, the speaker remembered the
causative form enina-ta 'to give birth'. The same happened with tfipuJia 'turn'
and eputa 'wake up', which were used as S = O ambitransitives; but causatives
tfipuJia-ta 'to turn something' and eputa-ta 'to wake up somebody' were cited
after some thought, and the following variants appeared:
(654) eputa-na
wake-lsg
'He woke me up.'
370 Aikhenvald
(655) neputa-wa
1 sg -I- wake. up-NONACC
º woke up.'
Another example of the same sort is the Sa verb yapita 'hang', which forms
a causative yapita-ta 'hang, make hang'. This Sa verb has also been attested as
S = O ambitransitive (656, 657).
nijiwa i-tapa-palu-ni
3pl + from 3sgnf-go-PURP-3pl
'Then he (father) went around, away from the children, away from them,
to avoid (go past) them.'
waliya-wabehe ni-kawyu-ta-paJu
smoking.grid-DIR + PAUS 3pl-smoke-CAUS-PURP
'Then he went by the road of his brothers, for all of them to take with
them the insides of the game to hang on the smoking grid, to smoke.'
reciprocal meaning:
ni-mai-na-wa 'They fought each other.'
3pl-fight-REC-NONACC
372 Aikhenvald
reflexive meaning:
nu-teluka-na-wa kutfiyu iyu
Isg-cut-REFL-NONACC knife with
º cut myself with a knife.'
Compare the use of alaka 'put on (clothes)' and aJlaka-na 'dress oneself:
In some cases the use of -na- 'reflexive' with intransitive Sa verbs can be
influenced by Portuguese (665).
(667) illustrates the reciprocal meaning of the -na- derivation with a plural
subject:
The reflexive meaning of the -na derivation is illustrated with (668) and
(669). (669) contains a rare example of the -na derivation on a causative of a
verb, which contains -nia 'inchoative, change of state'.
pipi-nia-ta-na-mia-wa ijiwa
lost-INCH-CAUS-REFL-PERF-NONACC from
'The rabbit did not know where to hide himself (lit.: make himself get
lost) from him (jaguar).'
A na- derivation can sometimes have a reflexive meaning even if the subject
is plural:
wali alita-na-wa
where tie-REFL-NONACC
'Then the monkey went and bit the vine where he (the jaguar) was tied.'
wakwe-na-mia-wa wakwa-na-mia-wa
untie-REFL-PERF-NONACC untie-REFL-PERF-NONACC
The monkey bit (the vine) until it tore, and he (the jaguar) was untied,
untied.'
verb and its topic". Focus in Warekena is marked on the verb as a special
derivation. Its function is to show that an object (O) constituent or an oblique
constituent belongs to (contrastive) focus. Warekena has two kinds of focus:
object focus marked with -pitfi, and oblique focus marked with -buJe.
Both focus derivations are fully productive and can be used headlessly. -Pitfi
derivations can be used with a meaning close to object nominalizations (see
sect. 15.4.1): e-pitfi 'food' (thing to eat), nu-wenifa-pitfi 'my purchases, things I
buy/bought', nu-wende-pitfi 'what I sell/sold'. -BuJe derivations can be used
with a meaning close to locative or instrumental nominalizations, e.g., nu-
kuJua-buJe 'glass, mug' (the thing for me to drink with); e-buJe 'plate' (the
thing to eat from/on); miwata-buJe 'companion to play with' (the one to play
with), nu-tawita-na-buJe (Isg-cover-REFL-OBL.FOC) 'blanket, the thing for
me to cover myself with'; nu-yubuta-buJe 'grinder'; futfi-buJe ka:ka (big-
OBL.FOC bread) 'yeast; the thing with the help of which bread becomes big'.
18.5.3.1. Object focus. Object focus is used when the object (O) is in the
(contrastive) focus. The suffix -pitfi is a suffix, because it does not undergo the
attraction to the negative proclitic ya-. The focused constituent is placed before
the predicate marked with -pitfi. The predicate retains its transitivity, and the A
constituent is still cross-referenced on it. Object focus -pitfi shares with passive
its main discourse function, namely, that it involves focusing of O. The object
focus derivation does not qualify as a passive.
Passive as a valency reducing derivation involves the following processes (as
suggested in Dixon 1994):
(a) the underlying O becomes S of a derived intransitive verb;
(b) A becomes a peripheral constituent and can be omitted;
(c) passive is explicitly marked on the verb.
Unlike passive, if the object focus is applied, the verb does not become
intransitive, since it still has an O, and the A constituent is still cross-referenced
on the verb, as illustrated in (677)-(678).
ne-pitj"i ne-pitj"i
3pl + eat-OBJ.FOC 3pl + eat-OBJ.FOC
'They have smoked meat (and not anything else), to be eaten by them.'
The object of the verb uJeta 'swallow' is a topic in (680), and so -pitfi
derivation is used:
uleta-pitj"i
swallow-OBJ.FOC
'The tide was big, the snake has come already, he arrived near her, this
very woman, that one to be swallowed by him. '
In (681), again, the parts of the body of the tapir to be eaten by the jaguar
are in focus, and so the -pitfi derivation is used on the verb:
The contrast between -paJu 'purposive' and -pitfi Object focus' can be illus-
trated by the following example taken from the story about a tapir and a jaguar.
In (682) -paJu is used for a general description. Then different parts of the
tapir's body become topics, as they become more individuated in the text, as
they are being consumed by the jaguar, and -pitfi is used in (683):
-paJu, and not -pitft is used when the action, and not the object, is in focus,
as in (684) and (685):
pe-palu-na
2sg + eat-PURP-lsg
' "Throw me up high, so that I get fat, so that you can eat me," (said the
monkey to the jaguar).'
18.5.3.2 Oblique focus. Oblique focus is marked with the help of a suffix
-buJe (which is a suffix since it does not undergo attraction to the negative
proclitic ya-). The function of oblique focus is to mark an oblique (peripheral,
i.e., locative or instrumental) constituent as a topic. It cannot be considered a
passive, for the same reasons as the object focus. The following examples
illustrate -buJe marking a locative constituent in focus:
The difference between -paJu 'purposive' and -buJe can be illustrated with
the help of the following 'minimal pair'. As Humberto explained, (696) means
that I am going to eat nothing but pure manioc flour; (697) means that I am
going to eat something else with manioc flour:
Similarly to -pitfi Object focus', -buJe does not qualify as passive because
the verb does not become intransitive, and the A constituent is cross-referenced
on it. Similarly to the -pitfi derivation, the -buJe derivation can be only used in
the case of S/A pivot.
18.6.1 Relative forms. Clitic -Ji marks the relative verb forms. The use of -Ji
as an adjectivizer (see sects. 15.3.1; 19) can be considered an extension of the
same morpheme used as a relative form marker. The difference between -Ji
'adjectivizer' and -Ji 'relative form marker' is that the former behaves more
like a suffix, and the latter is a clitic.
-Ji is used in the following three main functions:
(i) In cleft construction characterized by (1) left dislocation of the subject; (2)
-Ji marking on the verb; and (3) suppression of A/Sa cross-referencing prefixes
on the verb. The cleft construction is discussed in sect. 9.4; see example (438).
(ii) In interrogative clauses (discussed in sect. 10) (152—154).
(iii) In coreferential relative clauses (see sect. 14.3.1):
(iv) In converbal subordinate clauses, where the subject of the -Ji form may
coincide with that of the main clause (see sect. 14.3.2):
In the cases (ii) and (iv) -Ji does not trigger suppression of cross-referencing
markers.
If -Ji is marked on only one of the contiguous predicates, they may or may
not constitute a serial construction (see sect. 18.8.1 on serial constructions).
(702) is not a serial construction, since the two contiguous predicates, weya-Ji-
hl (want-REL-PAUS) 'who wanted' and atseta-ni (teach-3pl) 'he taught them'
have different subjects. A long pause after \veya-Ji-hi also indicates that this is
not a serial construction. (703) (a converbal clause) is a serial verb construction.
-Ji goes on the auxiliary-like component of the construction; in (703) it is the
modal verb weya 'want'.
uleta-yu
swallow-3sgf
'Now I want to marry your daughter, I killed a snake from her, when
this one wanted to swallow her.'
-Ji is also used to mark nominalizations (see sect. 15.4), which can be used
as predicates in verbless clauses (709), (745). (745) is a case of a headless use
of a relative form.
Warekena 383
wepi wa-pana-Ji
pupunha Ipl-plant-REL
'Carä, sweet potato, sugar cane, abiu, pupunha, this is what we plant
(lit.: our planting).'
inapa-mia-wa mundu
fmish-PERF-NONACC world
'Then the jaguar spoke, "It is now that I will die, the world is coming
to an end." '
18.6.3 Degree forms. Warekena verbs have two degree forms: -ba augmenta-
tive and -ma 'delimitative, just'.
Warekena 385
The augmentative -ba- (also used with adverbs, see sect. 20) is used with all
types of predicates (except Si0). It is frequently used with the emphatic clitic
-Ju and perfective clitic -mia. It means 'very much, excessively' (719-722):
(721) nu-kulua-ba-mia-lu
1 sg-drink-AUG -PERF-EMPH
º drank too much.'
(722) nu-tapa-pa-ba-mia-Ju
Isg-walk-RED-AUG-PERF-EMPH
º walked excessively.'
-ba- can be used with negation, and then it means 'not much, not enough'.
Then it is followed by the emphatic clitic -Ju:
(723) ya-atulapi-ba-lu-pia-na
NEG-full-AUG-EMPH-NEG-1 sg
º am not too full (I can eat more).'
Another meaning of -ba is 'truly, really' (724), (725). In negative sentences, the
negation then goes between -ba and -Ju, and the meaning is 'not really' (726).
(725) ube-ma-ba-mia-lu
all-DEL-AUG-PERF-EMPH
'It is really over.'
386 Aikhenvald
The suffix -ma 'delimitative, just, the right quantity /quality' is not very
productive. It is used with several S0 verbs; e.g., (727), (728), (729), and once
with a transitive verb (730). Its etymological counterpart, Tariana -ma, has the
meaning 'excessive degree'. In (730) and (728) -ma is used together with the
augmentative -ba and emphatic -Ju; the last position in (730) is occupied by the
3pl O = S0 pronominal clitic.
(727) jibu-ma-mia
night-DEL-PERF
'It was already just dark (enough for people to go to sleep).'
Compare also aJe 'be this way' and aJe-ma 'just this way'; (cf. aJe-ma-Ju-ta
'just like this').
(730) kune-ta-ta-ma-ba-lu-ni
fear-CAUS-RED-DEL-AUG-EMPH-3pl
'He (the evil spirit) frightens them (people) a lot (but does not kill
them).'
possible, if the action of the second verb has to be emphasized (704). The negation
marking occurs with the first component of the construction. Both components
receive the same subject marking. The components of a serial verb construction do
not have to share arguments other than subject. A serial construction contains more
than one phonological word, and each of them can appear with the pausal form
marker. There are no restrictions on using pausal forms for each of the components
of a serial verb construction (731, 733). The second component of a construction
can usually be repeated, but not the first (750).
Serial verb constructions have the following properties which show that they
are not a sequence of predicates, but a single predicate. These properties help
to differentiate between serial verb constructions and sequences of predicates
(including repetition), syntactic causatives and complement clauses,
(i) Serial verb constructions are used to describe a single overall event,
(ii) The components of a serial verb construction are contiguous and no other
constituent can intervene between them.
(iii) The components of a serial verb construction share the same subject which
is obligatorily cross-referenced on all of them.
(iv) The components of a serial verb construction share some of the aspectual
marking and they cannot have distinct aspectual marking.
Aspectual markers in Warekena provide an interesting argument in favour of
distinguishing between nuclear and core serialization (see Foley and Olson
1985). Clitics -mia 'perfective', -yaJu 'yet', -pafia 'future' belong to core layer
serialization, since they cannot be used independently to characterize the prop-
erties of different components of serial constructions. Clitics -yaJu 'yet' (543)
and -mia 'perfective' (731) usually go on the first component of a serial verb
construction. -Pafia 'future' goes on the second component of a serial verb
construction (501).
(733) comes from the story about Humberto's family, and this part of the
narrative is focused upon how the old woman, his wife, is unable to do anything
(work, grate manioc, etc.) because of her old age:
etyawa-mia-yu
old + FEM-PERF-3sgf
'She cannot work, she is old.'
(553), (556) show that -tß 'repetitive' can occur with either component of a
serial verb construction, depending on whether it refers to the repetitive char-
acter of the action expressed by the first verb (553) or the second verb (556) .
(v) The components of a serial verb construction have a single negation, whose
scope covers the whole serial construction (734, 735), and they can have only
one marker for purposive mood (595, 596, 736).
-PaJu 'purposive' usually goes on the first component of a serial verb con-
struction (in italics in (736)):
(vi) There is no clause boundary (marked with a presentative wa) between the
components of a serial verb construction.
Warekena 389
The following pair of examples illustrates the difference, both formal and
semantic, between a serial verb construction (737) and coordinated clauses
marked with the help of the presentative wa 'then' (738).
(vii) The serial verb construction has the intonation pattern of a monoverbal
group. A specific 'enumerative' rising intonation is prohibited in serial verb
constructions. In (528), rising intonation on nu-fa º went' shows that this verb
cannot be a part of a serial verb construction. This intonation is not found on
nu-weya-ha º wanted' which forms a part of a serial verb construction (528).
In (739) -fa and -yutfia form a serial construction, but kawyu-ta does not,
since it is separated from the serial construction by a constituent.
The following sentence does not satisfy the conditions (v) (scope of negation)
and (vii) (phonology), and so it is not a serial verb construction; it contains the
repetition of the same predicate and cannot be interpreted as a serial verb
construction.
we-miehe ya-mia-ipeta-piehe
leave-PERF + PAUS NEG-PERF-beat-NEC + PAUS
'Then the rabbit left him, left him, he did not beat him any more.'
(743) is not a serial verb construction, because the two contiguous verbs
receive an independent aspect and mood marking.
(595), (735) and (754) show that the object clitics go on the last component.
As was shown above in sect. 18.6, relative marker -Ji behaves slightly differ-
ently from other clitics, since it can optionally be used on both components of
a serial construction (704). Relative -Ji belongs to the core serialization, since
it characterizes the whole serial verb construction (703).
Serial verb constructions cannot be nominalized or undergo oblique or object
focusing (unlike Tariana, but similarly to Baniwa and Bare).
A serial verb construction may consist of two transitive verbs (754). It may
consist of one transitive and one intransitive Sa verb. If the first component
(e.g., a motion or stance verb) is intransitive, and the second one is transitive,
the whole construction is transitive:
If the first component (e.g., a phase verb or modal verb) is transitive and the
second component is intransitive, the whole construction is intransitive. The
whole construction has the transitivity of the second component.
A serial verb construction may consist of two S0 verbs (748) (in italics). In this
case, only the last one of the string has the S0 cross-referencing enclitic (749) (in
agreement with the general principles of cross-referencing, see sect. 18.4.2).
Serial constructions which consist of Sa and S0 verbs are extremely rare. One
example is in the sequence yutfi-Ji paJa-ha (strong-REL run-PAUS) '(who) is
stronger in running'.
Serial verb constructions cannot contain an Sj0 verb.
Serial verb constructions in Warekena are used to express the following
meanings: aspectual, modal, sequential, and cause-effect and oblique case-like
meaning.
• Aspectual:
(i) The motion verb fa 'go' has intentional-future meaning 'going to' (750) (the
serial verb construction is in italics).
(ii) Serial verb constructions with fa 'go' in Ipl form can have a cohortative
meaning (see sect. 11):
392 Aikhenvald
(iii) Serial verb constructions with the stance verb fia 'stay, sit, live' have the
meaning 'prolonged action':
ni-tsumeni wa yajapua-mia-ha
3pl-food then night-PERF-PAUS
Then they were eating their food, they ate (it), it was night.'
(iv) Serial verb constructions with the phase verb for 'start' have inchoative
meaning (745), and those with the verb for 'finish' have perfective meaning and
refer to a completed action (754).
• Modal:
(i) These include the verb weya 'want' and other verbs expressing desire, wish,
or liking:
(ii) Other verbs with modal meanings used in serial constructions are be 'can,
be able' (735), ise 'know, can':
• Simultaneous:
(759) guwacfeata le
cheat read
'He (rabbit) pretended to read (a piece of paper).'
• Cause-effect:
Serial verb constructions which contain two S0 verbs belong to this type
(748, 749).
(ii) In the verb of fearing. The verb of fearing consists of an S0 verb if it means
'be afraid' as a permanent state (without actually having to see the object one
is afraid of):
The verb of fearing which consists of akune 'afraid' and eda 'perceive'
means 'be frightened, fear something which is there as a real danger':
The verb of fearing is similar to a serial verb construction, since the same
subject constraint is fulfilled, and no other constituent can go between the two
components. It is negated similarly to a serial verb construction (766). It is differ-
ent from a serial verb construction in that the auxiliary-like verb follows the main
verb. Akune does not take the subject person-markers, which appears to be always
the case in serial constructions with an S0 verb in the first position (749).
Warekena 395
The link between serial verb constructions with an So verb, and adverbial
phrases can be illustrated by (770). The clitic pronoun -na 'Isg Ï or S0' occurs
on yutfi 'strong', the way it behaves in serial verb constructions, and the aspect
marker occurs on the verb. Yutfi is an adverb.
Adverbs are an open class of items whose primary function is verbal modifier.
Adverbs do not modify adjectives. They have no inflectional categories of their
own. There are two types of adverbs in Warekena:
(i) primary adverbs which constitute a smallish closed class.
These are, mainly, time and mode words. They are: umina 'a long time',
yaJiwa 'now', benamitfi 'long time ago', bena 'the day before yesterday', benata
'the day after tomorrow', yafia 'yesterday', tsutne 'almost', ubema Only', ma-
isena 'in vain', isina 'again,' and demonstrative adverbs discussed in sect. 16.4:
\vani 'here', wata 'there', and corresponding directional and elative forms. They
can occupy any position in the clause (61, 225, 772—775).
i-paka-mia-wa yu-tjiabu
3sgnf-split-PERF-NONACC 3sgf-belly
'The water boiled a long time, then her belly split.'
(ii) secondary adverbs, which are S0 verbs used in an adverbial function without
any special marking. Adverbs can be used as modifiers to verbs of all types.
They usually immediately precede or follow the predicate, and there is a
tendency that no other constituent should intervene between a predicate and its
Warekena 397
modifier. All adverbs often combine with -In- 'emphatic', -ba- 'augmentative'
and -ma- 'delimitative' (728, 775, 785). Semantically, these adverbs are
basically of mode type. They include one time word: yawaya 'early, tomorrow'.
Some verbal clitics can occasionally be used as adverbs; e.g., pafia 'future':
Particles
Warekena has three particles: wa 'presentative' (also used for coordination and
pivot changing; discussed in sect. 9), ai 'then' (405; a loan from Portuguese)
and a rarely used idi 'then' (it may be a loan word from Bare where it is very
frequent), e.g., (788), where idi also changes pivot, similarly to wa.
yapa yapewe
flat, stone flat, stone + LOG
Warekena 399
'Then the rabbit went to take tucum to eat, then he arrived on a flat
stone, the flat stone was a good one, on a stone.'
Negative proclitic ya- can be considered a particle. Warekena has a few conjunc-
tions: coordinating e 'and' (Portuguese loan), subordinating ate, te 'until' (Portu-
guese loan), \vaJi 'where'. Si 'úÃ (Portuguese loan) is used occasionally (716).
Phonology
22 Phonology
stop p t k
voiceless
stop b d (g)
voiced
fricative J h
affricate ts tj"
voiceless
affricate (ö)
voiced
nasal m n Þ
lateral l
flap
glide w y
p, t and k are voiceless unaspirated stops. Unlike Bare and Baniwa of Iijana,
Warekena does not have aspirated stops. These stops do not have any restrictions
as to their occurrence. Examples: pi-tfipana 'your hair', paJatana 'banana', tapa
400 Aikhenvald
'walk', pata 'work', kuJua 'drink', deka 'have', akune 'fear.' p can be optionally
realized as b in the word-initial position, if followed by a voiced consonant in the
second syllable, e.g., pa-buya or ba-buya One (period of time)', peda or beda
(pa-eda: IMP-perceive) 'it is seen'12.
b, d and g are voiced unaspirated stops, b and d occur in any position; both
are rather rare (as is the case with b in the majority of Northern Maipuran
languages); e.g., biyuJi 'evil spirit', be 'can', baJede 'long', benefi 'nothing',
bafaka 'tear'; yubua 'dig', yamadu 'evil spirit', duwUi 'crocodile', eda 'per-
ceive', g is a rare phoneme attested in a few words in the word-initial position:
ganu 'left-hand side', guwe 'bark' (ideophone?), guwadjata 'cheat', guJawa
'guava', guwifa 'peel' and in one word in an intervocalic position: anagudua
'dirty'. The existence of g is very unusual for a Northern Maipuran language.
Alveopalatal fricative / is attested in every environment and is frequent. It
can have an alveolar fricative s as a free variant. Examples: ifina 'name', fupe
'many', yafapua 'night', ifi 'what', fa 'go'.
Glottal fricative h has a very limited occurrence. It occurs exclusively in
pausal forms (see sect. 22.4.5) and triggers vowel nasalization, translaryngeal
vowel harmony, and other vowel changes; e.g., nuya-hä T, ibu-hu 'his head'.
It also occurs in loan words from Spanish: kunehu 'rabbit' (which has a variant
kune), naräha Orange'.
Voiceless affricates is and tf occur rather frequently, e.g., yutfia 'kill', teJefi
'three', tawape-tfi 'jungle',petfi On', tse 'know', isapia 'jump', tsawifa 'shoot an
arrow', matse 'lest'. In loan words from Portuguese and Lingua Geral, s is
treated as is, e.g., Diutsu 'Deus, God', isemana 'week', tsuJudawa 'soldier',
kamitsa 'shirt', kaisa 'marry', isiJuJa 'trousers', matsuka 'manioc flour', is and $
are frequently confused in non-loan words before the stressed high front vowel
/-; e.g., isina, tfina 'tell'; yatfipe, yatsipe 'ground'; -$i, -tsi 'repetitive', tfinu, tsinu
'dog'; and before the vowel sequence ia in a non-word-initial position; e.g.,
yutfia, yutsia 'kill', pi-tfiapuJi, pi-isiapuJi 'your bottom' (this does not occur in
loan words: IsUuJa ,*tfiJuJa 'trousers', from Lingua Geral). This confusion does
not take place before ia in the word-initial position, e.g., tfiapi 'intestines',
tfiabutfi 'his stomach', tfiapuJi 'his bottom'.
/ and if, but not / and ts can co-occur within one root; e.g., futfi (never
pronounced as futsi) (sequence tf-f has not been attested). / and ts can occur
together in one phonological word, where / belongs to the root and ts/tf to the
clitic tfi/isi 'repetitive'; e.g., fa-mia-tfi-wa, fa-mia-tsi-wa 'he went again'. Two/
can co-occur in one root; e.g.,/e/e 'dry', and so can two tf: tfitfi 'stink', tfiabutfi
'stomach', tf and ts cannot co-occur in one root or even in one phonological
word. Two ts cannot co-occur in one root or one phonological word.
There is also some degree of variation between / and tf before the unstressed
high vowel / in the suffix -tfi; cf. ta\vape-tfi, tawape-fi 'jungle', petfi, pefi On'.
This variation between the three sibilants j", tj" and ts in Warekena may be
interpreted as a symptom of a language death situation, whereby the phonologi-
cal complexity of the language is being reduced.
Warekena 401
22.1.2 Vowels. Warekena has four oral vowels, with long counterparts shown
in Table 5.
Every vowel also has a nasal counterpart with a limited occurrence: nasal vowels
occur in pausal forms (see sect. 22.4.4) as the result of a nasalization triggered by h,
and in rapid registers as the result of a phonological process: vN > v.
Central vowels can be realized as ts in post-tonic position in pausal forms
(see sect. 22.4.5).
Examples of high front i: Uuami 'his father', ifiwa 'from him'. Examples of
i: ni:da 'they see/perceive', umi:na 'long time', abi:da 'pig'.
Examples of central e: etene Old man'; ale 'how', epaJu 'he'. Long e: yame:
'far', e:ma 'tapir', ne:da º see/perceive'. In rapid speech registers short i and
402 Aikhenvald
e are neutralized to å in words with more than two syllables in word final
post-tonic syllables, e.g., etens Old man', piiaJe 'your tongue\fabapaJe 'liver'.
This does not happen in disyllabic roots, e.g., pane 'liver', tsape 'leaF.
High back rounded u has an allophone o, frequent in rapid speech registers;
e.g., yofana or yufana 'voice, speech, word', nofumiawa or nufumyawa 'my
wife'. Its long counterpart does not have this allophone: u:Je:Ji 'sounding',
\vadu:Je Our catch', yupu:Ji 'her eye'. In pausal forms, word final u can
alternate with / (see sect. 22.4.4: (801, 802)). ï occasionally appears in loan
words, in stressed syllables; e.g., flore 'flower', ora 'hour', kopu 'glass' (but
anu 'year'). Istoria 'story' has a variant isturia.
Examples of central unrounded a: ale 'how', ma 'say, do', afi 'manioc', a can
be reduced to 3 in word-final post-tonic syllables: nutanipaJd 'my ear'. Exam-
ples of long a are: a:fi 'fire', a:tapi 'tree', ma:pa 'honey', ka:ka 'manioc bread'.
Two long vowels are rare in one phonological word; e.g., u:Je:Ji 'sounding'.
Vowel length has a low functional load, and there is a considerable degree of
variation in current speech. Long vowels are regularly shortened in phonological
phrases with a stress shift (see sect. 22.4.3); e.g., e:ma 'tapir', pane ema 'tapir's
liver', and when derivational affixes are added; e.g., a:fi 'fire', afi-tfi 'firewood'.
Presentative wa can be realized as \va: when not a clitic (see sect. 22.4). Long
vowels can be pronounced as short; e.g., e:ma, ema 'tapir', ne:pitfi, nepitfi 'food,
something to eat', a:fi, afi Tire', paya.Ju, payaJu 'all'. Vowel length can 'float'
within a phonological word, i.e., the originally long vowel is pronounced as a short
one, and another vowel becomes long: yame:, ya:me 'long'. This variation in vowel
length can be considered a symptom of language obsolescence (cf. similar phenom-
ena in Tariana, also an endangered Northern Maipuran language).
- high + high
> > + cons
- cons - cons
4. Vowel nasalization
Phonetically nasalized vowels precede nasal stops (regressive nasalization).
Another case of regressive vowel nasalization occurs in rapid speech register,
where post-tonic ni is reduced to nd > (of the preceding vowel):
In loan words from Spanish and Portuguese syllables of Cv and CVN structure,
there are the following possibilities of adaptation of the loan words: (i) v > V(v)
in word-final position: nuve 'cloud', salau 'salon', kapitau 'captain'; (ii) CVN >
Cv before glottal fricative h: naräha Orange' (Spanish naranja); (iii) CVN > CVN:
mundu 'world', or CVN > CVNi: -wenita 'buy, sell' (Portuguese, Spanish vender
'sell'), depending on the degree of adaptation of the loan word.
If the phonological word contains the proclitic ya-, or wa- (sect. 7.1), the
bound enclitics are attracted to it. The possibility of attraction is the main
criterion for distinguishing an enclitic from a suffix.
The sequence proclitic -f enclitic(s) behaves as an independent proclitic. It
can be optionally cliticized to the following verb form in rapid speech:
ya-mia-ni-yue-pia tenepu
NEG-PERF-3pl-for-NEG road
'They did not know where to go, there is no road for them.'
Compare (794), where the pausal marker goes on the whole sequence procli-
tic + root + clitic + predicate, and ya-mia is cliticized.
There are the following rules of clitic sequencing in a clitic string: aspectual
clitics (e.g., -mia 'perfective') are followed by relativizer -Ji, which is followed by
the personal enclitics. In my corpus only sequences of no more than two enclitics
are attested, e.g., aspectual enclitic - personal enclitic in (795) and aspectual
enclitic - relativizer in (796):
(796) wiyua-mia-li
die-PERF-REL
'the one who died'
Enclitics always follow affixes. There is, however, one problematic case which
may be considered an instance of endoclisis. As was shown in sect. 18.2.2, aspec-
tual -wa 'unaccomplished' is not an enclitic, since it does not undergo attraction to
the negative ya- (examples in sect. 18.2.2). However, when it co-occurs with
aspectual -mia 'perfective', both sequences -\va-mia, which is to be expected, and
-mia-wa are possible, but with a semantic difference (see sect. 18.2.2).
Independent clitics include a number of items which can form an independent
phonological word under certain conditions (see below). The independent proclitics
are the sequences ya 'negative' + enclitic (illustrated above) and presentative wa
+ enclitic. Bound proclitic ya- 'negative' is extraprosodic, i.e., it does not affect
the stress placement in a phonological word: ya-nupa-pia 'he did not come'. An
unusual property of the proclitic ya- is that, if a pausal marker -hv is attached to
it, it forms an independent phonological word, e.g., sentential negation yahä 'no'.
This word is stressed on the last syllable: yahä 'no' (see sect. 22.4.3). Otherwise
pausal marker -hv is never stressed, e.g., wa-hä 'presentative-pausal'. Presentative
wa can be optionally used with a pausal marker. There are two ways in which wa
is used: as a proclitic, and as an independent phonological word (with or without
the pausal marker, see sect. 22.4.5). Wa as an independent phonological word is
often realized as wa: (33). Wa as an independent phonological word is frequently
used narrative-initially as a presentative (31, 68, 74), clause-finally to resume a
paragraph (43, 84, 103), to mark an important action (33, 101), and to mark the
change of pivot from S/O to S/A (see sections 3 and 9.2). Wa as a proclitic is
frequently used in a connective sense (11, 13, 22, 26, 85), or as a presentative
clause-initially, but not in the beginning of a narrative (84, 98, 99). Wa as an
independent phonological word and wa as a proclitic occur together in (26), (277)
and (425). In these cases, wa-hä is used in a resumptive sense ('and so, and then')
and wa- is used in presentative sense in (26) and (425) and to mark a sequence of
events in (277). Aspectual bound clitics are attracted to wa(-hä) when it is used in
copular clauses (60, 63; see sect. 7.1).
Phonological processes which occur at proclitic and affix boundaries are
different. For instance, at an affix boundary a- + e > e; e.g., wa-eda > weda
'we see, perceive'; and at a clitic boundary a + e > ae; e.g., ya-eda-pia 'he
does not see'; wa-eda 'then he saw' (see sect. 22.5).
Independent clitics are eya 'the one mentioned in the previous text', eni 'this;
near demonstrative' (797); eta 'that; distant demonstrative' (798); jfi 'what' (799);
e 'this, just mentioned' (406). They can be used as independent phonological words
(with or without a pausal marker -hv), if they have to be specially emphasized (eni
in 417, eta in 421, 423), or topicalized (e in 420), or are used headlessly (eni in
413, e and eya in 410). The independent clitics usually form a separate phonologi-
cal word, if a word-final marker -hv occurs on the previous phonological word or
408 Aikhenvald
one of the conditions for the use of a pausal form described in sect. 22.4.5 applies.
When cliticized, eya, e and ifi are proclitics; eni and eta can be used as either
proclitics or enclitics, depending on which constituent is more focused. For exam-
ple, in (415) the fact of the deer's death is important, and so eni 'this' is encliti-
cized to the verb, and in (419) turtles are in focus, and eni is procliticized to the
subject constituent, kuJimaJu-nawi 'turtles'. When used as enclitics, they form one
phonological word with the preceding item, and are extraprosodic. The following
vowel changes occur at clitic boundaries: vowel reduction: ae > a (797), ie > i
(798), vowel fusion: ai > e (799). Only the vowel fusion rule is similar to the
process which occurs at an affix-root boundary; other rules are different (see sect.
22. 5). When used as proclitics, they lose the first vowel.
In Warekena, all prefixes and the majority of suffixes are monosyllabic. There
are a few disyllabic suffixes; e.g., -wene 'elative', -iwe 'locative'. Bound clitics are
either monosyllabic, e.g., -Ji, or disyllabic, e.g., -mia 'perfective', -pia 'negative',
-yaJu 'yet'. Independent clitics are disyllabic, e.g., the demonstratives eni, eta. The
restriction concerning the maximum length of a root of two syllables seems not to
hold any more. Noun roots are usually disyllabic e.g., anu 'arm', api 'hand', tfinu
'dog'. Noun roots with three or more syllables are very frequent, and in the
majority of cases a historical and comparative analysis of these roots reveals that
Warekena 409
they contain a fossilized derivational affix (or classifier); e.g., a:tapi 'tree' (cf. -pi
'classifier for long objects'), nu-tani-paJa 'ear', nu-tfi-paJa 'leg' (-paJa 'body
part').The root fimapi 'bone' is a fossilized compound: fima 'fish' + api 'bone'.
The majority of verbal roots are either disyllabic, or monosyllabic; e.g., nupa
'come', yama 'drown', eda 'see/hear, perceive', fa 'go', ma 'do, say', e 'eat'.
Trisyllabic verbal roots contain a thematic syllable (see sect. 18.4.1 on the classi-
fication of verbs); e.g., yeJie-ta 'arrive', waya-ta 'talk', bitfi-ka 'go out'; however,
in some cases, the origin of the third syllable is hard to determine, e.g., yuJua 'fall',
yubua 'dig'. There is one four-syllable verbal root (probably a fossilized com-
pound) yawaputa 'answer'.
There is a tendency to avoid monosyllabic phonological words with short
vowels in Warekena (see sect. 22.4.1 on vowel lengthening in wa 'then' when
it is used as an independent phonological word). Every monosyllabic word tends
to be treated as a proclitic with respect to the following phonological word. For
example, in (800) the monosyllabic form of the verb e 'to eat' forms one
phonological word with its direct object pane 'liver'.
with respect to stress to yahä 'she eats' (yu-a-ha '3sgf-eat-PAUS', see sect.
22.5).
Prefixes and proclitics never affect the stress placement; e.g., nu-nupa
came', nu-tfinu-ne 'my dog'.
Suffixes divide into prosodic type, which affect the stress placement, and
extraprosodic type which do not affect stress placement.
When a prosodic suffix is attached to the root, the stress moves to the final
syllable of the root, i.e., the penultimate syllable of the word. Examples of
prosodic suffixes: -fi 'non-possessed, nominalization': pani-fi 'house', aya-fi
'food'; -tfi 'locative?': tawape-tfi 'the place with much jungle'; -ina 'agentive
nominalization': atsetena 'the one who teaches'.
In the case of extraprosodic suffixes of one syllable, the stress falls on the
antepenultimate syllable. Examples of extraprosodic suffixes: thematic syllables
in verbs; e.g., bftfi-ka 'go out', \vaya-ta 'speak', aJi-ta 'tie'; possessive suffixes
-Je, -ne, -te; e.g., nu-muJupa-Je; -Ji 'adjectivizer'; e.g., feJe-Ji >'black'; -iwe
'locative'; e.g., tenepuwe (tenepu-iwe) On the road'.
There is only one disyllabic extraprosodic suffix: -waba 'directional'. The
stress shifts to the final syllable of the root, i.e., the antepenultimate syllable of
the word, if the extraprosodic disyllabic suffix or a disyllabic enclitic is at-
tached to it: tenepu-waba (road-DIR) 'to the road'; yeJeta-mia-wa (arrive-PERF-
NONACC) 'he is coming'.
Other disyllabic suffixes are stressed on the penultimate syllable, and the root
they are attached to retains its original stress, but this is weakened (shown with "
as in the words that follow):- natfi 'locative nominalization'; e.g., wuninatfi 'a place
with much water', -pitfi Object focus'; e.g., küJua-pitfi 'something to drink'; -buJe
Oblique focus'; e.g., küJua-buJe 'something to drink with/from'; -nawi 'plural;
e.g., wUubeJu-nawi 'children'. The diminutive suffix -tui behaves in the same way,
and this is another piece of evidence in favour of the fact that vowel sequences are
not diphthongs in Warekena, e.g., aJetß-iui 'very small'.
The only exception occurs when a disyllabic affix is attached to a monosyl-
labic root, either a verbal or a numeral root. Then the stress shifts to the
antepenultimate syllable, e.g., e-pitfi 'something to eat', e-naba 'two (generic
class)', ba-buya 'two (cycles)'. There is a certain degree of variation in stress
placement in these cases, and occasionally the penultimate syllable is stressed,
e.g., babuya 'two (cycles)'. In verbs, a reduplicated syllable is always ex-
traprosodic, e.g., nu-tapa-pa (Isg-walk-RED) walk (much)'.
As was pointed out above, all verbal enclitics are extraprosodic, e.g., yeleta-
mia 'he is coming'.
(801) kuluka-paJuhi
make.hole-PURP + PAUS
'to make a hole'
The final boundary segment is extraprosodic, in the sense that it does not affect
the rules of stress assignment. Unlike other extraprosodic units, however, it always
has a secondary stress. It is not obligatory, i.e., it can be omitted under certain
circumstances (see sect. 3.1); e.g., ibu or ibuhü 'head', fimapi or fimapihl 'bone'.
The final boundary marker functions, among other things, as a token for recogniz-
ing phonological words in Warekena. For instance, in cases like yuJua-piani (fall-
NEG+this) 'this did not fall', wa-kuJuka-mia (then-tear-PERF) 'then he tore (it)',
it is possible to use the final boundary marker after the first word in each case,
yielding yuJupia-ha eni-hJ and wa-hä kuJukamia-ha. This shows the virtual
phonological independence of the cliticizable morphemes eni 'this', and wa 'then'.
In slow speech, no stress shift or stress weakening occurs. The word bound-
ary marker -hv in Warekena is used to signal the boundaries of phonological
phrases preceding pauses. A pausal form of the last phonological word in a
phonological phrase is used to signal a boundary under the following conditions:
(a) the constituent or a part of it is topicalized or is a part of an afterthought,
or
(b) it is under special emphasis, or
(c) it is followed by a pause.
If one or more of these conditions apply, pausal forms can also be used with
independent enclitics, as in (804), where both occurrences of eni-hl 'this' are in
a pausal form, since they are emphasized and topicalized; waJamatfiaha 'he
saves' is in a pausal form due to its utterance-final position. The first occur-
rence of eni-hi 'this' is followed by pause. (805) shows that a specific enclitic
(eni 'this') is treated as an independent phonological word, since the preceding
item is in the pausal form, and a significant pause follows. Verbs ni-wayata-hä
'they spoke', ni-ma-ha 'they said', ya-Ji-anetua-pia-ha 'is not good' are in
pausal form because they precede a pause. In (806) the pausal form is used to
indicate a pause after yu-yue-he 'to her'; it is also used to emphasize the
constituent pi-katsa-ha 'you will marry'. Wa-hä 'then' is in a pausal form, since
it is emphasized. In (807) a non-pausal form yue 'to him' is used since the
conditions necessary for the use of pausal forms are not met. YeJetahä 'he
arrived' is in a pausal form before a pause. Pausal forms are in italics in the
following examples.
waJamatfia-ha
save-PAUS
'This very God is good, it is he who saves.'
ya-Ji-anetua-pia-hä
NEG-REL-good-NEG-PAUS
'They spoke. They said, "This man is not good." '
The use of a pausal form for topicalization is illustrated by (808). The second
(or last) of the repeated constituents usually appears in pausal forms (as a kind
of afterthought) (809). More than one of the repeated constituents can appear in
a pausal form (110, 115).
Some items are only used in a pausal form. For instance, the general negation
yahä 'no' always appears in pausal form, as opposed to the negative proclitic
ya- 'not', which is always accompanied by the negative enclitic -pia (811).
Another item found only in pausal form is yumUehe 'when?'.
wa yu-ma-hä ya-hä
then 3sgf-say-PAUS NEG-PAUS
' "A snake did not come to you" (the man said), then she said: "No." '
Other optional pausal marking phenomena in the position before the end of a
clause or of a significant passage (813), or in right dislocation for clarification
(814), are:
(a) monophthongizing and centralizing vowel sequences before the pausal
marker: ia > ie: -mia-ha 'PERF-PAUS' -miehe in (813) and ia > e: -pia-ha
'NEG-PAUS' > -pehe (814); optionally ia > i: pia-hä 'NEG-PAUS' > -pihl
(503); -mia-ha 'PERF-PAUS' > mihl (22).
wa atuJapi-mehe
then full-PERF + PAUS
Then he was not full, he was not full, the jaguar. Then he has become
full.'
(b) centralizing back vowel a > e before the pausal marker (815); cf. also yehe 'no'
in (196), -wehe 'PERL + PAUS' in (299), nedehe (1 sg + see + PAUS) in (101).
(c) centralizing high vowel ; > e in the word-final position before the pausal
marker; e.g., (428) ninapa-mia-nehe (3pl + fmish-PERF-DEM:PR + PAUS) 'they
finished'.
(d) centralizing back vowel a > e in the word-final position; e.g., (482) \vayate
'speak'; or before a clitic; e.g., (232) yuJute-Ji-wa (lie-REL-NONACC) 'where
he lay.
(f) vowel dissimilation: aJe-he 'this-PAUS' > aJe hi (535), wa-hä 'then-PAUS'
> wahe (642).
ya-tse-pia-ha uwa-ha
NEG-know-NEG-PAUS climb-PAUS
'When people climb a tree, he cannot kill (them), he cannot climb.'
Thus, Warekena makes a distinction between different kinds of 'pauses' and this
corresponds to the prosodic domains of an utterance and a phonological phrase
respectively. The principle is similar in Bare. Both languages use different devices
for utterance-final and non-utterance-final pause-marking. However, in Bare it is
416 Aikhenvald
22.4.6 Intonation. Little is known as yet about the intonation patterns in Ware-
kena. All types of questions are characterized with a rising intonation, and
declarative sentences have a falling intonation on the last word. Enumeration is
characterized by a rising intonation on each component. Serial verb construc-
tions are a single intonational unit (see sect. 18.8.1). When a constituent has to
be emphasized, emphatic falling intonation is used (277).
(a) u + a > a
Examples like (iii) also show that sequences like Vi, Vu in Warekena cannot
be considered triphthongs (in fact, they should be considered vowel sequences).
(b) u + e > e
In slow speech register, i -I- e is sometimes realized as ie; e.g., ni-ema >
nema. or niema 'they call'.
There is one exception, when i + e > i:
(d) u + i > u
(e) e + i > e
(f) a + e > e
(g) a + i > e
Morphology
23 Morphology
23.2 Word classes. Major open lexical classes in Warekena are: nouns, verbs,
adjectives, adverbs. They differ as to their inflectional categories and syntactic
functions.
Nouns are arguments of verbs and heads in head-modifier constructions.
They have the following inflectional categories: inherent gender, number (op-
tional), possession (obligatory), inherent class (optional), peripheral cases
(obligatory, depending on the semantics of the nouns). They also have a number
of derivational devices distinct from other categories.
Verbs are predicates. They have the following inflectional and derivational
categories: person/gender/number, transitivity and active/stative, causative, re-
flexive, oblique and object focus; aspect/tense (optional), emphatic.
Adjectives are modifiers in head-modifier constructions. They have the fol-
lowing inflectional categories: number (optional), concordial gender (optional);
and a number of specific derivational devices. They are a large open class, and
are regularly derived from verbal roots (So verbs). The only nonderived adjec-
tives are pronominal (demonstrative) adjectives.
Adverbs are modifiers of verbs. They do not have any specific inflectional
categories. They can be shown to have, historically, a deverbal origin, going
back to serial verb constructions (sect. 18.8.2).
Closed lexical classes in Warekena are: quantifiers, numerals (they have
classifiers which distinguish them from quantifiers), pronouns, particles, con-
junctions, adpositions.
Warekena displays a certain degree of fuzziness between the lexical classes.
Adverbs and S0 verbs can be distinguished only according to their syntactic
functions, and are sometimes indistinguishable. Any noun can be used to func-
tion as an S0 verb. Quantifiers and numerals tend to collapse together, due to
partial loss of the classifier system in the situation of language obsolescence.
Adpositions can be considered as a subclass of obligatorily possessed nouns.
24 Ideophones
Notes
The Maipuran family, whose genetic unity has been clear since 1783 when
Pe. Gilij established genetic affinity between Maipure language in Vene-
zuela and Mojo in Bolivia, is known among South American scholars as
Aru k (or Arawak). Claims for a larger family including Maipuran and
other families such as the Arawa languages have not been substantiated (see
Payne 1991:360-365).
Third person singular masculine is usually realized by the absence of an
overt prefix in verbal, nominal and adpositional inflectional paradigms, and
so has been omitted from both sentence transcriptions and morpheme
glosses; see Table 1, sect. 15.2.1, for the full chart of prefixes.
The term pivot refers to the syntactic constraints on clause combination or
on the omission of coreferential constituents in clause combinations. If a
language treats S and O in the same way and A differently in terms of these
constraints, it is said to have an S/O pivot, or to be 'syntactically ergative'.
If S and A are treated in the same way and Ï differently, the language is
said to have an S/A pivot and to be 'syntactically accusative'. If a language
combines both types of constraints, and has an S/O pivot under some
conditions and an S/A pivot under others, it is said to have a mixed pivot
(see Dixon 1994: 143ff).
Syntactic causatives in Warekena are strikingly similar to syntactic causa-
tives in Bare; e.g., Bare:
Compare -{â/â as an affix used in place names and a few nouns, e.g.,
wene- 'Xie', yapi-tfi 'ground' (see also other examples in sect. 15.4.1); as
a part of collective marker -natfi, as well as in the postposition petfi On';
cf. locative -Jiku in Baniwa, which corresponds to a derivational suffix in
Tariana.
-Wa 'PERL' is used with deverbal nouns to form adverbial-like expressions,
e.g., dabana-wa 'in the first place, first, erstwhile' (768), inapa-wa 'last';
see sect. 18.8.2, about the problems which arise here:
422 Aikhenvald
nu-tfmu-ni
Isg-dog-POSS
'my dog: the dog I brought up'
nu-tjmu-te
Isg-dog-POSS
'my dog: the dog I found'
Acknowledgments
I am extremely grateful to Des Derbyshire, R.M.W.Dixon, W.P. Lehmann,
Rute Amorim, Tatiana Baltar and Cristiane Cunha for their comments and
assistance. I am also thankful to Valteir and Silvana Martins and Lenita and
Elias Coelho for their patience and hospitality. This grammar could not have
been completed without the patient assistance of Humberto and Manoel Baltazar
and Pedro Ängelo Tomäs, native speakers of Warekena of Xie.
Abbreviations
ADJ adjectivizer
AFF affix
AG agentive
AUG augmentative
CAUS causative
CL classifier
COLL collective
DEC deceased
DECL declarative
DEL delimitative
DEM demonstrative
DER derivational suffix
DIM diminutive
DIR directional
EL elative
EMPH emphatic
f feminine
PUT future
HAB habitual
IMM immediate
IMP impersonal
INCH inchoative
INT intensive
LOG locative
MASC masculine
NEG negative
NOM nominalizer
NONACC non-accomplished
nf nonfeminine
424 Aikhenvald
NPOSS non-possessed
OBJ.FOC object focus
OBL.FOC oblique focus
PAUS pausal
PEJ pejorative
PERF perfective
PERL perlative/transformative
pi plural
POSS possessive
PR proximate
PRES presentative
PROB probability
PROH prohibitive
PURP purposive
REC reciprocal
RED reduplication
REFL reflexive
REL relative
sg singular
REM.P remote past
REP repetitive
TOP topic
WARN warning
Warekena 425
References
Aikhenvald, Alexandra Y.
1994a "Classifiers in Tariana," Anthropological Linguistics 36:4: 1-62.
1994b "Grammatical relations in Tariana," Nordic Journal of Linguistics
17:2: 201-218.
1995a Bare. (Lincom Europe Materials: 100)
1995b "Person-marking and discourse in North-Arawak languages," Stu-
dia Linguistica 49, 2:152-95.
1995c "Topic advancing voice in Baniwa of Igana and Tariana," 20pp,
ms.
1996a "Classe nominal e genero nas linguas Aruäk," Boletim do Museu
Goeldi, Vol. 10(2) 1994:137-259. [Published in 1996.]
1996b "Words, phrases, pauses and boundaries: evidence from South-
American Indian languages," Studies in Language 20:487-517.
In prep. "A grammar of Tariana," ms.
Aikhenvald, Alexandra Y., and Rute Maria Coelho Amorim
1995 "Warekena in Brazil," Opcion, 18:29-44.
Anonymous
1933 Pelo Rio Mar. (MissOes Salesianas do Amazonas). (Esta-
belecimento de artes gräphicas. C.Mendes Junior, Rio de Janeiro).
Briizzi, Alconcilio Alves da Silva
1961 Discoteca etno-lingüistica-musical das tribos dos rios Uaupes,
Ifana e Cauaburi (Centro de Pesquisas de lauarete da Missäo Sale-
siana do Rio Negro, Säo Paulo).
1977 A civilizacäo indigena do Vaupes (Las: Roma).
Bull er, Barbara, Ernest Buller, and Daniel L.Everett
1993 "Stress placement, syllable structure and minimality in Banawä,"
International Journal of American Linguistics 59:276—89.
Civrieux, Marc and Reno Lichy
1950 "Vocabulärio de cuatro dialectos Arawak del rio Guainia," Boletin
de la Sociedad Venezolana de Ciencias Naturales XIII (77): 121—
159.
Comrie, Bernard
1981 Language Universals and Linguistic Typology (Oxford: Black-
well).
426 Aikhenvald
Mosonyi, Esteban
1968 "Introduction al analisis intraestrutural del idioma Baniva,"
Economia y Ciencias Sociales, 10:3:65—70.
Nimuendaju, Curt
1982 Textos indigenistas (Säo Paulo: Edisoes Loyola).
Payne, David L.
1991 "A classification of Maipuran (Arawakan) languages based on
shared lexical retentions," Handbook of Amazonian languages
vol.3, edited by Desmond C. Derbyshire and Geoffrey K.Pullum
(Berlin: Mouton de Gruyter), 355-499.
Rodrigues, Aryon D.
1986 Linguas brasileiras. Para o conhecimento das linguas indigenas
(Edi^Oes Loyola, Säo Paulo).
Schachter, Paul, and Fe T. Otanes
1972 Tagalog reference grammar (Berkeley: University of California
Press).
Schomburgk, Richard H.
1841 Reisen in Guiana und am Orinoco während der Jahre 1835-1839
(Leipzig).
Voegelin Carl F., and Florence M. Voegelin
1977 Classification and index of the world's languages (New York: El-
se vier).
Wallace Alfred R.
1853 A narrative of travels on the Amazon and Rio Negro, Reprinted in
1972 (New York: Dover publications).
Wise, Mary Ruth
1991 "Morfosintaxis comparativa y subagrupaciones de las lenguas
arawakas maipurän," Revista Latinoamericana de Estudios Et-
nolingiifsticos. Lingüistica Arawaka, 259-282.
428 Aikhenvald
Appendix
Warekena Text
(24) wa ni-yuleta-mia-wa
then 3pl-return-PERF-NONACC
of the road, in front," he said, "another one will be far, far on by the
edge of the road", so he left (them), he left (them) far, far on the edge
of the road.'
(119) wa-yuleta-wa
Ipl-return-NONACC
'He said to the family, he said to the family, "Now let's go back." '
ni-Ja-mia-wa
3pl-go-PERF-NONACC
Warekena 439
Margarethe W. Sparing-Chavez
Summer Institute of Linguistics
0 Introduction
1 General characteristics of Amahuaca
1.1 The case-marking system
1.2 Tense/aspect system
1.3. Operators that function as IRMs
2 General description of IR-constructions in Amahuaca
3 IR-operators in Amahuaca
3.1 Set A
3.2 Set B
4 Conclusions
5 IRMs in Narrative Discourse
5.1 Sample narrative
5.2 Analysis
5.3 Participant Orientation
0 Introduction
Ever since William Jacobson coined the term "switch-reference" in his seminal
paper of 1967 entitled: "Switch-reference in Hokan-Coahuiltecan," linguists have
detected this syntactic clause-linking device not only in the Americas, but all over
the globe: in New Guinea, Australia, and Africa. Switch-reference as defined by
Jacobson (1967:240) "consists simply in the fact that a switch in subject or agent is
obligatorily indicated in certain situations by a morpheme, usually suffixed, which
may or may not carry other meanings in addition." Meanwhile switch-reference has
become an important issue in typological studies, and the experts speak of "canoni-
cal" (Haiman), "prototypical" vs. "non-prototypical" (Comrie), "anticipatory" vs.
"non-anticipatory" and "real" vs. "unreal" (Givon) switch-reference systems. Al-
though the scope of the phenomenon described has changed, the basic assumption
that this type of clause-linking device primarily encodes referential relations has
remained the same.
Amahuaca, like other members of the Panoan1 language family of South America, has
a well developed clause-linking morphology which exhibits all of the characteristics of
444 Sparing-Chavez
classic "switch-reference" systems (Jacobson 1967) plus the additional feature of encod-
ing coreference between subjects and objects. It is this feature that makes the system
very productive. However, maintaining referential continuity is only one of the functions
of this system, and not necessarily even the most important one. Other functions include
encoding transitivity and temporal or logical relations between events. Therefore the
term "switch reference" does not accurately describe the system and I will refer to it as
"interclausal reference system "(hereafter IRS), following Franklin (1983).
My purpose in this discussion is twofold: 1. to describe the IRS of Amahuaca in
terms of previous typological characterizations of switch reference (e.g., Munro
1980, Haiman and Munro 1983, Haiman 1983, Connie 1983, Givon 1983, and
Longacre 1983), and 2. to question the assumption that clause-linking morphology
primarily establishes cross-clausal reference between subjects or agents. Instead, I
suggest that in Amahuaca to a large extent temporal and/or logical relations take
precedence over referential ones. I will attempt to show in this study that two
systems are basic to the IRS: case marking and tense/aspect, and that the referential
parameter is subsidiary to the temporal parameter. These claims disqualify Ama-
huaca as a "prototypical" switch-reference language, as described by Comrie
(1983:36). I therefore propose a more general typology of interclausal reference that
is not strictly concerned with the referential tracking function.
In Section 1, I describe some fundamental characteristics of Amahuaca and, as a
background to the investigation of interclausal relations, outline the case marking
and tense/aspect systems, as well as discuss a small number of suffixes which figure
in the IRS. In Section 2, I describe the basic properties of the IR suffix system.
Section 3 presents and illustrates the constructions in more detail with examples and
summary displays of the morphemes and their grammatical functions. In Section 4,
I summarize my findings and relate them to the studies previously mentioned.
Finally, in Section 5, I analyse a short text, and briefly discuss the pragmatic
functions of the IRS.
Amahuaca is spoken by approximately 1,000 people in the lowlands of southeast-
em Peru. The people live widely dispersed and there are some dialectal differences.
But to my knowledge they do not affect the IRS. My language data come from field
work in the settlements of Nuevo San Martin on the Inuya River and Laureano on
the Purus River, during several trips from November 1985 to June 1992 under the
auspices of the Summer Institute of Linguistics.
phono logically bound to the last constituent of a clause or sentence. It is not at all
unusual to attach tense-aspect-person-mood morphemes to a noun or pronoun, or
case markings to verb phrases. Likewise, clause-level evidential operators can be
suffixed to almost any constituent. There is strong indication that the whole sen-
tence, rather than the constituent marked, is the scope of the affixation.
As is typical in clause chaining languages, Amahuaca distinguishes between inde-
pendent and dependent clauses. Independent clauses are inflected for
tense-aspect-person-mood, and usually are sentence final. Dependent clauses carry
the IR morphology.
Again following the general trend, Amahuaca is best classified as an SOV language
(Wise 1979), but not in an exclusive sense. Whereas dependent clauses (particularly
clauses in a chain with IR morphology) strongly prefer SOV/SV word order, inde-
pendent clauses and sentences display more freedom. Their word order depends on
verbal aspect3 and pragmatic considerations (which in the context of this paper means
that a sentence constituent or a clause is considered prominent). While independent
sentences in unmarked aspect have SOV/SV word order, those in marked aspect display
OVS/VS word order. Other orders (OVS/VS in unmarked aspect and SOV/SV in
marked aspect) are clearly pragmatically marked (e.g., 3a,b below).
1.2 Tense/aspect system. Amahuaca has very few free temporal adverbs. Instead,
it has a detailed grammatical morphology that encodes tense and/or aspect, and other
temporal nuances. For example, the system includes bound temporal adverbs encod-
ing different degrees of time lapses between events. In this brief sketch I will outline
the tense system only as far as it relates to the IRS. The tense/aspect operators
described are those used in declarative clauses. There are different sets of operators
for questions and commands.
As mentioned in sect. 1.1, the tense/aspect system of Amahuaca is divided into two
sets: unmarked and marked. They differ formally as well as semantically.
The unmarked aspect operators (Table 2) have the following formal characteristics:
the forms are one single unanalyzable suffix encoding either aspect only, e.g., -non
'prospective aspect,' or tense + aspect + person, e.g., -xo '3rd person, narrative past,
perfective.' They can be preceded by bound temporal adverbs, e.g., -shinxo º-3 days
ago, 3rd person, narrative past, perfective,' and preceded or followed by plural subject
markers, e.g., -haivo 'continuous aspect, plural.'
The marked forms (Table 3) have the following formal characteristics: they consist of
two separate operators, which can either occur together on the same consitutuent, e.g.,
-haxqui (3), or they can occur on different constituents, e.g., -hax. .. -qui 'past perfect,
3rd person actuality aspect' (9—11). The forms encoding tense/aspect, are often analyz-
able, e.g., -ca-tzi 'future' consists of -ca 'go' + -tzi 'commitment.' Like the operators
of the unmarked category, those of the marked category can also be preceded by the
bound temporal adverbs, resulting in hypermorphemes such as -shinax/-shinnax º-3
days ago' (from -shin + -hax. )
1.2.1 Unmarked tense/aspect set (Table 2). There are two past tenses in perfective
aspect: the narrative past and the immediate past. The narrative past has three different
forms: -cu '1st and 2nd person,' and -xo '3rd person' encode recent past; -tai which does
not inflect for person, encodes One planting season ago.'
(4) a. Hiya -x -mun hun -0 jo -cu -hnu.
I -NOM -TH I -ABS come -1/2REC.PAST.PFTV -DECL
'It is I who came.'
b. Jaa -x -mun jan -0 jo -xo -hnu.
he -NOM -TH he -ABS come -3REC.PAST.PFTV -DECL
'It is he who came.'
The above past tenses (with -tai as an exception) can combine with the following
operators expressing time lapses: -shin º-3 days ago,' -yan '4 days till a planting season
ago,' -ni 'a long time ago.'
The forms for immediate past are: -ha '1st person,' and -qui '2nd and 3rd person.'
(5) a. Moha -mun hun -0 jo -ha -nu.
now -TH I -ABS come -1IMM.PAST.PFTV -DECL
'Now I have (just/actually) arrived.'
448 Sparing-Chavez
1.2.2 Marked tense/aspect set. Where unmarked aspects encode events that are not
perceived as actually taking place at the present moment, marked aspects encode events
that someone is actually engaged in or will be engaged in. The most complex form in
this set is -hax (9a). It has imperfective as well as perfect characteristics. (As Comrie
has shown (1976:63), that is not self-contradictory.) It is imperfective in that it pays
attention to the internal structure of the event, and perfect in that the past event is
relevant to the present moment.
450 Sparing-Chavez
1.3. Operators that function as IRJVls. As mentioned in the Introduction, two sys-
tems are basic to the IRS: case marking and tense/aspect. A combining of the
operators of the two systems constitutes the IRMs. In addition a small number of
verbal suffixes encoding motion and possible events also figure in the IRS.
1.3.2 Suffixes of motion. There are five suffixes in this set. One of them, -tan,
expresses motion, 'go and come back,' and/or 'immediate action.' Four of them
express direction: -cahin and -vahin express 'movement away from a focal point,'
-quiran and -vuran express 'movement towards a focal point.' All of them imply
perfective aspect and encode sequential events in the IRS (13-17).
1.3.3 Suffix expressing possible event, -ti 'can,' 'be able' (45,46).
marked verb. Only where necessary for clarity the subject of the marked clause is
inflected for singular or plural.
In cases where there is overlapping reference in participant sets, i.e., changes from
plural to singular (we/I, they/he) or vice versa (I/we, he/they), the SS-morpheme is
used. This appears quite natural for the first person since the first person plural is
treated as a single form, but it also applies to the third person. This type of
construction is not very common in Amahuaca. I only found one example in some
sixty pages of oral narrative text, and upon trying to elicit examples, the native
speakers were in disagreement with each other. 13
Clause chaining is an extremely prominent and obvious aspect of Amahuaca
discourse structure. In particular, narratives typically consist of chains of several
clauses joined by IRMs. It is not unusual to find a series of ten or more clauses
which consist of little more than a chain of verb phrases followed by an independent
clause (sentence 4 in Section 5.1 is a typical example). Amahuaca narrative sentence
structure closely resembles Wojekeso of New Guinea and Guanano of Colombia,
whose "distinct and final verb at the end of the sentence is like an engine that pulls
a train of cars" (Longacre 1983:186). While the relationship between a marked
clause and its sentence-final controlling clause is one of dependence/independence
by virtue of the verb affixation of the final controlling clause for tense/aspect-per-
son-mood (the marked clause depends on the controlling clause for completion), the
relationship between neighboring marked clauses appears to be one of coordination;
the clauses are of equal rank. There exists, of course, a certain degree of dependency
between a marked clause and its adjacent non-final clause; the clause on the right
has some control over the one on the left. Comrie speaks of "stacked dependency,
i.e., where one clause is dependent on another that is in turn dependent on another"
(1983:23-25,37). However, since the marked clauses are the same structurally but
contrast with the final/independent controlling clause, it appears advantageous to
consider the marked clauses coordinates. If we consider, for instance, the clauses 2c,·
d, and e in Section 5.1, we note that the morphology of the two adjacent marked
clauses (2c,d) is very similar, but contrasts considerably with the final controlling
clause (2e).14
Again in line with the general pattern, it is the verb of the marked clause that
receives the IRM. There are, however, a few exceptions in which a noun or pronoun
receives the IRM (27,28).
With regard to the discussion on iconic marking,15 the Amahuaca constructions
display a divided picture. Several compound IRMs of Sets A and B abide by iconic
principles in that the IRMs are verbal (tense related) affixes attached to verbs.
However, those displaying case-marking morphology do not seem to abide by iconic
principles, at least not from a synchronic point of view. Case marking semantically
applies to nouns and pronouns, and one would expect them only to be formally
marked. The Amahuaca constructions at large, as the discussion on clitics in Section
1 confirms, support Comrie"s view that "the morphological marking is a charac-
teristic of the dependent clause as a w h o l e , . . . " (1983:23).
454 Sparing-Chavez
The above sketch shows that Amahuaca displays most of the properties that are
typical of switch-reference systems, as described by Comrie (1983:21-33). The
unusual features in Amahuaca, as well as in some other Panoan languages, are:
1) Sensitivity to subject/object coreference as well as subject/subject coreference
and non-coreference.16
2) Incorporation of case-marking morphology into the IRS.
3) Expression of transitivity as well as temporal/logical relations.
3 IR operators in Amahuaca
As mentioned before, there are two sets of IRMs in Amahuaca: A and B. The basic
references that the IRMs establish in Set A are the temporal ones, and in Set B the
referential ones. The IR morphology in both sets is a combination of the tense/aspect
and case-marking morphology.
The temporal relations encoded in the IRS include absolute and relative tenses.17
The relationships between two events can be (a) sequential, (b) non-sequential or (c)
subsequent.
(a) Sequential events refer to events that follow each other in chronological order.
They contrast along aspectual lines: (1) The action expressed in the controlling verb
immediately follows and might be the consequence of the action of the marked verb.
The action expressed in the marked verb has perfective aspect, i.e., it is considered
complete before the action of the controlling verb begins. (2) The action of the
controlling verb follows the action of the marked verb. The marked verb expresses
perfect aspect, i.e., it emphasizes that it is relevant to the following action. As we
noted in the tense/aspect system, so in the IRS the operators can combine with
temporal adverbs to express degrees of time lapses between the events of the marked
and controlling verbs. Thus -hax 'SS, transitive controlling verb, sequential actions'
+ -shin, º to 3 days ago' combine into -shinax/-shinnax 'the action of the marked
verb occurred yesterday or 1 to 3 days before the action of the controlling verb' (20).
Despite the time lapse between the events, the notion of relevance is preserved.
(b) Non-sequential actions or events (I am using the two terms indiscriminately)
refer to simultaneous actions which can be completely parallel, or partially overlap-
ping, i.e., the action of the marked verb is continuing while the action of the
controlling verb occurs.
(c) Subsequent action in the context of this paper means that the action of the
controlling verb precedes the action of the marked verb.
Depending on the context, temporal relations can be interpreted as logical relations.
For instance, subsequent action often expresses purpose (35,37-39); sequential action
can express potential fact (22), reason-result (47,49), or concession-contraexpectation
(19); simultaneous actions can be interpreted as condition-consequence (32).
In the DS and OS constructions there are two ways to indicate a plural subject of
the marked clause: (1) The subject is identified through a noun phrase. In this case
the noun phrase receives the plural case-marker, either ergative (-vaun) if governed
by a transitive verb, or absolutive (-vo) if governed by an intransitive verb, or
Amahuaca 455
3.1 Set A. This set consists of IRMs encoding sequential and non-sequential events.
3.1.1 IRMs encoding sequential events. There are two distinct types of IRMs that
encode sequential events: (1) suffixes of motion, and (2) forms that are combinations
of tense/aspect and case-marking morphemes.
The following operators can be subdivided into pairs, whereby the direction of the
controlling verb is the criterion for the division.
When the controlling verb expresses 'moving away from a focal point,' -cahin or
-vahin are used.
-cahin encodes valence 1.
(14) Junu -naqui pacuu -cahin -0 -mun hina -n pocu
river -middle jump -SQ.VAL1 -SS -TH tiger -ERG cross
-vahin -xo -hnu.
-go -3PAST -DECL
'Having jumped into the middle of the river, the tiger crossed (it) and
went.'
-vahin encodes valence 2.
(15) a. Haa jo -hato nincaa -vahin -0 -mun hun jato
tapir come -NONSQ.SO hear -SQ.VAL2 -SS -TH I them
yohi -cu -hnu.
tell -1PAST -DECL
'When I heard the tapir coming, I went to tell them.'
The following example shows that -vahin replaces -cahin when the intransitive
marked verb governs a plural subject.
(15) b. Junu -naqui pacuu -vahin -0 -mun
river -middle jump(ITR) -SQ.VAL2 -SS -TH
pocu -can -vahin -xo -hnu.
cross -they -go -3PAST -DECL
'Having jumped into the middle of the river, they crossed (it) and went.'
When a controlling verb expresses 'moving toward a focal point,' -quiran and
-vuran are used.
-quiran encodes valence 1.
(16) Hiya nu -ri xucuu -quiran -0 vacon vama
me here -ward approach -SQ.VAL1 -SS wasp(s) drive.off
-hi -yu.
-come -IMP
'Come close to me and drive off the wasps.'
-vuran encodes valence 2.
(17) Nocoo -haito vuchi -vuran -0 -mun muchi
arrive -NONSQ.SO see -SQ.VAL2 -SS -TH surround
Amahuaca 457
We note that the case markers and in most cases also the theme markers have been
suffixed to the marked verb. Also, the perfective aspect of the tense marker has been
changed to perfect aspect with present relevancy. This is especially evident in (20).
Although the IRM encodes a time lapse between the two events, the first event is
relevant to the second.
Amahuaca 459
-hi, . . .
-NONSQ(SS)ITR
' "Come, see for me, mother, if it (the clay pot) is ready," she (said).
Going (the old woman) looked,...
(28) Vachin min -hcun -mun jan yohi -xo -hnu.
quarrel you -SQ(DS) -TH he say -3PAST -DECL
'He said that you had been quarreling.'
(Lit.: 'Quarreling is what you had been doing . .. ')
3.1.2 IRMs encoding non-sequential events (Table V). When the temporal rela-
tion between the clauses is non-sequential, -hi and -quin express SS, -haito expresses
SO, and -hain/-haivaun DS. The SS morphemes also encode whether or not the
controlling verb is transitive, but the SO and DS operators do not.
3.1.3 Subsequent event. When the action of the controlling verb precedes the
action of the marked verb and/or is the reason for it, -xanhni/-catzi or xanhquin are
used to encode SS function, and -non DS function. -xanhni/-catzi specify that the
controlling verb is intransitive. The difference between the two forms is that -xanhni
emphasizes the logical relationship between the events and -catzi the temporal
relations. The IRM -xanhquin encodes transitive controlling verb, and -non does not
encode transitivity. The form -no (without the final case-marking) resembles -cu (see
Section 3.1.1.2.3) in that it compounds with morphemes encoding 'location,' (and
also with the verb to be), in order to function as an adverb of place: nuno '(being)
here,'jano '(being) there.'
pima -xanhquin.
light -SS.SUBS/PURP.TR
º will make a fire, therefore I am chopping wood.'
(The controlling clause is pragmatically marked.)
3.2 Set B (Table VT). This set resembles Set A in that the operators also contrast
along temporal lines: sequential, non-sequential, and subsequent. Sequential event
markers have implicit perfective aspect. Considering that several of the IRMs in Set
B function as relative clause markers this is not surprising. Where the IRMs of Set
A primarily relate events to one another, which is achieved by the perfect aspect
(present relevance), in Set B the focus shifts to relating participants (subjects or
objects) to events, which is achieved through perfective aspect. (The basic difference
between the two sets then is one of focus.) On the referential parameter the sets
differ. The operators of Set A distinguish between coreference and non-coreference
of subjects (SS, DS) and coreference of subject and object (SO). The operators of
Set B encode coreference only, but distinguish between coreference of subjects (SS)
and object with subject (OS). Several of the marked clauses in Set B are relative
clauses.
Since the focus in Set B is on participants, I will use the referential parameter as
the basic dividing line for grouping the IRMs.
Amahuaca 465
3.2.1 Same subject There are three pairs in this group. They encode whether the
controlling verb is transitive or intransitive, and they differ along aspectual lines:
~hahton/-hahtox encode perfective aspect, -haiton/-haitox repetitive/continuative as-
pect, and -tihton/-tihtox possibility.
3.2.2 OS function. There are two pairs in this set. They differ along aspectual
lines: -ha/-havo, encode perfective aspect, -hail-haivo encode repetitive/continuative
aspect. The morphemes ending in -vo specify that the plural subject of the marked
clause is unidentified. When the plural subject is identified by a noun phrase, the
noun phrase is inflected for plural ergative case.
4 Conclusions
In the previous discussion I have described and illustrated those interclausal relation
constructions in Amahuaca that encode temporal/logical and referential relations be-
tween a marked and a controlling clause. The ERMs were divided into two sets, A and
B. Set A primarily displays contrasts on the temporal/logical parameter, distinguishing
468 Sparing-Chavez
between sequence, non-sequence, and subsequent events, and secondarily on the refer-
ential parameter, distinguishing between SS, SO, and DS relations. Set B primarily
displays contrasts on the referential parameter, distinguishing between S S and OS
relations, and secondarily on the temporal/logical parameter, distinguishing between
sequential, non-sequential, and subsequent events. The operators are compounds of
tense/aspect and case-marking morphemes.
Comparing Amahuaca with some of the "switch-reference languages" described
by Comrie, Givon, Kaiman, Longacre and Munro, we note that it shares the follow-
ing properties with those languages: (a) Amahuaca is an SOV language, but not in
an exclusive sense; (b) it has formal distinctions between dependent and independent
clauses; (c) the IRMs encode both referential and temporal/logical information; (d)
the IRMs are suffixed to the dependent verb (except in cases of ellipsis of the verb
or pragmatically marked environments in which the verb precedes the subject noun
or pronoun); (e) the marked clause precedes the controlling clause (except in cases
of pragmatically marked environments); (f) the notions of "subject" and "object" in
the IRSs have syntactic rather than semantic functions; (g) the IRMs are used for all
persons; (h) the marked and controlling clauses are almost always adjacent to each
other. Only on rare occasions, for the sake of pragmatic marking, a marked clause
refers directly to a final controlling clause.
Other differences from most languages with IRSs are: (a) the case-marking and
tense/aspect morphology is incorporated in the IRS; (b) there are distinct operators
for SS, DS and SO; (c) in Set A temporal distinctions are more basic than referential
ones; (d) some of the IRMs encode transitivity of the controlling verb. None of these
properties, however, are unique to Amahuaca.
The question that needs to be asked now is a typological one: can the Amahuaca
system be classified as a "real" interclausal reference system? The above summary of
properties that the language shares with so-called switch-reference languages indicates
that it indeed needs to be included in the broader typological framework of interclausal
reference languages. In order to pursue this issue further, I will apply Givon's criteria
for "anticipatory" switch-reference languages (1983:77-78) to the Amahuaca data (his
criteria partially overlap with the ones listed above):
(a) "All languages of type (a) DS/SS system are strict SOV languages." As noted
above, Amahuaca is SOV, but not in an exclusive sense.
(b) "In all those languages, the DS/SS morphology appears as verb-suffixes, i.e.,
also clause-final." This is basically so in Amahuaca.
(c) "In all those languages only medial, non-finite clauses are involved in the
SS/DS contrast." This argument applies to the Amahuaca data.
(d) "Non-finite clauses are less likely to exhibit their own verb-inflections, such as
tense-aspect-modality, mood, speech-act or pronominal agreement." This also applies to
Amahuaca. Despite the fact that the IRMs in Amahuaca are derivatives of tense/aspect
markers, and can combine with temporal suffixes to express time lapses between events,
they do not encode tense but temporal relations. The controlling verbs encode tense.
Amahuaca 469
(e) "In all these languages, the SS marker is consistently smaller in size than the DS
marker. Often the SS marker is zero." This argument, although it does not directly
address the question of a "real" or "unreal" IRS, is a crucial one with regards to the
Amahuaca data. As Tables V and VII show, several of the SS and DS IRMs are equal
in size. We need to decide whether or not the referential relations are explicitly coded
or not.
If we dissect the sequence and non-sequence IRMs separating the tense/aspect and
case-marking elements, we note the following pattern:
TABLE 7:Tense/aspect and case-marking elements in IRMs
SO •xo -0 hai-to- -0
OS SG -ha -0 -hai -0
PL -ha -vo -0 -hai -vo -0
DS SG -cu -n -hai -n -non
PL -ha-va -n -hai-vau -n -no -n -vo
Following Haiman's model (1983:107), the Amahuaca marked verb can be sym-
bolized as follows:
In order to understand what actually triggers the SS/DS contrast I am applying some
of the origin hypotheses that have been advanced in the past by different scholars to the
Amahuaca data. In Haiman and Munro's introduction to Switch-Reference and Univer-
sal Grammar (1983:xiv) we find a list of suppositions (i-vi). Also Givon, in the same
volume (p.78), postulates a hypothesis. I will apply these in the order listed, except for
Haiman's "gapping" hypothesis. I will discuss it at the end.
i. This asserts in part: "The reduction of an entire SS clause to the point where
it is an auxiliary-like affix on the verb of the reference clause, (is) a clearly
possible outcome of such a reduction process, . . . " The IRMs discussed in
Section 3.1.1.1 confirm this possibility. They are auxiliary-like affixes, their
referential relations are coded zero, and the forms combine with DS and SO
IRMs.
ii. "SS and DS markers originate as deictics such as hither and hence." The
origin of -cun and -non 'DS,' -hax 'SS, intransitive reference verb,' and -xon
'SS, transitive controlling verb' could be explained through this hypothesis.
The forms -hax and -xon, when suffixed to adverbs, modify them, and -cu and
-no (without the case marking) are integral parts of a series of adverbs
encoding 'discontinuity' (see the discussions in Sections 3.1.1.2 and 3.1.3).
iii. "SS and DS markers originate as case marker affixes." This supposition, of
course, cannot be entirely true for Amahuaca, since the case marking is only
one part of the IRMs.
iv. See below.
v. "SS markers originate as temporal successive markers; DS markers originate
as temporal overlap markers." On the surface the Amahuaca data appear to
disagree with this supposition, since the -hai forms encode coreference as well
as non-coreference. However, a closer look reveals that -haito (SO) and -hai
(OS) focus on coreference of subject with object, but at the same time imply
non-coreference of subjects. Also -to-0/-x/-n can be considered an explicit
SO/SS indicator (see discussion in 3.1.2.2.). This can be considered as over-
riding the DS quality.25 Interpreted as such, Longacre's hypothesis, on which
the above supposition is based, does apply to the Amahuaca data.
vi. " . . . , the causative may serve as a switch-reference marking mechanism,..."
This does not apply to the Amahuaca data.
Givon's hypothesis says in essence that the SS/DS morphemes may arise
diachronically from a contrast of subject pronouns in the succeeding clause and that
this morphological contrast, either between stressed vs. unstressed pronouns or
pronouns vs. zero, respectively, became cliticized on the preceding verb. This
hypothesis could apply to the DS and SO constructions. The DS marker -cun could
have derived from nocun 'we (stressed),' or Our (unstressed).' The SO marker
-hai-to could have derived from mato 'you, plural object,' or jato 'them.'
Amahuaca 471
Haiman's hypothesis (iv) says in essence that IRMs are the result of a coordination
reduction (or gapping) process. As stated above, in Amahuaca the subject is gener-
ally overtly expressed in the controlling clause, and only in certain DS constructions
is the subject also expressed in the marked clause. Thus there is backward gapping.
Since the subject is lacking in the marked clause, it is the verb that takes the
case-marking morphology.
The strongest support for the above hypothesis are the IRMs encoding plural
subject. As we noted in examples 26 and 34, the plural-marking -van/-vaun is
anaphoric, referring back to a non-expressed plural subject. Since the subject is not
overtly expressed, it is the verb that receives the marking for case and number.
The Amahuaca data also are relevant to one of Haiman"s propositions, which
states: "The characteristic index of cohesion is not "same x" but rather "necessarily
same x" as other clause, where "x" is subject, . . . " (1983:107). We find this
proposition most clearly exemplified in the SS constructions of Set B, where -tox/-
ton mark the coreference in a pronounced way, but it can also be applied to the rest
of the coreference markers and can be extended to include the SO construction of
Set A as well.
If we return now once more to the question of authenticity, it seems to me that
Amahuaca is a "real" clause-linking language, with IRMs that explicitly mark
coreference of subjects (SS), coreference of subjects with objects (SO, OS), and
non-coreference of subjects (DS). The IRMs, although analyzable, have become
frozen forms which do double duty in that they keep track of temporal/logical
relations as well as of referential relations.
In this paper so far I have described and illustrated the forms and usage of the IRS
in Amahuaca on sentence level. In the following section I briefly examine the effect
of the system on discourse level, focusing on participant orientation.
The extensive use of IRMs results in a highly complex cohesive system. The
complexity is due to the fact that the IRMs encode not only participant reference
(SS, DS, OS and SO) and temporal/logical relations of events, which can include
time spans between events, but also different kinds of events (transitive, intransitive,
and movement towards or away from a focal point). Consequently there are no
conjunctions, and the use of adverbs, pronouns and other connectors is kept to a
minimum.
The text to be analyzed is a mythological tale which has been recorded and
transcribed. The editing process has been limited to incorporating only those changes
a native speaker felt necessary to insure clear understanding of the story.
eyes (with her hands), (i) and she stood (j) and while she kept looking,
(k) the niece came from behind toward her (1) without causing her to
see, (m) and pushed her into the fire.'
(5) a. Chihi -qui j'an vuran -ha
fire -into she push -(SQ)OS
b. totocototoco -cax26 -mun -quiha
sizzle -(SQ)SS.ITR -TH -REPORT
c. jan rapashcara -cahtox26
she turn.into.sparks -(SQ)SS.ITR
d. jaa -razix xau -nix -quiha
REL -all.of.the.kind become.turtle -REM.PAST -REPORT
-qui -hnu.
-3ACT -DECL
'(a) She (the niece) having pushed her into the fire, (b) they say that she
(the old woman) sizzled (c) and having turned into sparks (d) the rest of
her turned into turtles, (they say).'
(6) Jan nato -x -mun -quiha haa -ni -xo
her navel -NOM -TH -REPORT turn.into.tapir -REM -3PAST
-hqui.
-DECL
'It is her navel, they say, that turned into a tapir.'
5.2 Analysis. The entire narrative consists of six sentences. The first and last
sentences are composed of one independent clause each. The rest have several
marked clauses and one final controlling clause. The longest sentence (4) consists of
eleven dependent clauses, and spans the major part of the narrative.
The text is best divided into four parts:2^
Part I, Exposition, sentence (1)
Part II, Inciting Moment, sentences (2) and (3)
Part III, Developing Conflict and Climax, sentence (4)
Part IV, Conclusion, sentences (5) and (6).
The exposition contains the major events in a nutshell: it introduces the partici-
pants (the old woman, the niece, and the child) and foretells what is about to happen.
The inciting moment includes the setting and the initial episode. It opens the
conflict between the hero and the villain. The niece, who fills the role of hero in the
narrative, asks the old woman, the villain, to nurse the child. The old woman
Amahuaca 475
pretending to do as requested, kills and eats the child. At the outset of the inciting
moment the two major participants interact. The hero then leaves the scene, allowing
the villainy to take place. The crucial episode, killing and eating the child, after
being described in detail, is summed up by the narrator in a brief statement (3), thus
underlining the fact that the world has been thrown into chaos.
At the outset of Part III, the developing conflict and climax, the narrator refers
back to the episode that caused the chaos, thus linking it directly to what is about to
happen in a reason-result relationship. This is the turning point in the story, the
villain becomes the victim, and the hero assumes her heroic role. By avenging the
villain's crime, the hero re-establishes world order.
Part IV, the conclusion, deals with one participant only. The villain, having been
reduced to ashes, turns into animals.
5.3 Participant Orientation. The old woman is the main participant. In the expo-
sition she fills the subject slot, and she is present throughout the story. The narrator
refers to the other major participant from the old woman's point of view: hain-vacu
'niece' means literally 'the daughter of a woman's sister.' Outside of the exposition
the main participant is referred to twice as 'mother' (huha) in direct address (2) and
(3) (this is the proper address to mother's sister in Amahuaca culture), and once as
Old woman' (xunivaun) (3). In the conclusion, reference is made to her three times
through pronouns. The remaining eleven references are carried by IRMs.
The niece, after having been introduced, is referred to twice as 'niece' (hain-vacu)
(3) (4), twice in direct speech she refers to herself with the first person pronoun hiya
'for me' (2) and (4), and three times she is referred to by IRMs.
The minor participant, the child, is introduced as the niece's child (Jan vacu xano),
and is referred to as 'child' (vacu) three times thereafter (twice in 2) and (3).
Three major principles appear to underly participant orientation in this narrative:
1. participants are explicitly introduced at the beginning of the narrative; 2. the more
important the participant, the less explicit is the reference to him (cf. Givon
1983:67); 3. strict economy is exercised in the usage of nouns and pronouns.
The most striking example of principle 3 is sentence (4): the old woman is in the
foreground and a new episode begins with a switch in roles: she, the villain-hero,
becomes the victim, and the victim (the niece who has been bereft of her children)
becomes the hero. However, the switch in roles is signaled only by the IRM -haito
'SO' which, of course, besides expressing that the actor (the old woman) becomes
the patient, also assumes a new actor (the niece). But not until ten clauses later is
the new actor referred to explicitly.
In the context of the story the IRM -haito 'non-sequential, subject is coreferential with
object' in the marked clause pihanpihan-haito-munhquiha 'while/because she kept eating
to the detriment of her niece' achieves two things on two levels: on the syntactic level
it gives the reason for the final event, and on the discourse level it predicts the reversal
of the roles of the two major participants. The speaker employed this IRM in a most
strategic way: he suffixed it to the first verb phrase in the paragraph that contains the
476 Sparing-Chavez
climax of the story. The function of the entire verb-phrase, which is a repetition of the
last event of the previous paragraph, is to introduce a new theme (the destruction of the
villain), and the function of the IRM in particular is to predict the role reversal of the
participants. The speaker could have employed -cun(DS) instead of -haito(SO\ because
there is a switch in subjects between lines 4a and b. However, this would not have
achieved the same effect. It appears to me that the marked clause has received special
prominence through the IRM -haito, which does not refer to the adjacent clause but to
the final independent clause. In 4j the IRM -haito is used again, but here in the standard
way referring to the clause on its immediate right. (In the sixty pages of oral narrative
text I have only seen the IRM -haito used in this way. Further investigation in this area
is necessary.)
Besides effecting economy, the IRM strategy achieves referential continuity. The
old woman remains present throughout the whole episode despite the fact that
semantically she changes roles from actor to patient.
This short narrative shows the tremendous potential inherent in the IRS in Ama-
huaca. The system, although complicated on the surface, is compact, systematic and
logical.
Acknowledgements
Notes
Abbreviations
1 first person
2 second person
3 third person
ACN action
ACT actuality aspect
ABS absolutive
BENEF benefactive
CONTR.FACT contra-factual
COM commitment
DECL declarative
DIST distant
DS different subject
ERG ergative
PUT future
GEN genitive
HAB habitual
INF infinitive
IMM immediate
IMP imperative
IRM interclausal reference morpheme
IRS interclausal reference system
ITR intransitive
LOG locative
MALF malefactive
NEG negative
NOM nominative
NONSQ non-sequential
OS object is coreferential with subject
PF perfect aspect
PFTV perfective aspect
PL plural
PRES present tense
PRET pretence
PROSP prospective aspect
QUEST question
REM remote
REP repetitive action
REFL reflexive
REL.CL relative clause
REC recent
SG singular
Amahuaca 481
References
Chafe, Wallace L.
1976 "Givenness, contrastiveness, definiteness, subjects, topics, and point of
view," Subject and topic, edited by Charles N. Li (New York: Aca-
demic Press), 25-55.
Cole, Peter
1983 "Switch-reference in two Quechua languages," in John Haiman and
Pamela Munro (eds.), 1—15.
Comrie, Bernard
1976 Aspect (Cambridge: Cambridge University Press).
1978 "Ergativity," Syntactic typology, edited by Winfred P. Lehmann
(Austin: University of Texas Press), 329-94.
1983 "Switch-reference in Huichol: a typological study," in John Haiman
and Pamela Munro (eds.), 17-37.
1985 Tense (Cambridge: Cambridge University Press).
Cromack, Robert E.
1968 Language systems and discourse structure in Cashinawa, Hartford Stud-
ies in Linguistics, No.23 (Hartford, Conn: Hartford Seminary
Foundation).
DeLancey, Scott
1980 An interpretation of split ergativity and related patterns (Bloomington:
Indiana University Linguistics Club).
DuBois, John W.
1987 "The discourse basis for ergativity," Language 63:805-55.
Faust, Norma
1973 Lecciones para el aprendizaje del idioma shipibo-conibo, Documento
de Trabajo No. 1 (Yarinacocha, Pucallpa, Peru: Institute Lingüistico de
Verano).
1985 "Gramätica yaminahua," unpublished ms. (Peru: Instituto Lingüistico
de Verano)
Finer, Daniel
1985 "The syntax of switch reference," Linguistic Inquiry 16:35-55.
Franklin, Karl
1983 "Some features of interclausal reference in Kewa," in John Haiman
and Pamela Munro (eds.), 29-49.
Amahuaca 483
Givon, T.
1982 "Transitivity, topicality, and Ute impersonal passive," Syntax and se-
mantics, Vol.15, edited by Paul Hopper and Sandra Thompson (New
York: Academic Press), 43-60.
1983 "Topic continuity in discourse: the functional domain of switch-refer-
ence," in John Haiman and Pamela Munro (eds.), 51-82.
Haiman, John
1983 "On some origins of switch-reference marking," in John Haiman and
Pamela Munro (eds.), 105-28.
Haiman, John, and Pamela Munro
1983 "Introduction," in John Haiman and Pamela Munro (eds.), ix-xv.
Haiman, John, and Pamela Munro (eds.)
1983 Switch-reference and universal grammar (Amsterdam: John Ben-
jamins).
Jacobson, William H.
1967 "Switch-reference in Hokan Coahuiltecan," Studies in South-western
ethnolinguistics, edited by Dell Hymes (The Hague: Mouton), 238—63.
Kneeland, Hattie
1979 Lecciones para el aprendizaje del idioma mayoruna, Documento de
Trabajo No. 14 (Yarinacocha, Pucallpa, Peru: Instituto Lingüistico de
Verano).
Lauriault, James
1957 "Some problems in translating paragraphs idiomatically," The Bible
Translator 8:166-69.
Longacre, Robert
1976 An anatomy of speech notions (Lisse: The Peter de Ridder Press).
1983 "Switch-reference systems in two distinct linguistic areas: Wojokeso
(Papua New Guinea) and Guanano (Northern South America)," in John
Haiman and Pamela Munro (eds.), 185-207.
Loos, Eugene
1963 "Capanahua narration structure," Studies in literature and language
Vol.IV, Supplement, edited by George Monteiro (Austin: University of
Texas Press), 697-742.
1967 The phonology of Capanahua and its grammatical basis (University of
Texas: Ph.D. dissertation).
Loriot, James, and Barbara Hollenbach
1970 "Shipibo paragraph structure," Foundations of Language 6:43-66.
484 Sparing-Chavez
Montag, Richard
1981 "Notas gramaticales," Diccionario cashinahua, Tomo II, compiled by
Susan Montag, Serie Lingüfstica Peruana, No.9 (Yarinacocha, Pucallpa,
Peru: Instituto Lingüistico de Verano), 551-607.
Munro, Pamela
1980 "Introduction," in Pamela Munro (ed.), 1-3
Munro, Pamela (ed.)
1980 Studies of switch-reference (University of California, Los Angeles,
Papers in Syntax 8).
Oswalt, Robert L.
1983 "Interclausal reference in Kashaya," in John Haiman and Pamela
Munro (eds.) 267-90.
Payne, Thomas
1980 "Who's kissing who: the fourth person in Yup'ik Eskimo," in Pamela
Munro (ed.), 65-68.
1990 "Transitivity and ergativity in Panare," Amazonian linguistics: Studies
in lowland South American languages, edited by Doris L. Payne
(Austin: University of Texas Press), 429-53.
Roberts, John
1988 "Amele switch-reference and the theory of grammar," Linguistic In-
quiry 19:45-63.
Russell, Robert
1965 A transformational grammar of Amahuaca (Pano) (Ohio State Univer-
sity: Master's thesis).
1975 Una gramatica transformacional del amahuaca, Estudos Panos IV
(Yarinacocha, Pucallpa, Peru: Instituto Lingüistico de Verano).
Shell, Olive
1957 "Cashibo II: Grammemic analysis of transitive and intransitive verb
patterns," International Journal of American Linguistics 23:179-218.
Shell, Olive, and Mary Ruth Wise
1971 Grupos idiomäticos del Peru (Lima: Universidad Nacional Mayor de
San Marcos and Instituto Lingüistico de Verano).
Woodbury, Anthony
1983 "Switch-reference, syntactic organization, and rhetorical structure in
Central Yup'ik Eskimo," in John Haiman and Pamela Munro (eds.),
291-315.
PART III
Comparative Study: Tupi-Guarani
Comparative Tupi-Guaram Morphosyntax
Cheryl Jensen
Summer Institute of Linguistics
0 Introduction
1 Proto-Tupi-Guarani person markers
2 Word classes
2.1 Subclasses
2.2 The linking prefix r-
2.3 Allomorphs of r-
2.4 Other interpretations of r-
3 Nouns
3.1 Possession
3.2 Co referential possessive markers
3.3 Grammatical suffixes
3.4 Modificational suffixes
3.5 Noun composition
3.6 Noun phrases
4 Postpositions
4.1 Normal markers
4.2 Coreferential markers
5 Cross-referencing on independent verbs
5.1 Cross-referencing on independent verbs: A and S markers
5.2 Cross-referencing on independent verbs: P markers
5.3 Cross-referencing on independent verbs: Portmanteau A-P prefixes
5.4 Cross-referencing on independent intransitive non-agentive verbs
5.5 Imperatives
6 Cross-referencing of dependent verb forms
6.1 Oblique-topicalized verbs
6.2 Subordinate verbs
6.3 Dependent serial verbs
7 Valence-changing devices
7.1 Causatives
7.2 Detransitivizers
8 Verbal aspect
8.1 Future, desiderative, and completion
8.2 Monosyllabic reduplication
490 Jensen
0 Introduction
The Tupi-Guarani language family is one of seven families (plus 3 isolated lan-
guages) in the Tupian stock, located in lowland South America. Whereas the peoples
of the other Tupian language families have a history of being somewhat settled
(being confined to the Brazilian states of Rondonia, Amazonia, and western Para),
the Tupi-Guarani peoples have been highly migratory. Members of the Tupi-Guarani
language family are located in 5 countries of South America: Brazil, Paraguay,
Bolivia, Peru, and French Guiana.1 Even so, they show a high degree of consistency
in their basic morphology. Compare, for example, three languages from the geo-
graphical extremities of the language family: Mbyä Guarani (southern Brazil),
Wayampi (Jari dialect spoken in northern Brazil), and Guarayu (Bolivia), together
with the reconstructed Proto-Tupi-Guarani forms:2
P-T-G GiM WaJ Gu
corn *abati avaxi awäsi aväci
jaguar *jawär jawa jäwa jawar
sun *kwaracy kwaray kwaräy äry
monkey *ka'i ka'i kä'i kä'i
rock *ita ita takuru ita
The differentiation in these descendant languages is the result of (1) neutralization of *b
and *w (WaJ); (2) loss of final consonant (GiM,Wa); (3) stress change from final to
penultimate syllable (Gu,Wa); (4) loss of *c (GiM,Wa); (5) palatalization and weakening
of *t preceding a high front vowel (Gu,GiM,Wa); (6) modification of stem (Wa); (7)
loss of cognate (Gu). These lexical items are easily recognizable as cognates in spite of
the great geographical distance separating the three languages.
The main objective of this paper is to give an overview of the morphology and, to
a lesser degree, the syntax of the Tupi-Guarani language family. Both phonological
(Lemle 1971) and morphological (C. Jensen 1989) reconstructions of Proto-Tupi-
Guarani have been published and provide a good framework for a comparison of the
descendant languages. In this paper I present the grammatical structures as they were
reconstructed for Proto-Tupi-Guarani, followed by examples of these structures in
various descendant languages. I then describe changes that have taken place in other
languages which deviate from the reconstructed forms. Since phonological change has
a major influence on morphological and syntactic changes, I have included Appendices
II to IV to serve as references for the reader. Reconstructed morphemes with supporting
data appear in Appendix I. Rodrigues' (1984/85) criteria for distinguishing Tupi-Guarani
languages by subgroups appear in Appendix V.
Since my principal interest is in the area of reconstruction and change, I have
concentrated my studies on those Tupi-Guarani languages which to a large degree
still represent the Proto-Tupi-Guarani system. A few languages in the family show
alternation significant enough to justify the hypothesis that they are spoken by
492 Jensen
IWayampi
Tupinambä
(extinct)
ECUADOR
Urubu-
Kaapor
Parin tintin _ . .
Guajajara
r Guarayu ;
.
-Paraguay afr,
·—jGuaranf ^i
Old Guaram
(extinct)
Comparative Tupf-Guarani Moiphosyntax 493
0.1 Overview
1. There are four sets of person markers (sect. 1) reconstructed for Proto-Tupi-
Guarani which occur with verbs, nouns, and postpositions. Set 1 markers (sect. 5.1)
refer to the grammatical subject and occur in the independent clause with agentive
intransitive verbs and with transitive verbs when the object is third person. Set 2
markers refer to the object of transitive verbs (sect. 5.2), the subject of intransitive
verbs (agentive and non-agentive) (sects. 5.4; 6.1-3), the possessor of nouns (sect.
3.1), and the object of postpositions (sect. 4.1) when these are not coreferential with
the subject of the independent verb. They are used in dependent verb constructions
(sects. 6.1-3; 9.1-2) as well as with independent transitive verbs when the object is
hierarchically superior to the subject. Set 3 markers refer to the subject of dependent
serial intransitive verbs (agentive and non-agentive) (sect. 6.3), the possessor of
nouns (sect. 3.2) and the object of postpositions (sect. 4.2) when these are corefer-
ential with the subject of the independent verb. Set 4 markers (sect. 5.3) occur with
independent transitive verbs when the subject is first person and the object is second
person.
2. There are two classes of inflectible stems, based on their combination or not with
a linking morpheme r- (sect. 2). There is an additional class of noninflectible stems.
3. Many nouns are obligatorily possessed, among them body parts (sect. 3.1).
4. Proto-Tupi-Guarani nouns are suffixed by a nominal case marker whenever
they function syntactically as nouns (sect. 3.3). Unsuffixed, a noun may also serve
syntactically as a verb.
5. Various modificational suffixes occur with nouns, including augmentative and
diminutive, devolved and anticipatory, genuine and imitative suffixes (sect. 3.4).
6. Complex nouns are formed by combinations of Noun-Noun and Noun-Verb
(sect. 3.5).
7. Noun phrases are fairly simple, and consist of three types: Genitive-Noun,
Noun-Appositive, Specifier-Noun (sect. 3.6).
8. The object of a postposition is indicated by a noun, a person marker of Set 2,
or a person marker of Set 3, the latter indicating coreferentiality with the subject of
the independent verb (sect. 4).
9. When the object of an independent transitive verb is third person, the verb is
prefixed with Set 1(A) person markers, which are also used to prefix agentive
intransitive verbs. When the object of the independent transitive verb is hierarchi-
cally superior to the subject, the verb is cross-referenced with Set 2(P) person
markers, which are also used to cross-reference non-agentive intransitive (that is,
494 Jemen
Stative) verbs. In the remaining cases (where the subject is first person and the object
is second person) the independent transitive verb is cross-referenced by Set 4 person
markers (sect. 5).
10. In languages which retain the system reconstructed for Proto-Tupi-Guarani, all
dependent verb forms are cross-referenced with person markers from Set 2, unless
their referent is coreferential with the subject of the independent verb. In this case
they take markers from Set 3 (sect. 6).
11. Dependent verb forms include oblique-topicalized, subordinate, and dependent
serial verbs (sect. 6).
12. Valence-changing devices include three types of causative and four techniques
for detransitivizing a transitive verb (sect. 7).
13. Verbal aspects include future, desiderative, completive, consecutive, frequen-
tive, frustrative, and intensive (sect. 8).
14. Nominalizations refer to the action itself, or to the agent, circumstances, or
patient of the action. These may co-occur with anticipatory and devolved mor-
phemes (sect. 9).
15. Five separate negation markers occur (sect. 10).
16. Demonstratives function with reference to persons, objects, time or location,
or to temporal, spatial, and other elements of the discourse (sect. 11).
17. Tupi-Guarani languages are characterized by particles (sect. 12) and ideo-
phones (sect. 13).
18. The numerical system is rudimentary (sect. 14).
19. Constituent order in absolutive-marked constructions tends to be verb-final.
Independent clauses show somewhat free word order, but basic orders tend to be
SOV or SVO (sect. 15).
20. Discussions of the significance of the linking prefix r- and of the development
of the person hierarchy appear in sects. 16 and 17, respectively.
0.2 Classification. The total number of Tupi-Guarani languages exceeds 40. Sev-
eral of these languages have been extensively documented, including two extinct
languages: Old Guarani (documented principally by Antonio Ruiz de Montoya in
1639 and 1640) and Tupinambä (documented by Joseph de Anchieta in 1595). For
certain other languages, only limited data are available. The Tupi-Guarani languages
are tentatively subdivided by Rodrigues (1984/85) into eight subgroups, which
appear in Table 1. Priest's (1987) reference to the Guarani linguistic family includes
Bolivian languages from Rodrigues' subgroups I and II.
Comparative Tupf-Guaranf Morphosyntax 495
disagrees with Rodrigues' subgroupings. I do not think that they are necessarily
inconsistent, since languages can be closely related historically, that is, members of the
same subgroup, as Rodrigues proposes, and still be quite different from each other,
because of subsequent individual changes, particularly if outside influences are in-
volved.
One limitation of this type of study, as Dietrich mentions himself, is its depend-
ence on sufficient data for each feature included. Some features listed as nonexistent
in the Jari dialect of Wayampi, for example, do in fact occur (oka, from *ukar, and
a 'e). Another feature suggested as regional, since it occurs in Wayampi and Emeril-
lon, is the negation morpheme rowä (or ruä, as we now spell it), which actually has
cognates in Tupinambä, Old Guarani, Parintintin, and Kamaiura (sect. 10.3). Insuf-
ficiency of data or inaccuracy in the reporting of features affects the numerical
outcome in the comparison of languages.
The choice of features and the value assigned to each one also affect the numerical
outcome in a comparison. I suggest that not all features should carry equal weight,
and that differences have their greatest significance when they cause misunderstand-
ing or misinterpretation from one language to another.5
As linguists studying the various Tupi-Guarani languages cooperate with Dietrich by
evaluating his results in relation to their particular language, by providing further input
into the significance of specific phonological and morphological features as a basis for
comparison, and by supplying missing data, he will be able to refine his system of
evaluation. This will increase its effectiveness in showing how closely the individual
Tupi-Guarani languages are related to each other. I suggest also that he add to his
phonological and morphological comparisons a measure of lexical comparison.
TV TV TV TV
IV-ag IV IV
N N
Post Post
(Note: The r- and n- occur obligatorily between the Set 2 person markers and
morphemes of Class II, as indicated.)
The person markers from Sets 1 and 2, as well as the free pronouns, occur in all of
the descendant languages on which the reconstruction was based. Set 1 markers occur
exclusively with transitive and agentive intransitive verbs; they refer to A and S,
respectively (see footnote 10 for abbreviations). Set 2 markers occur with transitive and
intransitive (agentive and non-agentive) verbs, nouns, and postpositions; they refer to
P and S. Set 3 markers occur in some languages with nouns, postpositions, and
dependent verbs; in other languages the full set occurs only with intransitive serial
verbs, only the third person prefix occurring with nouns, postpositions, and subordinate
verbs. (In still other languages, the set has been eliminated completely, with the
exception of the third person prefix o-.) They are agreement (coreferential) prefixes which
replace Set 2 markers when the referent (P or S) is also the grammatical subject of the
independent verb (or the verb which dominates a noun or postposition in a subordinate clause).
Set 4 are portmanteau markers referring simultaneously to first person A and second person P
in transitive verbs. The function of these sets of person markers will be discussed in greater
detail under noun and verb morphology (sects. 3-6).
2 Word classes
Basic to the description of Tupi-Guarani languages is the division of all inflectible
stems into two word classes for which there is no semantic or clearly defined
grammatical basis. Stems which are grammatically noninflectible, that is, obligato-
rily non-possessed, comprise a third class. Morphemes are classified as Class II if they
combine with a special linking morpheme r- (sect. 2.2). Class I morphemes never take the
r- prefix. These classes transect grammatical divisions of noun, verb, and postpositions, as
Comparative Tupf-Guaranf Morphosyntax 499
illustrated in Table 3. All members of Class II begin with a vowel; most, but not all,
members of Class I begin with a consonant.
Table 3: Class I and Class Ð stems
Class I (without r-) Class II (with r-)
*akarj 'head'
*ce akarj'my head'
*kununii akarj 'boy's head'
*apyti 'tie'
*ce apyti 'tie me'
*kunumT apyti 'tie the boy'
*u'u 'cough'
^v s y s ßô é *
*ce u u I cough
2.1 Subclasses. Word classes I and II both have subdivisions (Table 4) based on
the combination of its members with certain allomorphs of the third person and the
unspecified possessor prefixes. The unspecified possessor morpheme is used in cases
where the possessor or referent is grammatically obligatory, but out of focus. In the
following examples from Wayampi, 'medicine' is obligatorily possessed, either by
the person for whom it is prescribed (1) or by the ailment which it treats (2).
500 Jensen
la *i- - *0
Ib *i- - ""nasalization of initial C
Ila *c- *r- *t-
lib *t- *r- *t-
He *c- *r- *0
lid *c- *r- *elimination of initial V
III Cannot be inflected
Note that in subclass lib the third person and the unspecified forms are identical.
These subclasses apply to both noun and verb stems. However, the combination of
these prefixes (third person, linking and unspecified possessor) with verb stems is
restricted to specific syntactic structures which are described in sects. 5 and 6. The
examples in Table 5 are given without reference to such syntactic structures.
Table 5: Subclasses with prefixing, Proto-Tupi-Guarani
III non-inflectible:
*aman 'rain'
Although in most languages the members of subclass Ib all begin with /p/, there are
examples in Wayampi in which the rule of nasalization is applied to stems beginning
with/k/:
Sometimes a word which is in subclass lid in one language is in subclass Ila in another,
so that instead of eliminating the initial vowel, a prefix t- precedes it. For example, *epoti
'defecate' has the unspecified form poti in Tupinamba, but teposi in Wayampi. In Tupi-
narnba *emi'u 'food' has two unspecified forms: temi'ü and mi'ü. In Wayampi it has only
the form temi'ö. In Parintintin the form is mbi'u? It appears that in Mbyä Guarani all
members of what was originally subclass lid have been transferred to subclass Ila. For
example, tape 'path (unspecified)', tepoxi 'feces', tembi'u 'food'.
2.2 The linking prefix r-. The linking morpheme, which co-occurs with stems of
Class II, occurs under the following circumstances: 1) Noun preceded by the genitive,
indicated by a noun or by a first or second person marker of Set 2 (sect. 3.1). 2)
Postposition preceded by its object, indicated by a noun or by a first or second person
marker of Set 2 (sect. 4). 3) Verb preceded by a noun or by a first or second person
marker of Set 2, indicating absolutive cross-referencing (sects. 5.1—4). The linking
prefix does not co-occur with person marker sets 1, 3, or 4 (sects. 3 and 5), nor does
it co-occur with the third person prefixes of Set 2 (sect. 3.1): */- or *c- (which occurs
in various languages as s-, h-, or J0).
2.3 Allomorphs of r-. The allomorphs of the linking prefix are r- and n-, the latter
following the second person plural marker pe in many languages (Tocantins Assurini,
Guajajära, Kamaiurä, Parintintin, Tapirape, Wayampi). Considering that pe was de-
rived from a longer nasalized free pronoun *pe . . . e it is not strange that r- would have
502 Jensen
a nasalized allomorph. This allomorph also occurs in Guajajara with a third person
plural morpheme, \va, which Harrison (1986) states to be historically nasalized as
well. Tupinamba does not have the n- allomorph. Mbyä Guarani has replaced pe with
pene and uses the r- allomorph.
3 Nouns
3.1 Possession. Nouns in Tupi-Guaram languages are either possessible (for exam-
ple, body parts) or non-possessible (for example, 'rain'). If they are possessible, they
are members of Class I or II. The linking prefix r- occurs with Class II when the
noun is preceded by a genitive, expressed either by a person marker (Set 2) of first
or second person or by a noun. This is illustrated in Table 6 with the morphemes
10
*cy 'mother' (Class I) and *ub 'father' (Class II).'"
Table 6: Possessive marking, by stem class
Class I Class II
%lt v s
v s
*ce cy 'my mother' *ce r-ub 'my father'
*ore cy Our EX mother' *ore r-ub Our EX father'
*jane cy Our IN mother' *jane r-ub Our IN father'
ak s v s
*ne cy 'your SG mother' *ne r-ub 'your SG father'
*pe cy 'your PL mother' *pe n-ub 'your PL father'
*kuja cy 'woman's mother' *kuja r-ub 'woman's father'
The first person singular morpheme is expressed in various languages, due to
weakening processes, as s/e (Tb), he (Gj), e (Wa), and ye (Km).12
(6) sje sy sje r-ub (Tb)
(7) he-hy he-r-u (Gj)
(8) e-y e-r-u (Wa)
(9) jey je r-up (Km)
Unlike the first and second person markers, the third person prefix (*/'-, *c-, */-)
attaches directly to the stem, with no linking prefix involved, as in (10).
(10) *i-cy 'his/her mother' *t-ub 'his/her father'
If a noun is obligatorily possessible, reference must be made to this fact even when
the possessor is out of focus. In this case nouns take a special form (sect. 2.1).
Nouns which are grammatically nonpossessible are automatically members of Class
III. These are largely elements of the physical world, such as sun, sky, rain, jungle,
and animals. If these objects need to be referred to in a possessed form, a more
generic possessible morpheme is used, sometimes in apposition with the nonposses-
sible noun, as in (11) and (12).
(11) s-emiar-a tatu (Tb)
3-hunted.thing-NC armadillo
'his hunted thing, the armadillo'
(12) mamaz r-eimaw zapukaz (GjP)
mother LK-pet chicken
'mother's pet, a chicken'
504 Jensen
3.4 Modificational suffixes. There are at least three pairs of modificational suf-
fixes: (1) augmentative and diminutive, (2) devolved and anticipatory, and (3)
genuine and imitative.
(1) Augmentative and diminutive suffixes:
Augmentative *-wacu ~ *-ucu (51-62) and diminutive *-'/ (63- 67) morphemes
occur as suffixes on nouns. The augmentative morpheme has two allomorphs,
occurring after V-final and C-final stems, respectively.
(51) *parana-wacu 'large large-river'
(52) *jararak-ucu 'large jararaca snake (species name)'
Comparative Tupf-Guaranf Morphosyntax 509
3.5 Noun composition. There are various patterns for complex noun formation:
(1) *N-N in which the first noun modifies the second.
512 Jensen
This pattern is similar to the pattern for genitives in noun phrases (sects. 3.1, 3.6),
but certain features indicate that the combination is a compound word. For example,
when the second stem is a member of Class II, the r- linking morpheme attaches to
it in noun phrases but not in compound words.
(91) *akarj 'head' + *acy 'pain' > *akarjacy 'headache'
(92) *wyra 'bird' + *a 'down' > wyraa 'bird down' (Wa)
(93) *ka'i 'monkey' + *a'yr 'child' > ka'i-a'yr 'young monkey' (Lit.:
'monkey child') (Gj)
If (91—93) were separate words instead of compound words, they would be *akay-a
r-acy, \vyra ra, and ka ra 'yr. The nominal case suffix occurs with *akay when it is a
separate word, but not when it is the first part of a compound word.
Stems which combine to form a complex word are subject to the application of
morphophonemic rules (Appendix III. 1,2,5,6).
(94) men 'husband' + sy 'mother' > meny 'husband's mother, mother-
in-law' (Tb)
If (94) were a phrase, it would be men-a sy.
(2) *N-N in which the second noun indicates some distinguishing feature of the
first noun.
(95) *pira 'fish' + *aj 'tooth' > *piraj 'fish known for its teeth (piranha)'
(96) *arar 'macaw' + *akarj 'head' > *ararakarj 'macaw known for its head
(Red-and-green Macaw)'
(97) *pira + *aj > pirai 'piranha' (WaJ)
(98) *aba 'person' + *oba 'face' > abäobä 'person who has an unusual
face' (Tb)
(99) *arar + *akarj > *ararakarj 'macaw known for its head'
(3) *N-V in which the verb is a stative or intransitive verb. In composition the
verb acts as an adjective modifying the noun.
(100) *wyra 'bird' 4- *picun 'black' > *wyrapicun 'species of black bird'
(101) wainumby 'hummingbird' + pihun 'black' > wainumby pihun
'black hummingbird (Black -throated Mango)' (Ur)
(102) pirä 'fish' + bebe '(to) fly' > piräbebe 'flying fish' (Tb)
(103) from Wayampi is a combination of the second and third types of complex
noun formation: N + (N + V).
(103) pekö 'woodpecker' + akäg 'head'+ pirä 'red' > peköakämirä
'red-headed woodpecker (Yellow-throated Woodpecker)' (WaA)
Comparative Tupf-Guaranf Morphosyntax 513
3.6 Noun phrases. Noun phrases are quite simple in structure, since much of the
modificational load is handled on the morphological rather than the syntactic level.
Three types of noun phrases have been described for Tupi-Guarani languages:
(1) Genitive Noun:
(104) *i-cy akarj 'his mother's head'
4 Postpositions
Several postpositions have been reconstructed for Proto-Tupi- Guarani (Table 8):
Table 8: Postpositions, by stem class
Class I Class II
Table 12: Independent transitive verbs with third person P, with polysyllabic stems
A-P Class I Class II
'like' 'see'
Special allomorphs of the third person P prefix, *jo- (Class I) and *joc- (Class II),
occur with monosyllabic stems, as in Table 13.
Table 13: Independent transitive verbs with third person P, with monosyllabic stems
A ^ P C l a s sI Class II
'plane' 'wash'
The inclusion of the P prefix was obligatory in Proto-Tupi-Guarani, as it is in
various descendant languages. The third person P prefix occurs in transitive verbs in
Mbyä Guarani, Kaiwä, Chiriguano, Guarayu, and Tupinambä.
(121) *ja-c-ekar > ja-s-eka 'we seek it' (Gu)
(122) *ere-i-potar > ere-i-pota 'you like it' (GiM)
(123) *o-joc-ej > o-jos-ej 'he washed it' (Tb)
In both Mbyä and Paraguayan Guarani, the allomorph for Class II stems *c- has a
zero morpheme as a result of phonological changes (Appendix IV). As a result, the P
prefix only appears with Class I verbs. In Paraguayan Guarani (Payne 1994), this prefix
has become associated with the A prefix (124) and occurs even in cases where the P
prefix did not originally occur, that is, with Set 4 markers (example 125).
(124) *ere-i-nupä > rej-nupä 'you SG hit him' (GiP)
(125) *oro-nupa > roj-nupä hit you'
The P prefix has been eliminated in Wayampi, Urubu-Kaapor, Guajajära, Assurini,
Tapirape, Kamaiura, Parintintin, and Kayabi. Its elimination does not appear to have
Comparative Tupf-Guaranf Morphosyntax 519
any connection with other cross-referencing changes (since these have occurred in
the Guaranian languages and Wayampi and Urubu-Kaapor).
(126) *a-i-potar > a-putar º want it' (Gj)
(127) *a-i-potar > a-pota º want it' (Wa)
(128) *o-i-me'erj > o-me'erj 'he gave it' (Km)
In some languages in which the P prefix was deleted, remnants still appear in certain
words (C. Jensen 1987). In (129) and (130) from Guajajara and Urubu-Kaapor, the
P prefix was reassociated with the stem, producing consonant-initial stems. In (131)
and (132) from Wayampi, the attachment of the prefix to the stem produced bisyl-
labic stems; monosyllabic stems are somewhat undesirable in this language.
(129) *a-i-apo > a-j-apo > a-z-apo > a-zapo º make it' (Gj)
(130) *a-c-enub > a-hendu º hear it' (Ur)
(131) *a-jo-pin > a-ïñÀ º plane it' (Wa)
(132) *a-jo-'ok > a-jo'o º dig it up' (Wa)
Not all descendant languages of Proto-Tupi-Guarani use the first person inclusive
prefix *ja- with transitive verbs. Five languages use a prefix which could be recon-
structed as *ti-: ti- (Pt), ci- (Tp), si- (Kb, Gj, Wa). This has been shown to be derived
from a combination of *ja- and two other morphemes: *?- 'purpose' + *ja- 'UN.A'
+ */- '3P' (C. Jensen 1987). All of the languages in which this prefix occurs are
ones in which the Ñ prefix no longer co-occurs with the A prefix.
(133) ti-apo 'we IN made it' (Pt)
(134) si-juka 'we killed it' (Wa)
(135) si-eru 'we smelled it' (Kb)
Although it does not appear that this fused prefix (*£/-) should be reconstructed for
the entire Tupi-Guarani protolanguage, we must not overlook the fact that Aweti (a
closely related Tupian language, though not Tupi-Guaranian) also has a first person
inclusive prefix ti- for transitive verbs (Monserrat 1976). It has a separate prefix,
kaj-, for intransitive verbs.
A parallel morpheme for intransitive verbs, sa-, occurs in Kayabi as the independent
intransitive verb prefix. It also occurs in the Jari dialect of Wayampi in purpose
constructions. The latter provides strong evidence for its derivation from *t- + *ja-.
Rodrigues (1990) describes occurrences of the o- and Ja· prefixes with transitive
verbs in Tupinamb , in which the basic distinction is focus rather than person. Some
occurrences of o-, as in (139), are best translated as º, 2, and 3 in focus'. Some
occurrences of ja-, as in (137), can be translated as '3, out of focus, or generic'.
(136) pir ja-i-pysyk16
fish lIN-3-catch
'We caught fish.'
520 Jensen
by analogy after reinterpreting xepe as referring to P (xe-). At this point in time, pexepe
(157) is still used when the A is second person plural. This indicates that the free morphemes
have not gone all the way in referring primarily to P.
Kaiwa and Tapirape have similar forms. It appears that the po was at some point
reinterpreted as a P prefix, to which was added a prefix from Set 1, by analogy to
the forms which contained third person prefixes, as in Table 16.
Table 16: Analogical reinterpretation of forms with po in Kaiw and Tapirapi
PATIENT
1SG 1EX UN 2SG 2PL 3
AGENT
1SG oro- opo- a-i-
a-c-
1EX oro- opo- oro-i-
oro-c-
UN ja-i-
ja-c-
SET4
2SG ce (r-) ore (r-) ere-i-
ere-c-
2PL ce (r-) ore (r-) pe-i-
pe-c-
3 ce (r-) ore (r-) jane (r-) ne(r-) pe(n-) o-i-
o-c-
SET 2 SETS 1+2
A hypothesis for the development of the cross-referencing system which occurs in
independent transitive verbs is presented in sect. 17.
intended in the indigenous language. In (175) and (176), sje and e- are reflexes of
the first person singular *ce and function as the subject of the sentence. The noun
stems are aob 'clothes' and paje 'shaman'.
(175) sje aob º have clothes' or º am clothed' (Tb)
(176) e-paje º have shamanistic manifestations' (Wa)
There are no denominalizers in Tupi-Guarani languages. In those languages which
retain the nominal suffix *-a, its absence results in a verb (sect. 3.3), as in (178) from
Tapir ape.
(177) ne-r-a'yr-a 'your son' (Tp)
(178) ne-r-a'yt 'you have a son'
The alternation between /r/ and /t/ in the preceding two examples corresponds to a
rule which probably occurred in Proto-Tupi-Guarani, which devoiced final non-nasal
consonants (*b and *r) (Appendix III. 19).
To summarize the cross-referencing system which occurs with independent verbs:
The Tupi-Guarani languages use A (Set 1) markers for agentive intransitive
verbs and Ñ (Set 2) markers for nonagentive intransitive verbs (including
nouns functioning syntactically as verbs).
The cross-referencing system used with transitive verbs calls for A (and P)
markers when Ñ is third person, Ñ markers when Ñ is hierarchically superior
to A, and portmanteau forms (Set 4) when A is superior to non-third P.
This system remains essentially intact in all typical Tupi-Guarani languages except
for Urubu-Kaapor. Urubu-Kaapor has eliminated all absolutive cross-referencing on
transitive verbs (that is, all use of Set 2 and Set 4 person markers). Several
languages no longer use the third person Ñ prefix when Set 1 prefixes are used.
Kayabi has eliminated the use of Set 4 person markers, replacing these with Set 1
A prefixes with free pronouns to refer to the second person P.
oro-jo-upe (Kb)
lEX.COREF-RECIP-for
'When our children are born, we make a (special type of) porridge for
each other (of us).'
(192) ko pe kyna r-eko-i ra'e (Kb)
garden at 3SG.F. LK-be-OBTOP PRESENT
'She is at the garden.'
(193) weweri'i je-je'eg-i (Kb)
slowly ISG-speak-OBTOP
º speak slowly.'
(194) kwese sje r-a'yr-a sje r-ajyr-a s-epjak-i (Tb)
yesterday 1SG LK-son-NC 1SG LK-daughter-NC 3P-saw-OBTOP
'Yesterday my daughter saw my son.'
These examples also demonstrate that the adverbial construction which signals the
OBTOP construction may be a word (193, 194), a phrase (192), or a subordinate
clause (191).
The OBTOP construction no longer exists in the Guaranian languages, Wayampi,
and Urubu-Kaapor (C. Jensen 1990:132). In these languages the independent verb
forms occur even when an adverbial occurs in the initial position of the clause, as
in (195) from Wayampi and (196) from Chiriguano (Dietrich 1986).
(195) kwee o-'a (Wa)
yesterday 3-fall
'Yesterday he fell.'
(196) hokope o-mano xe ru (Ch)
there 3-die 1SG father
There my father died.'
A few remnant OBTOP forms nevertheless occur in all of these languages. Most
common are forms based on verbs which have alternate stem forms, such as *jub
and *t-ub 'to be lying down', and *jur and *t-ur 'to come'. In these cases the first
stem is used with independent verb forms, as in (197), and the second is used with
the OBTOP construction, as in (198). The form tui in (198) is derived from *i-tub-i,
and i-tury in (199) is from *i-tur-i.u The word i-hon in (200) is from *;-co-OBTOP.
(197) a'e pe a-ju (Wa)
that at ISG-be.located
º am living there.'
(198) a'e pe tui (Wa)
that at 3.be.located
'It is located there.'
528 Jensen
The Guaranian languages, Wayampi, and Urubu-Kaapor have replaced the absolutive
(Set 2) cross-referencing with the system characteristic of independent verbs.
The equivalent of *-VmV ~ *-rVmV in Guajajära is mehe.
(209) he-ker mehe (GjG)
ISG-sleep when
'when I sleep'
Some languages use coreferential prefixes from Set 3 with verbs when the S or P
of the subordinate clause is identical to that of the independent clause.
(210) o-ho re u-zai'o (GjG)
3-go after 3-cry
'After he left, he (the same person) cried.'
(211) o-ywu re u'yw-a r-eru-a 0-eru-'a-a (Kb)
3-shoot after arrow-NC LK-bring-SER18 3-CC-fall-SER
'After hei shot him2, he2 brought the arrow, falling with it as he came.'
Although Dobson does not report the zero prefix before -eru-'a- a (211), its
occurrence in this context is consistent with the data in other languages (derived
from *c-ero- 'ar-a). In such a construction, the transitive verb should be preceded by
an object.
markers, but have retained some form of the serial suffix. (The most productive form
of the suffix is -vy [bi], derived from the suffix *-a&o.)13
(225) a-eka a-iko-vy (GiM)
ISGA-seek ISG.S-be-SER
went about looking for it.'
Non-agentive intransitive serial verbs are suffixed by *-ramo — -amo. In (226)
from Kamaiurä (Seki 1989), and in Tocantins Assurini (Nicholson 1977), the com-
plete set of coreferential prefixes is used. In Tupinambä, the Set 3 prefix is used only
for third person as in (227).
(226) a-jot we-katu-ram (Km)
ISG-come ISG-good-SER
come to be good.'
(227) aba o-so o-oryb-amo (Tb)
man 3-go 3-happy-SER
'The man went and was glad.'
For first and second person, Set 2 markers are used in Tupinamba; sje katu-ramo
'being good', not \vi-katu-ramo.
1 Valence-changing devices
Valence-changing devices include three types of causatives, reflexive and recipro-
cal morphemes, and object incorporation.
7.1 Causatives
(1) Simple causative
There are three types of causatives in Tupi-Guarani languages. The prefix *mo- is
a simple causative. It may create a transitive verb from an agentive (228, 232) or
non-agentive (229, 233) intransitive verb, from a noun (230, 234), and, at least in
some languages, from certain suffixes (231, 235). The new verb created by mo- is
transitive. The mo- prefix cannot occur on transitive stems, since it is a transitivizer.
(228) *mo- + *co 'go' > *mono 'cause to go, send'
(229) *mo- + *pirarj 'be red' > *mopirarj or *momirarj 'make red, redden'
(230) *mo- + *kwar 'hole' > *mokwar or *morjwar 'make a hole'
(231) *mo- + *-'i 'small' > *mo'i 'make into small pieces'
(232) mo-pok 'cause it to burst' (Gj)
(233) mbo-aku 'cause it to be hot' (GiM)
(234) mu-her 'give (him) a name' (Ur)
(235) mo-ete 'honor, make great' (Tb)
The suffixes in (231) and (235) are described in sect. 3.4.
Comparative Tupf-Guaraiif Morphosyntax 533
In Tupinamba, the mo- prefix co-occurs with the third person object prefix i-
following a subject prefix: a-i-mo-no (from mo- 'CAUS' + so 'go' 'he sent him/it'. In
the Guarani languages, mo- does not co-occur with the object prefix in this context.
In various languages, and probably in Proto-Tupi-Guarani, the nasal consonant in
this morpheme caused a nasalization of the initial voiceless consonant of the follow-
ing morpheme (Appendix III.l). In Tupinamba, this rule had a restriction that there
be no other nasals (consonant or vowel) in that morpheme, as can be illustrated by
the word mo-pirarj 'cause to be red'. This restriction does not hold in Kayabi, as can
be seen by the cognate momirag.
(2) Causative of transitive verbs
The morpheme *-ukar appears to have been derived from a transitive verb. It
occurs after a transitive verb stem. In this construction the person commanding that
the action be done is the A of the verb, the recipient of the action is the P, and the
performer of the action, if stated, is the object of the postposition *cupe. In (237)
from Tupinamba, the commander is first person, the recipient of the action is the
jaguar, and the actual performer of the action is the brother. In (238) from Assurini,
the performer is not stated.
(236) *a-juk -uk r º caused it to be killed.'
Wayampi Assurini
7.2 Detransitivizers
(1) Reflexive
There are four basic ways of detransitivizing a transitive verb. If the reflexive
prefix *je- occurs between the person marker and the stem, this indicates that the
action is performed by the A upon himself, resulting in an intransitive verb.
(245) *o-i-mo'e 'he teaches him'
(246) *o-je-mo'e 'he teaches himself/he learns'
(247) a-upi º lift (it, him, her)' (Wa)
(248) a-ji-upi º lift myself/ I climb'
(249) o-juka 'he killed it' (Tb)
(250) o-je-juka 'he killed himself
(251) u-ze-zuka 'he killed himself (Gj)
The reflexive morpheme may co-occur with causatives, as in (252) and (253) from
Wayampi.
(252) o-si 'he is shy, embarrassed' (WaA)
(253) o-ji-mo-si-oka 'he is ashamed'
Comparative Tupf-Guaranf Morphosyntax 535
(2) Reciprocal
Reciprocal action is indicated by the prefix *jo-. This prefix occurs in the same
position as the reflexive prefix, and the result is again an intransitive verb.
(254) *o-jo-nupa 'they hit each other'
(255) ja-jo-kutuk 'we pierced each other' (Tb)
(256) o-so-soka 'they killed each other' (As)
(257) ja-jo-gweru 'we brought each other' (Ch)
In Wayampi and Guajaj ra only the reflexive prefix occurs. When the verb is
plural, the interpretation may be either reflexive or reciprocal, as in (258). In
Urubu-Kaapor (259) only the descendant form of the reciprocal suffix occurs, but
with a reflexive meaning.
(258) o-ji-nupa kupa (WaA)
3-REFL-hit PL
They hit themselves' or 'They hit each other.'
(259) ju-mu'e 'He learns.' (Ur)
(3) Object incorporation
Another way in which a transitive verb may be detransitivized is by the incorporation
of a non-possessible object between the person marker and the verb stem.
(260) *a-y-'u
1 SG-water-consume
º drink water.'
(261) a-pina-ety (Kb)
ISG-hook-drop
º fish.'
(262) a-ywa-epia (Gu)
ISG-heaven-see
º look toward heaven.'
(263) o-'y-asa (Wa)
3-water-cross
'He crossed the river.'
If the incorporated object is a possessible noun, the verb remains transitive, as in
(264, 266, 267). This may be made reflexive, as in (265).
(264) a-po-kysi º cut his hand.' (Kb)
(265) a-je-po-kysi º cut my own hand.'
(266) oro-po -çä º gave you medicine.' (Wa)
1 > 2SG-medicine-make
(267) o-akag-o23 'He decapitated it.' (Wa)
536 Jensen
8 Verbal aspect
8.1 Future, desiderative, and completion. In many languages the verb *potar, or
a reduction of it (tar or ta\ is used to indicate future. The same morpheme is used
at times to indicate a desiderative.
(274) a-ha-putar º will go.' (Gj)
(275) a-pyta-ta º will stay/ I want to stay.' (Wa)
(276) a-ha-ta º will go.' (Ch)
Urubu-Kaapor (Kakumasu 1986:385) has two morphemes based on *potar: -tar to
indicate desiderative, and -ta to indicate future.
(277) o-ho tar katu 'He really wants to go.' (Ur)
(278) o-ho ta tipe 'He intended to go, but didn't.'
Kaiw (Taylor 1984b) has three morphemes to indicate future, depending on the
degree of certainty (factual or hypothetical) and the distance into the future for the
factual morphemes. The three morphemes are: ta 'immediate future, factual', va'er
'distant future, factual', and ar 'hypothetical'.
(279) o-kwera por i-jase'o-ramo o-porahei-ta
3-recuperate well 3-throat-when 3-sing-PUT
'When his throat gets better, he will sing.'
(280) xe a-menda-ramo a-je-hogapo va'er
I 1 SG-marry-when ISG-REFL-make.a.house PUT
'If I marry her, I promise to build a house.'
In (279) and (280) we can see by the choice of future morpheme that the
recuperation and consequent return to singing is expected to occur soon, whereas the
marriage and housebuilding is expected to occur in the more distant future.
Future and desiderative morphemes are distinguishable in negated constructions in
Wayampi, although they are indistinguishable in non-negated constructions. The
negated future replaces the -ta morpheme with -'a, as in (282).
Comparative Tupf-Guaranf Morphosvntax 537
8.4 Frustrative. Various languages have frustrative morphemes, indicating that the
initiated action did not accomplish its intended result.
(304) i-memy tite (Wa)
3-give.birth FRUST
'She had a miscarriage.'
Comparative Tupf-Guarani Morphosyntax 539
9 Nominalizations
acting syntactically as a noun; that is, it refers to the action of the verb or, in the
case of nonagentive intransitive verbs, to its abstract quality. It has two allomorphs:
*-a, with consonant-final stems, and *-0, with vowel-final stems.
(317) *c-epjak-a 'his being seen'
(318) *ceker-a 'my sleep(ing)'
(319) *ne r-eko-J&T 'his being (in motion)'
(320) *i-katu-0 'his goodness'
(321) s-ekar-a 'the search for him' (Tb)
(322) bebe 'action of flying' (Tb)
(323) t-oryv-a 'party (action of being happy)' (Pt)
(324) kir-a 'sleep (action of sleeping)' (Pt)
In Kayabi (Dobson 1973), the -a suffix occurs with all stems, regardless of
whether they end in a consonant or a vowel.
(325) pe-porowyky-a 'your (PL) work'
(326) kwasiar-a 'drawing'
environment were consonant initial: *-car and *-cab. In Kamaiurä and Assurini the
phoneme *c was eliminated, making the reflexes of *-car and *-cab vowel initial (339,
340), identical with the allophones which follow a consonant-final stem. Rather than
suffixing a vowel-initial allomorph to a vowel-final stem, speakers of these two
languages extended the use of another pair of consonant-initial allomorphs, *-tar and
*-tab, to occur with vowel-final stems (341, 342).
(339) *i-apo-car-a > i-apo-ar-a
(340) *i-apo-cab-a > i-apo-aw-a
(341) i-apo-tara 'its maker' (As)
(342) i-apo-tawa 'the instrument for making it' (As)
Some languages have extended the use of the nominalizer of circumstance to
include the nominalization of action, replacing the nominal case suffix (see sect.
9.5). Among these are Wayampi (C. Jensen 1983) and Guajajära, as indicated in
(343) and (344). If the nominal case suffix had been used in (343) the form in
Wayampi would be e-mo-katu-0-kwer.
(343) *ce mo-katu-cab-wer >
ISO CAUS-good-CIRC-DEVOLV
e-mo-katu-a-wer (WaA)
1 SG-CAUS-good-NOM-DEVOLV
'my healing (my being made good)'
(344) *i-jukä-cäb > i-zuka-haw 'the killing of him' (Gj)
Four languages which have eliminated absolutive cross-referencing in verbs have
also made changes in the cross-referencing system used with these nominalizations.
Urubu-Kaapor (345, 346) uses only Set 1 markers for nominalizations. Mbyä Guarani,
Chiriguano, and Kaiwa use Set 1 prefixes for nominalizations of circumstance (347)
and Set 2 person markers to cross-reference the P in nominalizations of agent (348).
(345) a-kwa-ha 'my knowing of it' (Ur)
(346) u-sak-iha 'the one seeing him' (Ur)
(347) o-mbo'e-a 'the action (or place) of his being taught' (GiM)
(348) i-mbo'e-a 'his teacher' (GiM)
In Wayampi this nominalizer may be used to refer to S of the intransitive clause (365),
A or P of the transitive ckuse (366, 367; context determines interpretation, if A and P are
both third person), possessor (368), or object of the postposition (369):
(365) o-pyta ma'e (Wa)
3-stay NOM
'the one that stayed'
(366) e-nupa ma'e (Wa)
ISO-hit NOM
'the one that hit me (with an instrument)'
(367) a-nupa ma'e (Wa)
ISG-hit NOM
'the one that I hit'
(368) i-posi'a-tawa ma'e (Wa)
3-breast-yellow NOM
'the yellow-breasted one'
(369) a-me'e i-jupe ma'e (Wa)
ISO-give 3-to NOM
'the one to whom I gave it'
At the other extreme, languages such as Guajajära (Bendor-Samuel 1972:119) and
Tapirape (Almeida 1983:32) only allow this nominalizer with agentive and non-agen-
tive intransitive verbs. In these languages the nominalizer serves a complementary
function with the nominalizer -ar. These two nominalizers cannot be analyzed as true
complements of each other, however, since their basic nature is different. All the other
nominalizers which have been described involve a stem level nominalization. The verb
stem is cross-referenced in the same way as nouns and dependent verbs, using person
markers from Set 2. The nominalizer *-ba 'e, on the other hand, nominalizes a complete
verb construction, which includes the same type of cross-referencing as occurs on
independent verbs (357, 359). When the stems of these two types of nominalizations
are negated, they use different morphemes (sect. 10.1-2). The *-ba'e nominalization
uses the negation morpheme appropriate to independent verbs (370), and the *-ar uses
the morpheme appropriate to dependent verbs (371).
(370) n-oo-i ma'e (Wa)
NEG-go-NEG NOM
'the one that didn't go
(371) mojag-ar-e' em-a (Kb)
make-NOM-NEG-NC
One who doesn't make it'
544 Jensen
9.4 Adverbial nominalizer. The suffix *-cwar or *nwar indicates 'that which is
characterized by the preceding circumstance (indicated by an adverb or a postposi-
tional phrase)'.
(372) kope-wat 'the ones from here' (Kb)
(373) Cuiaba-pe-wat 'the ones from Cuiabä' (Kb)
(374) karamoe-war One from long ago' (WaA)
(375) -pe-swär-a 'that which is in the hand' (Tb)
(376) jawa-re-wa One who works with regard to jaguars,
jaguar hunter' (WaJ)
The same is true of Kaiw , which has identical forms -ha in the present and
distinguishable forms -hare and -hawe in the past (Taylor and Taylor 1966).
10 Negation
10.1 The negative morpheme *n- . . . - / . The predicate of the independent clause
in Tupi-Guarani languages is negated by a split affix. The prefix *n- precedes the
person markers and has three allomorphs: */j- before vowel-initial morphemes, *na-
before consonants, and *n- or *ni- before the semivowel *y. When the *-/' suffix
follows a vowel, the two phonemes form a diphthong (Appendix III. 17).
(386) a-maraka º sing.' (Km)
(387) n-a-maraka-ite24 º don't sing.'
(388) sje r-osarj º am patient.' (Tb)
(389) na-sje r-osarj-i º am not patient.'
(390) ja-kua 'We (GENERIC) know.' (Wa)
(391) ni-ja-kua-i 'We don't know' or 'Nobody knows.'
In Tupinamba, which allows the sequence *Cj [Cy], the ni- allomorph does not
occur. The n- allomorph attaches directly to the semivowel j.
(392) n-ja-ker-i 'We didn't sleep.' (Tb)
It is possible that this was also the case in Proto-Tupi-Guarani, and that the ni-
allomorph developed in languages where sequence restrictions made it necessary.
In Guarayu the na- allomorph co-occurs with the semivowel: na-ja-mondo-i 'we
don't send it'.
When a noun occurs as the verb of the sentence, it is negated by this morpheme.
(393) paje 'shaman' (Wa)
(394) i-paje 'He has shamanistic qualities.'
(395) n-i-paje-i 'He does not have shamanistic qualities.'
In Wayampi, where si- has replaced ja- as the first person inclusive A prefix in
transitive verbs, the negative allomorph appropriate to ja- still occurs, as in (396).
(396) ni-si-'u-i 'We didn't eat it.'
Bendor-Samuel (1972:86) states for the Pindare dialect of Guajajara that an allo-
morph na'- occurs with the third person prefix i-, as in (397).
(397) na'-i-kag 'He is not strong.'
The negation prefix is often accompanied in Guajajara by one of two negative
post-verbals, -z (from *j) or kwaw.
(398) n-a-enu-katu-z (GjP)
NEG-1 SG-hear-well-NEG
•I don't hear well.'
546 Jensen
10.2 The negative morpheme *e'ym. The morpheme *e'ym is used in many Tupi-
Guarani languages as a suffix on nouns to mean 'without' or 'lacking', as in (401).
It also negates dependent verb forms (402, 403) and nominalizations (404, 405).
(401) a'eramu te-yar-e'em-a-mu (Kb)
therefore ISG-boat-NEG-NC-?
'Then I was without a boat.'
(402) o'iran awiäwa r-ur-e'ym amoä ere-jot
tomorrow airplane LK-come-NEG if 2SG-come
moneta je-nite-n (Km)
talk ISG-with-FUT
'If the airplane doesn't come tomorrow, are you coming to talk with
me?'
(403) o-ji-monyi i-kua e'y ame (WaA)
3-REFL-scare 3-know NEG CONSEQUENCE
'So he was afraid because he didn't know (it).'
The relative order of the negation morpheme and the nominalizer is opposite in
Kayabi (404) and Wayampi (405).
(404) mojag-ar-e'em-a (Kb)
make-NOM-NEG-NC
One who doesn't make it'
(405) y r-eko-e'ym-a (Wa)
water LK-be-NEG-NOM
'a place lacking water'
According to Rodrigues (1953), both orders are possible in Tupinambä, as in the
following cognates of (404):
Comparative Tupf-Guarani Morphosyntax 547
mojaq-ar-e'ym-a (Tb)
mojaq-e 'ym-ar-a
However, the first is more common.
In Wayampi this morpheme negates a purpose clause in which the subject of the
proposed action is different from the subject of the main clause. This clause is
introduced by the prefix t- and the negation suffix follows the verb stem.
(406) o-mo-katu ja'yr-er t-o-posiko e'y F-ee (WaA)
3-CAUS-good child-COLL PURP-3-mess NEC 3-with
'He put it away so the children wouldn't mess with it
In Kamaiurä this morpheme is used to make a negative assertion.
(407) kamajura e'ym a-ko25 (Km)
Kamaiurä NEG ISG-be
am not Kamaiurä.'
According to Seki (1978), this morpheme occurs in Kamaiurä with nouns, nominali-
zations, and dependent verbs.
In Urubu-Kaapor the negation system has been reduced to the point that the
descendant form of *e'ym, i.e., ym, is the only productive negation morpheme other
than the free response form (sect. 10.5).
(408) e-raho ym 'Don't take it.' (NEG with IMP) (Ur)
(409) sawa'e ym '(He is) not a man.' (NEG of noun)
(410) paite ym '(It is) not far.' (NEG of adverbial)
(411) u-hyk we ym 'He hasn't arrived yet.' (NEG with independent V)
(412) a'i äka ym 'the headless old woman' (NEG of noun)
10.3 The negative morpheme *r«a or *rui. Examples of this morpheme were
found in fewer languages than the other negation morphemes. However, the fact that
it occurs in 5 of the 8 subgroups (Appendix IV) is a good indication that it can be
reconstructed for Proto-Tupi-Guarani. It negates adverbials (including postpositional
phrases). In some languages it also negates verbs or nouns.
Ruiz de Montoya (1892) actually lists two separate negative morphemes for Old
Guarani: ma (also referred to as ruguä /ruwaT) and ruguäj /ruwäj/. It is possible that
the second is actually a combination of ruä and the suffix -/ (sect. 10.1). It co-occurs
with na- (414, 416).
(413) awyje ruwä pa (GiO)
enough NEG INTER
'Isn't it enough?'
(414) na ce-ma'e ruwäj
NEG ISG-thing NEG
'It isn't mine.'
548 Jensen
10.5 The free response negative morpheme *ani. The form used as a free re-
sponse to a question in most Tupi-Guarani languages is derived from *ani.
(434) am, e-raho ym 'No, don't take (him).' (Ur)
(435) ani, n-a-a-'ai 'No, I won't go.' (Wa)
Mbyä Guarani has two morphemes: 'ani, a strong negative imperative, and 'any,
the free response.
11 Demonstratives
Demonstratives have a wide function in Tupi-Guarani languages. The same mor-
phemes may refer to persons, objects, time or location, or they may make reference
to elements of a discourse. They may occur with other morphemes, such as the
temporal subordinate morpheme, postpositions, the suffix *-7 'DIMINUTIVE', and
the suffix *-ete 'GENUINE'.
Rodrigues (1981) charts the basic demonstrative stems for Tupinambä using the
parameters of visibility (visible versus invisible) and distance (near the speaker, near
the hearer, or far from both). Nine of the Tupinambä morphemes have cognates in
other languages. I shall list these, together with words from various languages which
are based on these reconstructed morphemes, in order to show the flexibility of
meaning. One other morpheme which was not documented for Tupinambä, *pe, also
appears in various languages. In the reconstructed forms I was able to use the
parameters of visibility and distance, but was usually only able to make a two-way
distinction (near/far) in the latter.
550 Jensen
*wi or *wirj 'that one (visible), which may be far from the hearer also'
wi, wirj 'that one (visible)' (Tb)
wiba'e 'this, that (visible)' (Tb)
wiirja 'this, that' (Tb)
wime, wilme 'there (visible)' (Tb)
wime 'there (far from speaker and hearer)' (Wa)
wyi 'that one, farther away' (Wa)
awT 'this one, closer' (Wa)
gwi 'that' (Pt)
gwino 'with that one' (Pt)
12 Particles
The Tupi-Guarani languages are rich in particles. They occur at specific places in
the clause. One especially common place for them to occur is post-initial, that is,
immediately after the first element (word, phrase, or subordinate clause) of the
sentence.
(436) ihe ke a'e u-sak (Ur)
1SG FOCUS 3 3-see
'He saw me.'
(437) ore te si t-oro-posiko (WaA)
1EX EMPH EXCL PURP-lEX-work
'We are the ones that ought to work!'
Another place where particles typically occur is at the end of a clause.
(438) naikoi 0-erekwar ko ky'y (WaA)
NEG.EXISTENTIAL 3-wife PL CHANGE.OF.STATE
'None of their wives existed anymore.'
(439) u-zekwa wä zi'i kury (GjP)
3-dawn PL PAST CHANGE.OF.STATE
'Then it began to dawn for them.'
Bendor-Samuel (1972:146-157) describes particles for Guajajara in initial (440),
post-initial (441), medial (442), and final (439) positions.
(440) aze arapuha ur
SUPPOSE deer 3.come
'Suppose a deer comes.'
(441) u-ahem ze o-ho pe pe
3-arrive THEY.SAY 3-go there to
'They say that he arrived there.'
(442) u-'aw u-pa kwez pe pe
3-lie.down 3-prone.position PAST there at
'He lay down over there just now.'
Comparative Tupi-Guarani Morphosyntax 553
Bendor-Samuel says that the medial particle follows any verb phrase, subject, object,
or auxiliary verb, but that it precedes any adjuncts. He describes nine positions of
final particles. Particles in these positions can co-occur, as in (443).
(443) -ho zaryz a'e wä rihi no
3-go grandmother 3 PL YET CHANGE.OF.TOPIC
POSITION: 1 2 5 7
'Granny and others went still.'
In Wayampi (A. Jensen 1993) it is not unusual for several particles to co-occur in
post-initial position.
(444) awi te ruä si po ko e-mo-katu (WaJ)
he EMPH NEC EXCL INTER DETERM ISG.P-CAUS-good
'He was the one that healed me, wasn't he?'
The series of five particles in (444) indicates a rhetorical question in which the
speaker believes that he is correct in his conclusion. The first particle, te, is often
best translated with a cleft sentence in English. The negative particle ruä conveys
the speaker's belief about his conclusion. The exclamation particle si conveys an
emotional response. The po indicates a question. The ko indicates that he wants an
answer to his question. The five particles together form a rhetorical question in
Wayampi.
12.1 Particles indicating the speaker's feelings. Many of the post-initial particles
convey to some degree the speaker's feelings. (Some of these same feelings are
communicated in English through intonation.)
Among the particles which occur in this position in Wayampi are the following:
ipe 'unrealized goal', ko 'deliberation', si 'surprise, exclamation', so 'disapproval',
te 'emphasis', to 'anger' (only in Jari dialect), and kua 'resolve'.
Particles of this sort in Guajajara (Bendor-Samuel 1972:150) include the follow-
ing: ete 'emphasis', kutu 'reaffirmation', mua'u 'mistake, deception', poko
'deliberation', ru'u 'uncertainty', tezyz 'frustration', tomo 'deliberation', and zepe
'incomplete success'.
Particles of this type in Mbyä Guarani (Dooley 1982) include: ko Opinion', po
'doubt', nda 'amusement', tavy 'discontent', tu 'discontent', katu 'lack of patience'.
Particles in Guarayu (Newton 1978) include: m'ä 'certainly', revo 'perhaps,' and
rule 'seeing but hard to believe'.
time: rihi 'still, yet (looking forward to change of state or action), ri'i 'action in the
past', ra'e 'unfortunate or otherwise amusing action, often in the past', zipi 'habitual
past action, often incompletely successful', and kury 'action or state at point of
change'.
Taylor (1984b:66) reports four past markers for Kaiwä, based on distance into the
past and whether the speaker witnessed the action. These are kuri 'recent past,
attested', va'ehve 'remote past, attested', ra'e 'recent past, unattested by speaker',
and araka'e 'distant past, unattested by speaker'. The particle araka'e also occurs in
Guarayu, meaning 'distant past.'
The same distinctions which are made in Kaiwä are also made in Tocantins
Assurini (Nicholson 1978:67). The particles in this language are raka 'recent past,
attested by speaker', rakokwehe 'remote past, attested by speaker', ra'e 'recent past,
attested by someone other than the speaker,' and rakwehe 'remote past, attested by
someone other than the speaker'.
In Wayampi there are two verification particles which imply past action: ipo
'unattested by the speaker' and je 'no longer verifiable (more distant past)'. These
occur immediately after the first constituent of the clause.
Cognates of ipo and je occur in various languages. The morpheme ipo is translated
as 'maybe' in Tupinamba and 'uncertainty or doubt' in Tembe. Its cognate aipo in
Guarayu is translated as 'hearsay'. Cognates of the morpheme je are translated as
'they say' or 'it is said that' in Tembe and Kamaiurä, and as a hearsay particle in
Urubu-Kaapor.
13 Ideophones
Ideophones are a part of good storytelling in many Tupi-Guarani languages. Some-
times sounds are employed in ideophones which are not part of the phonetic inventory
of normal vocabulary. In Kayabi (Weiss 1989), these include voiceless and whispered
Comparative Tupf-Guaranf Morphosyntax 555
vowels, front rounded whispered vowels, lengthened vowels, bilabial voiceless trills,
and word-final glottal stops. The following are examples of ideophones.
puuUUU Tiring of gun' (Ur)
tu tu tu 'thumping' (Ur)
tä tä tä tä 'banging' (Ur)
warn warn warn 'whine of mosquitos, flapping their wings rapidly' (Kb)
ky: tykyty 'macaw flying flapping its wings' (Kb)
pyutik 'shooting (an arrow) once' (Kb)
uuuu tarä 'arrival and landing of a large bird' (Wa)
py' moro py' moro 'moving and stacking objects' (Wa)
Py üü Object falling' (Wa)
14 Numbers
The numerical system of Proto-Tupi-Guarani is very rudimentary. Only three
numbers have been reconstructed (Lemle, 1971):
*ojepetei One'
*mokoj 'two'
*mocapyr 'three'
Whereas the cognates for 'two' and 'three' are quite straightforward, the only part
of the *ojepetei construction that occurs in all languages is pe. This means that the
form in various languages could actually be made up of the basic stem plus a
combination of affixes, which have become a permanent part of the word.
(445) pe One' + 'i 'DIMINUTIVE' > pe'i One' (Wa)
Forms in other languages include osepe (As), ojepe (Tb), pitei (Gj), ojipefl (Pt),
ojepete (Km), and pete 7 (Ur, Gi).
The concept of pairs seems to be important in Tupi-Guarani languages. The word
for 'pair', *iru, has resulted in forms like the following from Wayampi:
(446) niröi Odd number' (Lit.: 'not a pair')
(447) irote 'four' (Lit.: 'true pair')
(448) iröiröte 'multiple pairs,' that is, '6 or 8'
In languages of subgroup IV, the term for Odd number' has been reinterpreted to
mean 'three'. For example, in Assurini, nairoihi, and in Guajajara na'iruz.
Urubu (Kakumasu and Kakumasu 1988:71) has developed a numerical system up
to 20 using entirely indigenous vocabulary.
15 Constituent order
In all types of dependent constructions the verb usually occurs in the final position
of the clause and the morpheme signaling the construction occurs as a suffix on the
556 Jensen
SOV, OSV, VSO, VOS, OVS, SVO. (See also Derbyshire and Pullum 1986:17 for
their comments on Urubu-Kaapor constituent order.)
Harrison (1986) reports that in 200 pages of text the following orders were found for
Guajajära: VSO - 19, VOS - 4, SVO - 3, SOV - 2, OVS - 0, OSV - 0. His results
demonstrate that it is very unusual to have all three constituents in a transitive sentence.
Only 28 occurrences appeared in 200 pages of text. Nouns tend not to be repeated once
they have been introduced into the discourse.
In Wayampi all six word order possibilities occur, although OSV is almost nonex-
istent. The word order is dependent on discourse factors, with new or definite
information (as subject or object) occurring before the verb and old or indefinite
information occurring after the verb (A. Jensen 1982, C. Jensen 1980). In most
instances, logic will tell the hearer which noun is the subject and which is the object.
When there is doubt, the hearer will ask.
The reported word orders (Grimes 1996 and Kakumasu 1986) would indicate that
the basic word order for the family should be SOV or SVO. Considering that
dependent constructions are usually verb final, I suggest that SOV was the original
order for independent verbs as well. This area has not been sufficiently studied for
us to say what degree of flexibility was allowed for independent verbs in Proto-Tupi-
Guarani or what would have been the motivating factors for such flexibility.
16.1 That r- is epenthetical. According to the first hypothesis, the r- prefix was
originally an epenthetical consonant. This is a reasonable suggestion from the stand-
point that certain other morphemes have allomorphs beginning with r when preceded
by a vowel-final morpheme (Appendix 1:23,24,40,44). The r- would keep the vowels
of the two morphemes separate. It would be necessary to determine whether the
epenthesis would operate at morpheme boundaries or between words.
Let us review what kinds of rules have already been described in this language
family for vowel or consonant sequences on the morphological level (C. Jensen 1989
and Appendix III). There have been two rules proposed for the insertion of epen-
thetical vowels to separate consonants at morpheme boundaries where a stem
combines with an affix (Appendix III.3,15). No rules have been proposed regarding
epenthetical consonants. When two stems combine, creating a consonant sequence,
the first consonant is eliminated in most languages of the family (Appendix III.6).
Rodrigues demonstrates for Tupinambä that when a low vowel is followed by a low
stressed vowel (in morphemes such as the serial verb suffix and certain nominaliza-
tion suffixes), the second vowel takes on the quality of the first vowel, and the first
vowel is deleted (Appendix III.7,10,12). These rules account for allophonic vari-
ations in Kayabf as well. In other rules he shows that in a sequence of two vowels
558 Jensen
where one vowel is a high vowel, the high vowel becomes asyllabic (Appendix
III.11,16). He has also proposed rules for the insertion of a semivowel: [y] when the
first vowel is a high front vowel and [w] when the first vowel is a back vowel
(Appendix 111.14,18). A rule for epenthesis of r- on the morphological level would
conflict with the other rules for handling vowel sequences. Furthermore, data dem-
onstrate that many vowel sequences do exist at morpheme boundaries without any
epenthesis taking place. Consider, for example, *o-apy > oapy (or owapy) 'it burns',
*ere-y'u > erey'u 'you drink', *oro-epjak > oroepjak see you SG'. In sum, r- is
not epenthesized at the morphological level, and a rule of epenthesis would actually
conflict with other phonological rules which operate at this level.
If the epenthesis operates on the word level, we are still faced with the fact that
its occurrence is restricted to situations in which the first of the two words is either
a noun or a person marker from Set 2. Syntactic rules in Proto-Tupi-Guarani are
such that any noun (functioning as such in a sentence) terminates with a vowel (sect.
3.3) and therefore automatically provides the right environment for the formation of
a vowel sequence. However, a verb may also end in a vowel, but no r- is inserted
between it and the following word. Furthermore, there are other sequences of words
in which the first word is a noun but no epenthesis occurs. In (455) from Tupinambä,
the word order is subject, independent verb, object, serial verb. Vowel sequences
occur at the boundary of the subject and verb and at the boundary of the verb and
object, yet in neither case is an epenthetical vowel inserted.
(455) kujä o-s-aro o-memyr-a s-ereko-feo (Tb)
woman 3A-3P-care.for 3COREF-child-NC 3P-keep.with-SER
'The woman cares for her child, keeping it with her.'
If the insertion of r- were governed purely by phonological rules, no distinction
should be made by grammatical role (noun, verb, postposition) or the relative order
of grammatical constituents. The r- does in fact have a grammatical interpretation.
The r- only occurs on a transitive verb if the verb is directly preceded by its object
(which takes the form of a first or second person marker or a noun, when absolutive
cross-referencing is employed). Examples (456) and (457), which have the r- pre-
ceding the verb stem, are variations of (455) put into a temporal subordinate clause.
The first clause can only mean that the woman is the object of the caring and the
second can only mean that the child is the object of the caring.
(456) kuja r-aro-reme 'when (someone) cares for the woman'
(457) o-memyr-a r-aro-reme 'when she cares for her child'
In sum, the r- cannot be an epenthetical consonant because its occurrence is
defined on a grammatical basis. At the most, it could have originally been a
phonologically conditioned morpheme, occurring before all vowel-initial stems un-
der the appropriate grammatical conditions.
Comparative Tupf-Guaranf Morphosyntax 559
16.4 The question of allomorphs. Payne (1994) points out that "all the construc-
tions where [r-] occurs are P-oriented, or at least not -oriented." This is the case when
P and A are relevant categories (i.e., in verb constructions). Because *r- is P-oriented,
560 Jensen
16.5 The Inverse hypothesis. Payne (1994) suggests that r- is the marker of an
inverse system (in transitive verbs). I agree with this, but must emphasize that this
is not the primary function of r-. As I will argue in sect. 17, the development of a
direct/indirect system in independent verbs is relatively recent. Before its develop-
ment, the r- morpheme was already alive and well, functioning in various contexts,
where a noun or pronoun (first or second person) was tightly knit to the following
word (verb, noun, or postposition).
Furthermore, as an inverse marker the r- is largely redundant and limited in scope.
It is redundant in the sense that in most cases the person markers indicate clearly
whether the cross-referencing is direct (for example, *a- 'first person singular A') or
indirect (*ce 'first person singular P'). Even with the second person plural markers
*pe- and *pe, marking A and P respectively, there would not have been ambiguity
in Proto-Tupi-Guarani. First of all, the P marker was stressed. Second, the A prefix
was always followed by the third person P prefix as in (463).
(463) *pe-c-epjak 'You PL saw him.'
(464) *pe n-epjak '(Someone) saw you PL.'
Thus, even without the *r- morpheme (n- in this context), there would be no
ambiguity.
In some descendant languages, such as Wayampi (465, 466), where the P person
marker is no longer stressed, and where the third person object prefix is no longer
used, the *r- morpheme does eliminate ambiguity between A and P.
(465) pe-esa 'You PL saw him.'
(466) pe-n-esa '(Someone) saw you PL.'
Even so, the r- is limited in scope as an inverse marker, as it eliminates the
ambiguity only with Class II stems. With Class I stems the ambiguity still exists.
Comparative Tupf-Guaranf Morphosyntax 561
16.6 Other considerations. In the OBTOP and dependent verb constructions, where
the cross-referencing system is ergative-absolutive, the r- has the function of clarifying
in transitive sentences which of the two nouns is the A and which is the P. If the noun
immediately preceding the transitive verb is P, the verb receives only the r- morpheme,
as in the following example from Tupinamba.
(468) kwese sje r-a'yr-a sje r-ajyr-a r-epjak-i
yesterday 1SG LK-son-NC 1SG LK-daughter-NC LK-see-OBTOP
'Yesterday my son saw my daughter.'
If the noun immediately preceding the transitive verb is the A, the cross-refer-
encing on the verb is still absolutive, but this is marked by the third person prefix
*c- or */-, as in the following example from Tupinamba.
(469) kwese sje r-ajyr-a sje r-a'yr-a s-epjak-i
yesterday 1SG LK-daughter-NC 1SG LK-son-NC 3-see-OBTOP
'Yesterday my son saw my daughter.'
I suggest that the P in the first example is part of the verb phrase, while the P in the
second example is an example of object raising, with the P still being marked in the
verb phrase by s-. The r- should then be interpreted as a phrase indicator.
Another important consideration is the significance of the two marking systems for
agentive intransitive verbs. In the following examples, the independent verb (470) takes
the Á-oriented (Set 1) prefix while the dependent verb (471) takes the P-oriented (Set
2) markers.
(470) *ere-iko 'You (SG) are (in motion).'
(471) *ne-r-eko-rVmV 'when you SG are (in motion)'
Harrison (1986) refers to this difference as one of control (in independent verbs)
versus non-control (in dependent verbs). We might say that the independent verb is
more verbal than the dependent verb, and that the independent subject is more
agentive than the dependent subject. Payne (1994) suggests that the r- may have
been a genitive marker in Pre-Proto-Tupi-Guarani. Perhaps dependent verbs were at
one time treated as nominale. There is certainly a resemblance between two of the
serial verb suffixes and the nominalization suffixes of agent and circumstance
562 Jensen
(Appendix I, morphemes 21, 30, and 31). There is also a similarity in some lan-
guages between the temporal/conditional morpheme and the attributive case
morpheme which occurs with nouns (Appendix I, morphemes 23 and 44).
In sum, r- is a morpheme which operates on the phrase level to express a close
relationship between two words, such as object-verb, genitive-noun, and object-postpo-
sition. Its use with verbs may have been an extension of a genitive marker in Pre-
Proto-Tupi-Guarani. Its significance in relation to verbs is probably best understood by
its occurrence with agentive intransitive verbs. Whereas it may now be the marker of
an inverse system in transitive verbs, the direct-inverse distinction is probably a later
development, which accompanied the development of the person hierarchy in inde-
pendent transitive verbs. The r- morpheme may originally have occurred with all
vowel-initial stems; those Class I stems which in Proto-Tupi-Guarani are vowel initial
may have lost an initial consonant.
It would be good to know the reactions of native speakers of Tupi-Guarani
languages who have a sufficient degree of linguistic sophistication to discuss the
structure of their language. Their impression as native speakers could be significant
to our understanding of the r- morpheme.
Transitive
(473) o-o e-r-esa 'He went to see/meet me.'
3 -go 1 SG-LK-see/meet
Set 1 Set 2
Guarani nominalizations of intransitive verbs (474) take Set 1 prefixes, as do nomi-
nalizations of action or circumstance in transitive verbs (475). Nominalizations of agent
(476) retain Set 2 markers to cross-reference P.
Guarani nominalizations
Intransitive
(474) ere-o-a 'your going'
2SG-go-NOM
Set 1
Transitive
(475) ere-mbo'e-a 'your teaching'
2SG-teach-NOM
Set 1
(476) ne-mbo'e-a 'your teacher'
2SG-teach-NOM
Set 2
Table 22 is a list of the grammatical structures which underwent change by substituting
Set 1 person markers and the languages in which the substitution appears. By comparing
the extent of change in the various languages, we can see that the transition was progres-
sive, beginning in the least dependent verb forms (OBTOP and temporal subordinate) and
moving to the most dependent (nominalizations).
Table 22: Proposed ordering and grouping of changes
Oblique-topicalized Gu, Wa, GiM, Kw, Ch, Ur
Temporal subordinate verbs Gu (partial), Wa, GiM, Kw, Ch, Ur
IV Serial verbs Gu (partial), Wa, GiM, Kw, Ch, Ur
IV Nominalizations GiM,Kw,Ch,Ur
TV Nominalizations of action, circumstance GiM,Kw,Ch,Ur
TV Serial verbs Kw,Ch,Ur
TV Nominalizations of agent Ur
Harrison (1986:427) proposes that the ergative-absolutive system which appears in
dependent verbs is older than the person hierarchy which occurs in independent verbs.
He suggests that originally the basic word order was SOV, and that the free pronoun
closest to the verb eventually became more firmly attached to it: s-IV and o-TV. The
result would be an absolutive cross-referencing system. Harrison describes this as
564 Jensen
classical CB (Central Brazil) cross-referencing. From this system the person hierarchy
developed in independent verbs, by a process which Harrison refers to as creeping
accusativity.
When we propose that all cross-referencing in Pre-Proto-Tupi-Guarani was absohitive,
we hope to find evidence to support this claim. I suggest four types of evidence:
1) remnants of older constructions found in languages in which recent
changes have eliminated or substantially modified those constructions,
2) a model which shows how this change might have come about,
3) evidence of cognate forms and cross-referencing system outside the
Tupi-Guarani family, in the Tupi stock,
4) unique characteristics of the Oblique-Topicalized construction.
We will examine each of these four possibilities.
Remnants of the Serial verb system in languages in which the system has been
modified by (a) changing from Set 2 to Set 1 prefixes in intransitive verbs, or, in the
case of Wayampi, elimination of prefixing, and (b) whole or partial loss of the serial
verb suffix:
Comparative Tupf-Guaranf Morphosyntax 565
17.2 A model for the development of the person hierarchy. The cross-refer-
encing system which occurs on independent transitive verbs can be derived from an
absolutive system by 1) the addition of Set 1 prefixes preceding P prefixes when P
is third person, 2) the redefinition of the extent of usage of first and second person
P prefixes in a person hierarchy rule in which hierarchically superior P is marked, a
rule which excluded the use of second person P person markers when P is hierarchi-
cally inferior, and 3) development of a set of portmanteau forms where the Set 2
prefixes were excluded.
Object
Subject Object
1SG 1EX UN 2SG 2PL 3
1 X X X ne pe i-/c-
v s
2 ce ore X X X i-/c-
3 ce ore jane ne P? i-/c-
Stage 2. At some point agentive intransitive verbs developed a new prefixing system
(Set 1 prefixes) which distinguished them in independent verbs from non-agentive
intransitive verbs.
*a-ker º sleep' *a-iko º am (being)'
*oro-ker 'we EX sleep' *oro-iko 'we EX are'
*ja-ker 'we IN sleep' *ja-iko 'we IN are'
*ere-ker 'you SG sleep' *ere-iko 'you SG are'
*pe-ker 'you PL sleep' *pe-iko 'you PL are'
*o-ker 'he sleeps' *o-iko 'he is'
Comparative Tupf-Guaranf Morphosyntax 567
My justification for proposing that the occurrence of Set 1 prefixes in intransitive verbs
preceded their occurrence in transitive verbs is based on the observation that more recent
changes occurred first in intransitive verbs in descendant languages (sect. 17).
Stage 3. At the next stage of development, these Set 1 prefixes became attached
by analogy to transitive verbs whenever the P was third person (indicated by *;- or
*c-), without eliminating the P prefixes. The Set 1 prefixes referred to A. When P
was first or second person, there was no change. This is illustrated in Table 26.
Table 26: Addition of Set 1 Prefixes to Transitive Verbs, by analogy
with Intransitive Verbs, Pre-Proto-Tupi-Guarani
V. Class I V. Class II
Object
1SG 1EX UN |2SG 2PL 3
Cross-reference P Cross-reference A + P
Once this system became established, the i- and c- became redundant, since the Set
1 prefixes only occurred when P was third person. Eventually the third person P
prefixes were eliminated in several languages.
*aipotar > apotar
568 Jensen
The languages which eliminated third person P prefixes in independent transitive verbs
include Wayampi, Urubu-Kaapor, Guajajara, Tocantins Assurini, Kamaiura, Parintintih,
and Kayabi. In some languages in which the P prefix was deleted, remnants still appear
in certain words (sect. 5.1). The loss of the P prefix does not appear to have any
connection with further cross-referencing changes, since languages from Subgroups I
and VIII show cross-referencing changes, whereas languages from Subgroups IV
through VIII have eliminated the P prefix.
Table 28 is a modification of Table 27 which reflects the elimination of the third
person P prefix in some languages.
Table 28: Intermediary stage following the addition of A prefixes and
elimination of P in some languages, Pre-Proto-Tupi-Guarani
Object
1SG 1EX UN [2SG 2PL 3
Cross-reference P Cross-reference A (+ P)
As in Table 28 Object
1 3
Cross-reference P Cross-reference A (+ P)
Ste
P 1 Object
1 2 3
Cross-reference P Cross-reference A (+ P)
Step 2 Object
2
Cross-reference A (+ P)
Although this result did not happen in a direct way in any Tupi-Guarani language,
Step 2 illustrates Urubu-Kaapor, which only cross-references A.
Comparative Tupf-Guaranf Morphosyntax 569
Subject Object
1SG 1EX UN 2SG 2PL 3
1 ÷ ÷ >: cross-reference
2 >ß X X A(+P)
3 cross-reference P 7
This hierarchy is a simplification of what actually takes place in Tupi-Guarani
languages in the cross-referencing of independent transitive verbs. It was used by
Harrison (1986:420) to demonstrate for the first time the relevance of Silverstein's
agency hierarchy to Tupi-Guarani languages, and in doing so, it is natural that he
would speak in general terms. Of the languages cited in this paper, the one whose
cross-referencing system most closely fits the description of the "1>2>3" rule is
Kayabi, which uses first person A prefixes when Ñ is either second or third person.
Within Tupi stock, Aweti also fits this description. (This is one of the two languages
which are most closely related to the Tupi-Guarani family.) This definition of the
person hierarchy has two weaknesses in regard to Tupi-Guarani languages: 1. It does
not explain what happens when A and Ñ are equal in number (that is, third person). In
this environment, *o-i-potar and *o-c-epjak occur. The forms *i-potar and *c-epjak are
not acceptable as independent verbs. 2. It ignores the special forms which occur in
most languages when the A is first person and Ñ is second person.
Let us consider a refinement of the above rule: hierarchically superior Ñ is
cross-referenced on the verb. We are assuming that the development is still in effect
by which the A is marked by Set 1 prefixes whenever Ñ is third person. Nothing is
stated about what happens when A is hierarchically superior to a non-third person Ñ
(that is, a second-person P). Under the original ergative-absolutive cross-referencing
system, the second-person prefixes *ne and *pe could be used as the second person
singular object markers regardless of whether the subject was first or third person,
as in (477-480).
(477) *ne potar º/we like you SG.'
(478) 'He/she/they like(s) you SG.'
(479) *pe potar º/we like you PL.'
(480) 'He/she/they like(s) you PL.'
570 Jensen
However, under the refined person hierarchy rule, these prefixes can be employed
only when the P (2nd person) is hierarchically superior to A, as in (478) and (480).
This is only'the case when A is third person. When a first person A acts on a second
person P, as in (477) and (479), the cross-referencing is not governed by the person
hierarchy rule, since P is hierarchically inferior to A. On the other hand, there is no
indication that we should use A markers, since this set (1) is used only when P is
third person. There is a gap, as indicated by the question marks in Table 31. (The
x's in the table indicate that reflexive forms rather than transitive forms occur.)
Table 31: Person hierarchy where hierarchically superior objects are
marked, Pre-Proto-Tupi-Guarani
Subject Object
ISO 1EX UN 2SG 2PL 3
? ?
1 X X X cross-reference
A ( + P)
2 X ÷ ÷
3 Cross-reference P
This is the intermediary stage following the introduction of the person hierarchy rule
in Pre-Proto-Tupi-Guarani.
17.2.3 Filling the gap. To summarize developments up to this point, the person
hierarchy rule excluded the use of P markers *ne and *pe when the A was first
person, because in this case P is not hierarchically superior to A. However, the new
A prefixes were employed only when the P was third person. Some arrangement had
to be made to cover the cases where A is first person and P is second person.
Stage 5. To fill this gap, two portmanteau prefixes were developed:
*oro- º A acting on 2SG P'
*opo- º A acting on 2PL P'
This completes the system reconstructed for Proto-Tupi-Guarani independent tran-
sitive verbs, summarized in Table 32.
Comparative Tupf-Guaranf Morphosyntax 571
Subject Object
ISO 1EX UN 2SG 2PL 3
1 X X 3L Cross-reference Cross-reference
A-P Set 4
2 >é ÷ ÷ A(+P)
Set 1 ( + 2)
3 Cross-reference P
Set 2
The two portmanteau prefixes seem to be a sort of compromise, not aligning
themselves with either the original ergative-absolutive system which marks Ñ (with
Set 2 markers) or to the new system which marks A (with Set 1 markers). They are
unlike Set 2 person markers in that they do not require the linking r- prefix between
the stem and the prefix, as can be seen by comparative examples (481-484) of a
Class II verb.
(481) *ce r-epjak '(You/he/she/they) saw me.'
(482) *ne r-epjak '(He/she/they) saw you SG.'
(483) *oro-epjak º/we saw you SG.'
(484) *opo-epjak º/we saw you PL.'
On the other hand, the *oro- and *opo- prefixes are unlike the Set 1 person
markers in that they refer simultaneously to A and P. The prefix *oro- is homony-
mous with the Set 1 prefix *oro- ºÅ×', but it has a different meaning. The Set 1
prefix refers only to first person exclusive A. The Set 4 prefix occurs when Ñ is
second person singular, and A is first person, either singular or plural. Those
languages which have both A and Ñ prefixes when the object is third person have a
contrast, as in (485) and (486), but those languages that deleted the Ñ prefix have
one word with two different meanings, as in (487) from Wayampi.
(485) *oro-i-potar 'We EX like him/her/it.'
(486) *oro-potar º/we EX like you SG.'
(487) oro-pota 'We EX like him/her/it' or (Set 1),
º/we EX like you SG (Set 4).' (Wa)
The *oro- and *opo- prefixes are unique in themselves. Perhaps it is the fact that
these two prefixes occur in a sort of buffer zone between two systems that allows the
variants of *opo- to show greater diversity in individual Tupi-Guarani languages than
any other person prefix (Table 33). Variations in this zone do not undermine either of
the two systems.
572 Jensen
In this case, the reinterpretation of *opo- as po- analogous to *;'- and *c- reflects
a mental association of this prefix with the new system of conjugation. What
remains unexplained is why similar variations did not develop for *oro- º A acting
on 2SG P'.
In Guarani, Assurini, and some dialects of Guajaj ra and Kaiw , *opo- has neutral-
ized with *oro- into a single prefix, oro-, while Wayampi has substituted poro- for
*opo-. All other languages retain *opo- or a direct phonetic consequence of it.
It can be seen that although nearly all Tupi-Guarani languages have special prefixes
when A is first person and P is second, there is a lot of variation in the forms. I suggest
that this was permissible because of the uniqueness of these forms having been
excluded from two wider-scope cross-referencing rules: one which called for the
marking of hierarchically superior objects, and the other which called for the marking
of subjects when the object was third person.
To summarize the cross-referencing system which developed for independent
transitive verbs: When P is third person the verbs are marked for A (Set 1), followed
in some languages by the prefix for third person P (Set 2). A person hierarchy rule
requires that hierarchically superior P be marked by person markers from Set 2.
When both P and A are third person, the first condition applies and the A prefix
must be used. When hierarchically superior A acts on a non-third person P (that is,
first acts on second), neither of the above conditions apply; this gap was resolved by
the development of a unique system of portmanteau markers (occurring in all
languages but Kayabi and Urubu-Kaapor). This complex cross-referencing system
has been extended to all syntactic structures which have undergone change in
Wayampi and the Guaranian languages. Urubu-Kaapor has eliminated the person
hierarchy rule as well as the set of portmanteau markers, so that all transitive verbs
cross-reference A.
The model of superimposing Set 1 prefixes on forms already prefixed for third
person P, as in Table 26, explains how transitive verbs might have begun to use A
prefixes. Table 31 shows what the system would be like if a rule developed which
required that hierarchically superior objects be cross-referenced. The gap, illustrated
by question marks, shows the scenario which may have permitted the formation of
the portmanteau prefixes. It also allows us to explain why the variety of forms in the
different languages is permissible.
17.3 The search for cognates outside the Tupi-Guarani family. If we look be-
yond the Tupi-Guarani language family to the Tupian stock for some evidence of
what the original cross-referencing system was like, we can only conclude that the
development of the personal pronouns, as well as the person markers of Sets 1, 2,
and 4 are much more recent than we might have expected.
In Satere-Mawe (Graham, Graham, and Harrison 1984), the Tupian language
probably most closely related to the Tupi-Guarani family, there is a comparable
system with a person hierarchy. Possible correspondences of person markers from
Set 1 are the following: a- SM, *a- P-T-G 'ISO'; wa- SM, *ja- P-T-G ºÉÍ'; uru-
574 Jemen
SM, *oro- P-T-G ºÅ×'; to- SM, *o- P-T-G '3'. From Set 2, the third person P
forms /- and h-, comparable to */- and *c-, co-occur with the Set 1 markers. In
Satere-Mawe they occur in other situations as well, where they are not called for in
the Proto-Tupi-Guarani languages. The Satere-Mawe prefixes for second person, e-
'2SG' and ewe- '2PL', have some similarity to the Set 1 P-T-G prefixes *ere- and
*pe-, but they are a more likely match with the Set 3 prefixes *e- and *peje-,
especially since occurrences of w in Satere-Mawe correspond to j in Proto-Tupi-
Guarani (Rodrigues, 1984/1985). Satere-Mawe also has a prefix aro- which
corresponds to the Proto-Tupi-Guarani prefix *oro- º A acting on 2SG P' from Set
4. The person hierarchy in Satere-Mawe is similar to that in Proto-Tupi-Guarani,
except that it has been extended to subordinate clauses as well.
If we compare the Tupi-Guarani cross-referencing system with Aweti, we find that
Aweti has one set of prefixes for intransitive verbs, and another for transitive verbs.
There is partial overlap between the two. The transitive verbs have a person hierar-
chy that fits the definition: "1 > 2 > 3". Intransitive prefixes which may
correspond to those in Proto-Tupi-Guarani are a- '1SG', üæï- ºÅ×', e- '2SG', o-
'3'. Prefixes in transitive verbs include those referring to A: a- '1SG', üæï- ºÅ×',
e- '2SG', and perhaps ti- ºÉÍ' (see sect. 5.1); and those referring to P: e- '2SG',
and üæï- ºÅ×'. Once again e- is more like the Proto-Tupi-Guarani prefix from Set
3 *e- than the one from Set 1 *ere-. The free pronouns 'en '2SG' and e 'ipe '2PL'
could correspond to *ene and *pe . . . e.
Going beyond these two languages, the possible cognates with the Tupi-Guarani
language family are limited. The prefix e- '2SG' occurs in Karo, Munduruku, Gavi o,
Surui, and Shipaya. There are also cognates of the second person singular pronoun
*ene: en in Karo, eet in Gavi o, en in Munduruku, een and eende in Surui, and ena in
Shipaya. It is also possible that the third person coreferential in Gavi o, a-, corresponds
to *o- (Set 3). Gabas (1994) summarizes that Karo is ergative, Surui is at least in part
ergative, and Gavi o and Munduruku are nominative-accusative.
Rodrigues (1985:380) has shown the cognates set out in Table 35. I have added
the Proto-Tupi-Guarani forms.
Table 35: Cognates in Tupian languages
Tupi-Guarani Tupi, non T-G
P-T-G Tupinamb Tupari Munduruku
1SG *wi- wi- w-, o- we-, ï-
2SG *e- e- e- â-
3 *i- i- ë- ß-
Tupari and Munduruku are both of Tupian stock, but are not Tupi-Guarani. The first
and second person prefixes correspond to the coreferential set (Set 3) in the Tupi-
Guarani system and the third person prefixes to those in Set 2. This shows also that
perhaps Set 3 prefixes originally had a broader scope (indicating more than corefer-
entiality) and that /- and t- may have been part of this set, being associated only later
with the personal pronouns and their reductions (Set 2). Rodrigues has shown that
Comparative Tupf-Guaranf Morphosyntax 575
these prefixes are cognates not only of Tupian languages, but also of Cariban
languages.
In sum, the languages of the Tupian stock are of little help to us in discovering
the original cross-referencing system for Pre-Proto-Tupi-Guarani except to say that
personal pronouns, their reductions in Set 2, and the person markers from Sets 1 and
4 apparently developed after the time that Pre-Proto-Tupi-Guarani separated from
most of the other Tupian languages. The development of the person hierarchy was
limited to the Tupi-Guarani languages, Aweti, and Satere-Mawe. It is of interest,
however, to consider that (previous to Pre-Proto-Tupi-Guarani) the */- and *t-,
prefixes which later indicated normal (not coreferential) third person, might have
originally been part of a set which included what later became coreferential markers
in Tupi-Guarani languages: *wi- and *e-.
that one semivowel (j) replaced another (w) in some languages by analogy with the
-i allomorph, which occurs with consonant-final stems. However, it is more difficult
to account for the discrepancy between the -n in some languages and -w in others. I
suggest that the discrepancy is evidence of the OBTOP construction being an older
form, and that the -w and the -n might even have come from different sources, both
of which were in the process of being eliminated.
Finally, this construction was eliminated in all of the languages described in C.
Jensen (1990) as having undergone cross-referencing changes: Mbyä Guarani,
Kaiwä, Chiriguano, Wayampi, and Urubu-Kaapor. As a result, except for a few
remnants, no independent verbs in these languages have absolutive cross-refer-
encing, except as part of the person hierarchy.
In sum, the unusual features of the OBTOP construction could be evidences that
it was in the process of being replaced, even at the time of Proto-Tupi-Guarani.
Where grammatical constraints do not require its use, the normal independent verb
forms occur. The OBTOP may well be a remnant of the original Pre-Proto-Tupi-
Guarani system for independent verbs. If this is the case, cross-referencing for
independent verbs was absolutive, as it is in dependent verbs, and there may have
been an independent verb suffix. The development of a person hierarchy in inde-
pendent verbs was a move away from absolutive cross-referencing, which is
consistent with further cross-referencing changes in Mbyä Guarani, Kaiwä,
Chiriguano, Wayampi, and Urubu-Kaapor.
Except for the third person prefix of Set 2, the person markers from Sets 1, 2, and
4 appear to be relatively recent, having cognates only in Aweti and Satere-Mawe.
The second person singular pronoun *ene has cognates in the Tupian stock. The
source of other personal pronouns (on which Set 2 person markers are based) is not
known.
Comparative Tupf-Guaranf Morphosyntax 577
NOTES
The loss of final consonants may not seem particularly significant, but it
causes the elimination of certain allomorphs of suffixes which only occur with
consonant-final stems, making these forms in other languages potentially
unrecognizable to speakers of a language such as Wayampi that lacks them.
The loss of nasalization could create difficulty for Wayampi hearers, because
of the substitution of oral for nasal morphemes, especially when homonyms
result from the loss. The anticipatory morpheme in Wayampi, rä, which has
no other allomorphs, shows little superficial resemblance to the Assurini
allomorph woma.
No data analysis of a comparative nature can be done without extensive
preliminary field work by many linguists in many languages. Work by An-
chieta (1595) on Tupinambä, by Ruiz de Montoya (1639, 1640) on Old
Guarani, and by Hoeller (1932) on Guarayu were done many years ago; other
works are relatively recent. A large variety of linguistic material has been
written on Tupi-Guarani languages: dictionaries, complete grammatical de-
scriptions, linguistic papers on some particular aspect of a language,
interlinear texts, etc. Each of these has contributed in some way to the
reconstruction of Tupi-Guarani morphology. Berts' (1981) meticulously pre-
pared dictionary of Parintintin has provided much important information for
the reconstruction of allomorphs, although the examples in context are few.
The "Formulario Tupi," designed by Harrison in 1977 for a Tupi workshop
and based on Rodrigues (1953), was filled out by linguists working with
Tocantins Assurini (Nicholson 1977), Guajajära (Harrison and Harrison
1977), Urubu-Kaapor (Kakumasu 1977), and Wayampi (C. Jensen 1981); the
completeness of the paradigms in various syntactic contexts has been a
valuable resource. (In that workshop Harrison planted the seeds of my passion
for comparative Tupi-Guarani work). Grammatical descriptions and interlin-
ear texts have been the main sources for the examples contained in this paper.
I wish to congratulate all my Tupi-Guarani colleagues for their valuable
contributions, and to thank them for the material which has made possible a
morphological reconstruction. Special thanks are due to La Vera Betts, Wolf
Dietrich, Rose Dobson, Carl Harrison, Jim Kakumasu, Yonne Leite, Dennis
Newton, Aryon Rodrigues, Lucy Seki, and Helga Weiss, for data or for
clarification of data furnished in personal communications; and to R. M. W
Dixon, for his suggestions and encouragement. Special thanks is due to Des
Derbyshire, whose skill as an editor I have come to highly respect. Fieldwork
in Brazil is conducted with the permission of FUNAI (Fundacäo Nacional do
Indio). A majority of the grammatical Pro to-Tupi-Guarani morphemes in-
cluded in this paper were reconstructed as part of my master's thesis,
completed at the UNICAMP (State University of Campinas) in 1984 under
the orientation of Aryon Rodrigues. The Working Conference for Amazonian
Languages, held at the University of Oregon in August, 1987, provided an
environment for further interaction with other Tupi-Guarani linguists and for
Comparative Tupf-Guaranf Morphosyntax 579
the form of clause-final clitic pronouns (Harrison 1986). At the sentence level,
these prefixes function on a nominative basis, since they are coreferential with
the grammatical subject (A or S).
22 In Urubu-Kaapor the symbol ÷ is used to refer to the phoneme /s/. This
phoneme was produced diachronically, palatalization of the phoneme *k when
adjacent to a high front vowel.
23 Example (267) is from *akarj 'head' + *'ok 'remove'. This is the result of
metathesis followed by the simplification of consonant clusters (Appendix
III.5-6).
24. The Kamaiura negation suffix -ite was probably formed by the combination
of the suffix *-/ 'NEG' and the particle te which occurs in various languages
and may be related to the morpheme -ete 'true, genuine'. Compare the free
response negative morpheme anite (Appendix 1.67) from *ani. The fact that
Seki (1978) recorded a vowel rather than a semivowel in such examples as
(387) does not necessarily preclude the formation of a diphthong with the
preceding vowel (Appendix III. 16).
25 I do not understand why different negation morphemes are used in examples
(407) and (423), unless kamajura in (407) has an adverbial function.
26 A change parallel to this hypothesis has occurred in Wayampi with the loss
of the consonant *c. A negative allomorph, na-, normally occurring only with
C-initial morphemes, is still used where the third person prefix has become
zero: *na-c-oryb-i > na-0-ory-i 'he is not happy'. One consonant likely to
have been eliminated at the beginning of stems would be the glottal stop,
which is rare in that position in Tupi-Guarani languages.
27 The system which occurs in independent clauses was referred to by Harrison
(1986) as a nominative-accusative system with a control/non-control intransi-
tive verb split and a person hierarchy split. In this paper I am describing it as
an active/inactive system, because the prefixes of the agentive IV correspond
with the A of the TV, while the prefixes of the nonagentive IV correspond
with the Ñ of the TV.
28 The / in t-ub-i and t-ur-i is a phonological irregularity of these two stems and
does not have any meaning. The /- and c- prefixes given in the various
protoforms are the third person prefixes from Set 2. The prefixes are not
always preserved in the remnant forms.
582 Jensen
ABBREVIATIONS
References
Nicholson, Velda
1977 Formulario padrao Tupf: Assurini, Archived at the Summer Institute of
Linguistics (Brasilia), Funda9ao Nacional do Indio (Brasilia), Museu
Nacional (Rio de Janeiro), and the Universidade Estadual de Campinas.
1978 Aspectos da Lingua Assurini (Brasilia: Summer Institute of Linguis-
tics).
Payne, Doris, (ed.)
1990 Amazonian linguistics (Austin: University of Texas Press).
1994 "The Tupi-Guarani inverse," Voice: Form and Function, edited by Barbara
Fox and Paul Hopper (Amsterdam: John Benjamins).
Priest, Perry N.
1987 "A contribution to comparative studies in the Guarani linguistic family,"
Language Sciences, Vol. 9, No. 1. 17-20.
Rodrigues, Aryon Dall'Igna
1953 "Morfologia do verbo Tupi," Letras (Curitiba), 1:121-152.
1981 "Estrutura da lingua Tupinamba," ms.
1984/1985 "Redoes internas na famflia linguistica Tupi-Guarani," Revista de
Antropologia (Säo Paulo), 27/28:33-53.
1985 "Evidence for Tupi-Carib Relationships," South American Indian lan-
guages: retrospect and prospect, edited by Harriet E. Manelis Klein and
Louisa R. Stark (Austin: University of Texas Press).
1990 "You and I = Neither you nor I: The personal system of Tupinamba
(Tupi-Guarani)," Amazonian linguistics, edited by Doris Payne (Austin:
University of Texas Press), 393-406.
Ruiz de Montoya, Antonio
1639 Tesoro de la lengua Guarani (Madrid: Juan Sanchez).
1640 Arte y Bocabulario de la lengua Guarani (Madrid: Juan Sanchez).
1892 Arte de la lengua Guarani [with commentaries, edited by Paulo Restivo]
(Stuttgart: Guilielmi Kohlhammer).
Seki, Lucy
1978 "Negasao em Kamaiura," Paper presented at the Universidade Estadual de
Campinas.
1987 "Aspects of Kamaiura morphosyntax," Paper presented at the Working
Conference on Amazonian Languages, Eugene, Oregon.
1989 "Sistema de marca9äo de caso do Kamaiura," Paper presented at the
Seminärio de Tipologia, Universidade Estadual de Campinas.
1990 "Kamaiura (Tupi-Guarani) as an active-stative language," Amazonian
linguistics, edited by Doris Payne (Austin: University of Texas Press),
367-92.
Comparative Tupf-Guarani Morphosyntax 587
Taylor, John
1984a "A interroga9ao na lingua Kaiwä," Estudos sobre linguas tupi do
Brasil, Serie linguistica Vol.11 (Brasilia: Summer Institute of Linguis-
tics), 123-56.
1984b "Marcafao temporal na lingua Kaiwä," Estudos sobre linguas tupi do
Brasil, Serie linguistica Vol. 11 (Brasilia: Summer Institute of Linguis-
tics), 37-122.
Taylor, John and Audrey Taylor
1966 Statement of Kaiwa grammar from clause to morpheme level, Archived
at the Summer Institute of Linguistics (Brasilia), the Fundagao Na-
cional do fndio (Brasilia), and the Museu Nacional (Rio de Janeiro).
Weiss, Helga
1972 Kayabi verbs, Archived at the Summer Institute of Linguistics (Braslia),
the Fundagao Nacional do indio (Brasilia), and the Universidade
Estadual de Campinas.
1989 "Notes on ideophones and onomatopoeia in Kayabi," Revue de
Phonetique Appliquee, Occasional publication of the Universite de
1'Etat, Mons, Belgique, 389-97.
588 Jemen
Appendix I
Reconstructions
The majority of morphemes in this list were reconstructed by the author of this
paper. Morphemes 1-6, 12-15, 18-23, 25-38, 43-47 were reconstructed as part of
my master's thesis in 1984 (C. Jensen 1989), under the orientation of Aryon Ro-
drigues. Morphemes 39-40, 48-56, 58-59 were reconstructed during the same
period but were not included in the thesis. The reconstructions of morphemes 16-17,
22 are revisions of forms reconstructed in my thesis and appear in C. Jensen (1990).
Morphemes 24, 41-42, 68-72, 74, and 76-77 were reconstructed for this paper.
Morphemes 8, 10, 57, 73, 75, and 78—80 were reconstructed by Miriam Lemle
(1971); morphemes 7, 9, and 11 by Aryon Rodrigues (p.c.); and morphemes 63—67
by Allen Jensen (1993). Orthographical symbols have been standardized for the
reconstruction. The citing of accented and non-accented forms does not necessarily
mean that the non-accented (orthographically unmarked) forms are phonetically
unaccented. It means simply that no accent was recorded. In cases where a mor-
pheme has different forms, depending on whether it is at the word boundary or
whether another affix occurs between it and the word boundary, this is indicated by
alternating forms, one with two hyphens.
Person markers (Set 1)
1. *a- '1SG Set 1'
a- (As,Ch,Gi,Gj,Gu,Kb,Km,Kw,Pt,Tb,Ur,
Wa)
a- (Tp)
2. *ere- '2SG Set 1'
ere- (As,Gi,Gj,Gu,Kb,Km,Kw,Pt,Tb,Tp,Ur,
WaJ)
re- (Ch)
ne- ~ -ere- (WaA)
3. *oro-'lEXSetl'
oro- (As,Gi,Gu,Kb,Km,Kw,Pt,Tb,Wa)
ro- (Ch)
uru- (Gj)
ara- (Tp)
eliminated in Ur
Comparative Tupf-Guaranf Morphosyntax 589
4. *pe-'2PLSet l'
pe- (As)Ch,GiM,Gj,Gu,Km,Kw,Pt,Tb,Tp,
Ur,Wa)
pe- (Kb)
5. *ja- ºÉÍ Set 1'
ja- (Ch,GiM,Gu,Km,Kw,Pt,Tb,Ur,Wa)
sa- (As)
za- (Gj)
c - (Tp)
sa- (Kb.WaJ) derived from t-ja-
ÏÃÏ- (As,GiM)
ara- (Tp)
apo- 41SG A with 2PL P' (Kw)
apu-'1SG A with 2PL P' (Gj)
apa- ' 1SG A with 2PL P' (Tp)
oropo- '1PL A with 2PL P' (Kw)
urupu- '1PL A with 2PL P' (GjP)
eliminated in Kb,Ur
Dependent verb suffixes
21. *-abo ta ~ -a 'dependent serial verb'
-afeo
-afeo ~ -abo (GiO,Pt,Tb)
-awu (Kb)
-wo (Tp)
-o (Gu)
-ta
-ta (As,GiO,Gj,Kb,Pt,Wa)
-a
-a (As,GiO,Kb,Pt,Tb,Tp)
-pa (Gj)
-m (Km)
eliminated except for remnants in Ch,GiM,Kw,Ur,Wa
22. *-i ~ (undetermined C) Oblique-topicalized'
-i (As,GiO,Kb,Km,Pt,Tb,Tp)
-j (As,Kb)
-w (Kb,m,Tb)
-n (Gj,Ur remnants)
eliminated except for remnants in Ch,GiM,Gu,Kw,Ur,Wa
23. *-VmV rVmV 'temporal/conditional (when/if)'
-erne reme (Tb)
reme (Wa)
mehe (Gj)
-amu ramu (Kb)
-amo ramo (As)
-ramo (GiM,Kw,Tp)
-amo ramo (GiO)
-ame rame (Pt)
-raha (Ur)
Comparative Tupf-Guaranf Morphosyntax 593
Nominalizations
30. *-ar ~ -car tar 'agent'
"-ar
-ar, -ar (As,Gj,Gu,Pt,Tb,WaA)
-at ar- (Kb,Km)
-ar a (GiM,GiO,WaJ)
/w
-ar (Tp)
-a (Ch)
*-car
-car (Gu)
-sar (Tb)
-har (Gj,Pt,Ur)
-har- ~ -ha (GiO,Kw)
-ar- ~ -a (WaJ)
-ar (WaA)
*-tar
-tar, -tar (As,Tb)
-tat ~ -tar- (Km)
-tar- —ta (GiO)
31. *-ab ~ -cab ~ -tab 'circumstance'
*-ab
-ab (Pt,Tb)
-ap ~ -aw- (Kb,Km)
-aw (As,Gj)
-äw (Tp)
-aw- ~ -a (Ch,GiM,Wa)
-aw- (Gu)
*-cab
-sab (Tb)
-ca (Gu)
-hab (Pt)
-haw (Gj)
-hap (Ur)
-ha haw- (Kw)
-aw— -a (Wa)
Comparative Tupf-Guaranf Morphosyntax 595
*-tab
-tab (Pt,Tb)
-tap taw- (Km)
-taw (As)
-ta (Wa)
In some languages this, rather than morpheme 43, refers to the action or
quality of a verb.
32. *-pyr 'patient'
-pyr, -pyr (Gj,Gu,Pt,Tb)
-pyt ~ -pyr- (Kb,Km,Tp)
- ~ -pyr- (GiO)
-py (GiM)
eliminated in Ur,Wa
33. *emi- Object'
emi- (As,Ch,GiM,GiO,Gj,Gu,Kb,Km,Pt,Tb,
Tp.Wa)
34. *-ba'e 'clause nominalizer'
-ba'e, -ba'e (GiM,GiO,Gu,Kw,Tb)
-bae (Ch)
-be'e (Pt)
-wa'e (As)
-ma'e (Gj,Kb,Km,Tp)
ma'e (Wa)
me'e (Ur)
35. *-cwär, *nwär 'nominalizer of adverb or postpositional phrase'
-swar, -nwär (Tb)
-cwa (GiO)
-war (As,WaA)
-wat war- (Kb)
-wa (Ch,GiM,WaJ)
-nar (Gu)
Unspecified possessor prefix
36. *0 nasalization of initial C ~ t elimination of initial V 'Unspecified
possessor'
0- (Ch,GiO,GiM,Gu,Pt,Tb,Tp,Wa)
nasalization of #C (Ch,GiO,GiM,Gj,Gu,Kb,Pt,Tb,Tp,Wa)
596 Jensen
t- (Ch,GiO,GiM,Gj,Gu,Kb,Pt,Tb,Tp,Wa)
elimination of #V (Ch,GiO,Gj,Gu,Pt,Tb,Tp,Wa)
Modificational suffixes
37. *-wacu ucu 'augmentative'
-wacu ~ -ucu (GiO, Gu) In Gu -ucu only appears in
some old forms.
-wasu ~ -usu (Tb)
-wahu uhu, -hu (Ur)
-hu uhu, -hü uhü (Pt)
-oho (As)
-wacu (GiM)
-wasu (Ch,Kw)
-wasu, -u (WaJ)
-wasu ~ -u, -usu (WaA)
-wahu hu (Gj)
-uu (Kb)
-(0)0 (Tp)
38. *-'i 'diminutive'
-'i, -' (GiM,Pt,Tb)
-'i (As,Tp,Wa)
-a'i (Gj)
-i'i ~ -'i (Kb)
39. *-pwer ~ -wer 'devolved, former'
-pwer ~ -wer (Tb)
-kwer wer (As)
-kwer er rjwer wer (Gj,WaA)
-kwe — (n)e —(r)e rjwe —we (GiMJCw.WaJ)
-kwer ~ -rjwer (Ur)
-kwer ~ -rjwer ~ -ruer ~ -awer (Pt)
40. *-ram ~ -warn ~ -am 'anticipatory, future'
-ram ~ -warn ~ -am (Tb)
-ruam arjwam ~ -aruam (Pt)
ra (Ch,Gi,Gu,Wa)
-warn (Gj)
-rom ~ -worn (As)
41. *-ete 'true, genuine'
-ete or -etc (As,GiM,Gj,Gu,Km,Tb,Wa)
-ete (Ch)
Comparative Tupf-Guaranf Morphosyntax 597
74. *wi or *wirj 'that one (visible), which may be far from the hearer also'
wi, wig (Tb)
wyi, äwi (Wa)
rjwi (Pt)
75. *mö, *amö 'there, another, some (invisible)' [Lemle]
mo, amö (GiM,Tb,Wa)
mo, amua (Pt)
omo (GjP)
amu (Kb)
amo (Ur)
76. *pe 'that one'
pe (GiM,Gj,Pt)
pebe (Gu)
pea (Ch)
upea (Kw)
peme'e (Ur)
77. *wa 'they, those'
wä, awä (Tb)
wä (Kb)
wä (Gj)
Numbers
78. *ojepetei One' [Lemle]
ojepe (Tb)
osepesowe (As)
pitei (Gj)
ojipeji (Pt)
ojepete (Km)
pete'i (Ur)
jepei (Gu)
pe'i (Wa)
peti (Ch)
petei (GiM)
79. *moköj 'two' [Lemle]
moköj (Gi,Km,Pt,Tb,Ur,Wa)
mokoi (As)
mokoi (Ch)
mukuz (Gj)
Comparative Tupf-Guarani Morphosyntax 603
Appendix Ð
Orthographical references:
Vowels (oral and nasal sets)
*i *y [i] *u
*e *a *o
Consonants
*p *t *k *' [?]
*c [ts] *c [ts]
*m *n *rj
*fe *r
Semivowels
*w *j [y]
Orthographic conventions:
The high central vowel *[i] is written orthographically as y.
The palatal semivowel *[y] is written as j.
The glottal stop *[?] is written with an apostrophe '.
Vocalic changes:
*a > (Tp)
*o > a (Tp)
*a > a (voiceless mid central vowel) (Gj)
Some *o > u (Gj.Ur)
Some *o > a (As)
*u > ï (As)
loss of nasalization (As,Gj,Tp)
Comparative Tupf-Guarani Morphosyntax 605
Consonantal changes:
*P
*pu, *pw > f (Kb)
*pw > kw (As,Gj,Tp,Wa)
*PJ > Pi tPJ] (Gu), c (GiM), [c] (Tp), c [c] (Km), s (As,Gj,Kb,Ur,Wa)
*t
*ti > ti (Tb,Pt), ci (Gu), ci [ci] (Km), xi [ci] (GiM), si (As,Gj,Ur,Wa)
*k
ik > ÷ [s] (Ur)
*k# > g [rj] (As, some Wa)
some k > g [g] medially (Kw), finally (Kb)
written as v in (GiM,Gu,Kw,Pt,WaA)
*b# > p (Kb,Km)
*b > w (As,Gj,Kb,Ur,WaJ)
*m, *n, *rj
In some languages (Ch,Gi,Gu,Kw,Pt,Ur) the oral allophones are written
separately as mb, nd, ng.
rj written as g (As,Gj,Wa)
rj written as g (Kb, Pt)
*c, *c
*c [ts] > ÷ [c] (GiM), [c] Tp, s (As,Gu,Tb), h (As,Gj,Ur), 0 (Kb,Tp,Wa)
*c [ts] > s (Tb), h (As,Gj,Kw,Ur), 0 (GiM,Kb,Tp,Wa)
c written as s (Gu)
*w
*w > gw (Kw), gu (GiO,GiM)
Sometimes written as (Kb)
*j pronounced in various languages as [y], [z], or [dz]
*j > z (Gj, Tembe)
*j > s (As)
*j > x [o] (Tp)
In some languages the nasalized allophone [n] is written separately, as nh
(GiHKw,Pt).
Syllable final, j is written in some languages as i or as i (Kb).
606 Jensen
*Ã
*ô# > t (Kb, Km)
Some *r# > n (As)
Comparative Tupi-Guaranf Morphosyntax 607
Appendix IQ
syllabic
consonantal
voiced
nasal
continuant
posterior
labial
high
syllabic + 4. +
nasal
posterior
labial
high
ï- 1
+ cons [ + nas]/l+nas] -fsyl + syl à -syl
[-voice L
-cons -syl -nas [-nas
-syl J + voice. )
In Tupinamb this rule applied when there was a consonant cluster produced at
morpheme boundary by the combination of two morphemes. The rule is based on
the interpretation that the nasal accent did not co-occur with a final nasal consonant.
Accordingly, this rule is followed by another (6) which simplifies consonant clusters.
> akaq * meb > akaij meb > (rule 6) 'flat head'
3. Epenthesis of/y/
+ syl
0- + high / + cons + [ + cons]
+ post -lab
-lab
According to this rule an epenthetical vowel y [i] is inserted between two conso-
nants at a morpheme juncture of the type stem + affix.
'ar 'day' + -bo 'diffuse locative' > 'arybo 'in the day'
4. Weakening of labial consonants
+ cons
cons >[ + cont] + voic
lab _ —nas J
Comparative Tupf-Guaranf Morphosyntax 609
'ab * puku > 'ab * buku > 'long hair' (to rule 6)
kuab * me'er) > kuab * be'ey > 'show, give knowledge' (to rule 6)
okar * pyter > okar * byter > 'the middle of the plaza' (to rule 6)
5. Metathesis
Note that this rule from C. Jensen (1989) replaces Rodrigues' rule 6.
— syl
Structural description: [ + cons] — cons
— voic .
2
According to this rule, when the suffix of the dependent serial verb -abo or the
nominalizer -ab follows a low vowel, the first vowel of the suffix is assimilated to
the position of the preceding vowel. This happens only occasionally with the nomi-
nalizer as the normal allomorph following vowels is -sab in Tupinamba.
o+ so +abo > o+50 + obo > 'and went' (to rule 8)
o+ man + abo > o + mano + obo > 'and died' (to rule 8)
s + e'e + abo > s + e'e + ebo 'and scraped it' (to rule 10)
610 Jensen
-fcons -fsyl
+ voic nas] + acc syl
—nas +nas acc
> s + e'e + ebo > s + e'e + ebo > 'and scraped it' (to rule 12)
apiti+äbo > apiti+abo > 'and killed it' (to rule 11)
> + manö + omo > o+mano + omo > 'and died' (to rule 12)
u 'eat' + ar 'NOM' > 'u + ar > 'eater' (to rule 11)
11. Asyllabification
+ syl > [-syl] / + syl
+ high •face
> ero 'CC' + sem 'come out' > enosem 'come out with it'
14. Insertion of semivowel /j/
— cons + syl + syl
0 — syl / + high + high
4- voic . — post . . -lab .
-lab
In Tupinamba, the semivowel j is inserted between a high front vowel and a high
front or central vowel.
i '3' + ita 'rock' > ijita 'his rock'
/ '3' + ypy 'beginning' > ijypy 'its beginning'
15. Epenthesis of/i/
+ syl + cons
4-high cons] — voic
. — post. . + cont.
16. Diphthongization
+ syl — cons
4- high -* — syl
— post + voic
— ace -lab
In Tupinamb , a vowel cluster in which the second of two vowels is a high front
vowel (i) results in a diphthong. This occurs with the partitive locative suffix,
negation of verbs, incorporation of object prefix, and the addition of a prefix to a
stem beginning with a high front vowel.
ku 'a 'waist' + / > ku 'aj 'at the waist'
n 'NEC' + a 'ISO* + karu 'eat' + / 'NEC' > nakaruj º didn't eat.'
a 'ISO* + i '3P' + potar 'want' > ajpotar º want it.'
ï '3' + iko 'be' > ojko 'he is'
17. Absorption of /i/
+ syl — cons
+ high — syl
— 0 1
. — post . . + voic .
In Tupinamb , there is optional insertion of/w/ between two vowels when the first
is a back vowel.
ï '3 COREF' + ete 'body' > oete or owete 'his own body'
ï '3 COREF' + obake 'in front of > oobakÝ or owobaka
'in front of himself
kuab 'know' > kuab or kuwab 'know'
19. Devoicing of final consonants (1.3 in Rodrigues)
Comparative Tupf-Guaranf Morphosyntax 613
Appendix IV
Rodrigues, 1984/1985,
with additions by C. Jensen
Subgroup I
1. loss of final consonants
2. *c and *c retained as separate phonemes
*c > [c] or [s]
*c > h or 0
3. reinterpretation of sequences as single phonemes *pj, *pw, *kw [C. Jensen]
4. elimination of consonant sequences [C. Jensen]
5. merger of *pw and *kw as [k ] (or &)
6. *pj > c or z
7. in Chiriguano, change of stress from final to penultimate syllable
Subgroup Ð
1. loss of final consonants
2. merger of *c and *c > c or s
3. merger of *pw and *kw > k\v or k
4. *pj remains intact
5. change of stress from final to penultimate syllable
6. palatalization of *tl / [C. Jensen]
Subgroup ÉÐ
1. retained final consonants
2. merger of *c and *c > c or s
3. *pw remains intact
4. *pj remains intact
5. stress unchanged
Subgroup IV
1. final C retained with or without modifications
2. merger of *c and *c > h
3. merger of *pw and *kw > kw [kw]
4. *pj > c or c
Comparative Tupf-Guaranf Morphosyntax 615
5. *j > c, c, s, or z
6. merger of 4 and 5 in some languages
7. merger of *b and *w with possible reinterpretation of final allomorphs
[C. Jensen]
8. palatalization of *f/ i [C. Jensen]
Subgroup V
1. retained final consonants
2. merger of *c and *c > h or 0
3. *pw > /(bilabial)
4. *pj > s
5. *j expressed as affricate
6. merger of *b and *vv with reinterpretation of final allomorphs
[C. Jensen]
7. pronominal markers for third person include masculine, feminine, and plural
8. palatalization of *?/ i [C. Jensen]
Subgroup VI
1. retained final consonants
2. merger of *c and *c > h
3. *pw > kw (Parintintih), v, or/(Tupi-Kawahib)
4. *pj remains intact
5. *j remains intact
6. pronominal markers for third person include masculine, feminine, and plural
7. palatalization of *tl i [C. Jensen]
Subgroup ÕÐ
1. retained final consonants
2. merger of *c and *c > h or 0
3. *pw > hw or h, retained as separate from *kw
4. *pj > s
5. *j intact
6. merger of *b and *vv with reinterpretation of final allomorphs
[C. Jensen]
7. palatalization of *// i [C. Jensen]
Subgroup ÕÐ!
1. loss - partial or complete - of final consonants
2. merger of *c and *c > h or 0
3. merger of *pw and *kw > kw [kw]
4. *pj > s
616 Jensen
5. *j intact
6. in most languages, merger of *b and *w [C. Jensen]
7. palatalization of *</ /
Comparative Tupf-Guarani Morphosyntax 617
Appendix V
Rodrigues 1984/85
peb 'low, flat', moköj 'two', mano 'die', me'er) 'give',ye'e;; 'speak', apo 'make', ata
'walk', -epjak 'see', ma'e 'look'.
9. The word petym (and not pe) 'tobacco' (literally, 'planted tobacco').
CUMULATIVE INDEX TO VOLUMES 1-4
This index incorporates and supersedes the one published in volume 3. Coverage has been
extended and improved, and many new subjects have been indexed for all four volumes.
Bold italic numerals followed by a colon refer to volumes; plain Arabic numerals refer to
pages; boldface page ranges in an entry indicate whole chapters devoted to that topic or written by
that person. Thus '7:viii, 15, 33-127; 2:73-9' means, 'in volume 1, pages viii and 15 and the
entire chapter in pages 33 to 127 inclusive; in volume 2, pages 73 to 79 inclusive'. For very gen-
eral topics like 'verbs', references to the general topic are first listed by volume ('3:80, 127-9,
185, ...') and then as subclassified by language (e.g., '(Guajajara) 7:410, 43Ã) and by subtype
('benefactive verbs, 1.297, 585, 595').
The different spelling conventions of English, Spanish, and Portuguese give rise to many ortho-
graphic variations for names of languages (and tribal groups). Systematic regularization has not
been attempted here, and alternate spellings for names of languages are not covered exhaustively.
Some nonstandard spellings are included where difficulties in identifying languages might arise,
but in addition it will be helpful to bear in mind the following alternations when searching for a
language name:
b~ v c~k c~s ch~tx g-j
gu~hu~w j~h x~sh y~j z~s
A purely English case-insensitive alphabetical ordering is used: accents on vowel letters are
ignored for alphabetization, so ò is equated with c and ç with n. The prefixes "Me" and "St." are
alphabetized literally (not equated with "Mac" and "Saint").
(Sanuma)2:154-64; (Canela-Kraho)7:185-6;
(Unibu-Kaapor) 1:382-384; (Guajajara) 7:416;
(WaiWai)4:45; (Paumari) 3:12;
(Yagua) 7:448-51; 2:288-9,291-2, (Wai Wai) 4:81,120-23,177-82;
310-11, 337, 337-9, 340, 378-83 (Warekena) 357-66;
adverbs and adverbiale, 7:48-9, 65, 87, 111, (Yagua) 2:253-9, 361-7
209-10,578,600-1; 2:37,204-7; Aguaruna language, 7:458,463
3:88-90, 97-8, 107-9,130-2; 4:35-6, Aikhenvald, Alexandra, 4:5, 11, 225-440
66-8, 79, 125-9, 183-5, 199-212, 396-8; Aimore" people, 7:6, 7
(Apalai) 7:48-9,65, 87,111; Akawaio language, 3:9, 23
(Arawakan) 7:578, 600-1; Akewere (= Tocantins Suru0 language, 4:495
(Macushi) 3:88-90, 97-8,107-9, Algonkian languages, 7:425
130-32; Alicea, Neftali, 7:463
(Pirahä)7:209-10; Allin, Trevor R., 3:359, 362
(PreAndine Arawakan) 7:578, 600-1; Almeida, Francisco Antonio de, 7:4
(Sanuma) 2:37, 204-7; Almeida Serra, Ricardo Franco de, 7:7
(Wai Wai) 4:47-8, 79,125-8, 129, Amage (= Amuesha) language, 7:12
183-5, 199-212; Amahuaca language, 4: viii, 8,12, 14, 443-85
(Warekena)4:396-8 Amahuaca people, 4:444
adverb phrases, 7:188, 302-3, 396; Amanay6 language, 4:495
2:210-11,417; 5:129-32,338; Amapä (territory, French Guiana), 7:471
4.396-8 Amapari dialect of Wayampi, 4:508, 512,
(Canela-Krahö)7:188; 514, 534-5, 537, 544, 552
(Macushi) 3.129-32; Amarakaeri language, 3:365
(Paumari) 3:338; Amazon (river), 7:14-15,440,472;
(Pirahä)7:302-3; 2:251-2
(Sanuma) 2:210-11; Amazonas (state, Brazil), 3:161, 200,
(Urubu-Kaapor)7:396; 471-2; 4:491
(WaiWai) 4: 35-6, 125-8; Amazonia (region), 7:1-10; 4: 5-8
(Warekena)4:396-8; Amazonian Indian art, literature, science, etc.,
(Yagua) 2:417 4:6
adverbial clauses, 7:262-8; 2:324-6; Amele language, 4:478
3:69-70, 73-82, 225-35; 4.77-80; Amerind languages, 3:13
(Macushi) 3:69-70, 73-82; Amisha (= Amuesha) language, 7:12
(Paumari) 3:225-35, 320; Amoishe (= Amuesha) language, 7:12
(Pirahä) 7:262-8; Amorim, Rute, 4:424
(Wai Wai) 4:77-80; Amsterdam (Holland), 3:355
(Yagua) 2: 324-6 Amuescha (= Amuesha) language, 7:12
Africa, 4:443 Amuesha language, 7:12-13, 505, 536, 559,
agency hierarchy 7:416,420-1,424 567-642 passim; 3:13, 355-499 passim
agglutination, in Maipuran 3:376 Amuesha people, 7:462
agreement, 7:185-6, 416, 513-8, 532-5; Amuexa (= Amuesha), 7:12
2:253-9,361-7; 3:12; 4:11,13,27,81, Anamb6 language, 4:495
120-23,177-82,357-66; anaphora, 7:68-72, 163-4, 217-8, 255-62,
(Arawakan) 7:513-8, 532-3, 534-5; 358-63; 2:89,330-4; 3:61-7,223-5;
Index 621
location markers, (Sanuma) 2:75,165, 4:8, 11, 13, 225, 228, 240, 320, 334, 369,
202-4; (Yagua)2:411-12 380, 386,419,422,423
locative phrases and clauses, 1:208, 286-7, Maipure language, 4:422
578,601; 2:161-3,209-10,338-9; Maku language, 4:386
3:112,253-4,315-7,380; 4:282-3 Makusi = Makusi, see Macushi
(Macushi) 3:112; Malaya (nation), 2:4
(Maipuran) 3:380; Mallinson, Graham, 1:455,474
(Paumari)3:253-4,315-7; Manaus (city, Brazil), /: 2, 4, 280; 3:6
(Piraha) /: 208, 286-7; Mandahuaca language, 3:433
(PreAndine Arawakan) /: 578, 601; Maniba language, 3:366
(Sanuma) 2:161-3,209-10; Mansen, Karis, 3:359
(Yagua) 2: 338-9 Mansen, Richard, 3:359
Lokono language, 3:13, 355-499 passim Mantey, Julius R.,/: 247
Longacre, Robert E., 4:468,479 maps of South America, /: vi; 2: vi;
Loos, Eugene, 2: 384; 4.12 3:viii,361; 4: viii, 226,492
Lorenzo (= Amuesha) language, 7:12 Mapuera river (Brazil), 4:9, 25
Loukotka, Cestmir, 1:8, 11 -14; 2: 252; Maquiritare language, 2:17
3:360, 363 Maracacume (river, Brazil), /: 326
Lovejoy, Thomas E., 3:5 Marajo (island), /: 3
Lowe, Ivan, 3:162 Maranhäo (state, Brazil), 1:128, 437; 4:495
Lutzenberger, Jos6,3:7 Martins, Valteir and Sylvana, 4:424
Lyons, John, 2:385 Masamae language, 2:252
Mashco-Piro languages, 3:362
Macahuachi, Mamerto, 2:459, 468 Mason, John Alden, /: 557; 3:360
Machiguenga language, /: 567-642 passim; Matacuni (river, Venuzuela), 2:17
3:13, 355-499 passim Matisoff, James, 2:440; 3:13
Macro-Carib languages, /: 12; 2:252; 3:13 Mato Grosso do Sul (state, Brazil), /: 472
Macro-Chibchan languages, 1:12, 200 Mato Grosso (state, Brazil), 1: 8, 13,472
Macro-Ge languages, 1:12, 15 Matsos language, 4:478
Macro-Ge-Pano-Carib languages, /: 16 Matteson, Esther, 3:358-9, 365-6, 368-71,
Macro-Panoan languages, 1:12 375, 379-80, 433,457,459, 472, 504, 513,
Macro-Tucanoan languages, 1:12-13 515-16, 556-7, 573, 576, 581, 587,
Macu language(s), 1:13 590-1, 593-4, 599-600, 637
Macushi language, /: 23; 3:4, 9-10, Mawayana people, 4:25
23-160; 4:6,9,14-15 Mayoruna, see Matsos
Madagascar (nation), 7:17 Mbaya-Guaicuru language, 1:13
Madeira (river, Brazil), /: 4 Mbyä (Brazilian Guarani) language, 4:13,
Madiga language, 3:11 491, 495, 501, 507, 508, 516, 518, 541-2,
Madre de Dios (area, Peru), 3:362 549, 553, 556, 562-5 572,576, 579
Maia, Marcus, 4:5 McCarthy, John,/: 321
Maici (river, Brazil), /: 200 Meira, SeYgio, 4:4, 5
Maicuru (river, Brazil), 1. 33 Mearim (river, Brazil), 1:437
Maiongong language, 2:17 Meggers, Betty J., /: 1, 3-4
Maipuran languages, 1:12-13, 469-642 Mello, Fernando Collor de, 3:7, 8
passim; 3:9-10,12-14,322,355-499; Mendes, Chico, 3:7
Index 633
Mendes, Francisco, /: 7 (Apalai)7:123-4;
Merrifield, William, 3: v (Paumari) 3: 349;
Mexico (nation), /: 17 (Pirahä) 7:317-21;
Migliazza, Ernest C, 2:7,17 (Sanuma) 2:173-81, 226-7;
Ministry of the Interior (Brazil), 2:6 (Wai Wai) 4:156-65;
Miranda (city, Brazil), 1:13,472 (Warekena) 4:416-19
Mishara language, 2:252 Mosonyi, Esteban, 3: 355, 366, 375
Missäo Evangelica da Amazonia, 2:15; 3:23 Mosonyi, Jorge, 3: 355, 359, 362, 375, 378
MIT (= Massachusetts Institute of Tech- Moynahan, Brian, 7:9
nology), 4.5 Mozombite, Laureano, 2:468
Mojo language, 4:422 Mucajai (river, Brazil), 2:17
Mond6 language, 1:13 Mucatyurirya, Estela, 2:459
mood and modality, 7: 100,106-7,184-5, Munduruku language, 7:414,415, 425, 426;
246-7, 295-6, 389-90, 522-3, 529, 613, 2:9; 4:574
615; 2:195-7,314-6,385-6,400-3; Mufioz, Juan, 3: 359
3:122-3,282-3,285-6; 4:352-1; Munro, Pamela, 7:461, 463; 2:277,459;
(Apalai) 7:100, 106-71; 4:468,470
(Arawakan) 1:522-3, 529, 613, 615; Mura language (dialect of Pirahä), 7:200
(Canela-Krahö) /: 184-5; Muran language family, 4:3,15
(Greek) 7:295; Mura-Piraha language, 7:200
(Latin) 7:295; Murinyapata language, 7:425
(Macushi) 5:122-3; Museu Emflio Goeldi (Belom, Brazil), 4:4
(Paumar03:282-6; Museu Nacional, Rio de Janeiro (Brazil),
(Pirahä) 7:246-7, 295-7; 7:199,437; 4:5
(PreAndine Arawakan) 7:613, 615; Muysken, Pieter, 7:288
(Sanuma)2:195-7;
(Urubu-Kaapor) 7:389-90; Nadeb language, 7:13, 16; 2:8-9; 3:4
(Warekena) 4:352-7; Na-Dene languages, 3: 13
(Yagua) 2:314-6, 385-6,400-3 Nadobo (= Nadöbö) language, see Nadeb
Moore, Dennis A., 7:10, 13; 2:9; 4:4 Nambicuaran languages, 7:12
Morique language, 3:374 Nambikuara language, 7:12,16; 2:8-9
morphology, N "ndeva language, 4:495
(Arawakan) 7:469-566, 504-11, Napo (river, Peru and Ecuador), 7:4,440;
569-604; 2:252
(Arawakan) 7:500-537; National Museum of Rio de Janeiro (Brazil),
(Canela-Kraho) 7:192-7; 7:199,437; 4:5
(Macushi) 3:9; National Science Foundation (NSF), 7:114;
(Maipuran)3:379-80; 3:vi,355; 4:579
(Pirahä) 7:322-3; Nazaro (town, Brazil), 4:225
(PreAndine Arawakan) 7:569-604; negatives and negation, 7:62,64-5,64-7,
(Tupi-Guarani)4:491; 65-7, 67, 89-90, 149, 160-3,161, 234,
(Urubu-Kaapor) 7:385; 235, 249-55, 357-8, 520-4, 615;
(Wai Wai) 4: 165-214; 2:78-89,317-30; 3:34-5,55-61,80,
(Warekena) 4:419-20; 218, 220-23; 4:91, 95, 195, 201, 211-12,
(Yagua) 2:445-56 morphophonology, 264-9, 324, 385, 388-9, 399, 494, 545-9,
634 Index
600; 594-5;
(Apalai) 7:64-7, 89-90; (Sanuma) 2:142-4;
(Arawakan) 7:520-4, 615; (Urubu-Kaapor) 7:376-8;
(Canela-Krahö) /: 149, 160-3; (Wai Wai) 4:49, 76-7, 88-96,129,183,
(Macushi) 3:34-5, 55-61, 80; 185-6;
(Paumari) 3:218, 220-3; (Warekena) 319-22;
(Pirahä) 1:234-5, 249-55; (Yagua)2:280-l, 291-2, 335-7, 349,
(PreAndine Arawakan) /: 615; 352-360,449-50
(Proto-Tupi-Guarani) 4:494, 545-9, 600; Northern Carib languages, 3:23
(Sanuma)2:78-89; Northern Foothills (region), 3:3
(Urubu-Kaapor) 7:357-8; Northern Maipuran languages, 4:11, 225, 228,
(WaiWai)4:91,95,201,211; 240, 320, 369, 380,419,423
(Warekena) 4:264-9, 324, 385; nouns, 7:230, 369-72, 371, 529, 574-78;
(Yagua)2:317-30 2:128-9,144-9,156-7,351-2,
Negro (river, Brazil), /: 4; 4:225, 228, 386 445-452; 3:88,254-9,307-8; 4:80,
Newton, Dennis, 4:578 85,107, 173,187-98, 217,493, 503-14;
National Endowment for the Humanities (Arawakan) 7:529, 574-78;
(NEH), 3: vi, 355; 4:579 (Macushi) 3: 88;
New Guinea (nation), 1:11,4:443 (Paumari) 3:254-9, 307-8;
Nhamundä river (Brazil), 4: vi, 25 (Pirahä) 7:230;
Nheengatu (Lingua Geral Amazonica) creole (PreAndine Arawakan) 7:574-78;
language, 1.280; 4:11, 225, 228, 330, (Sanuma) 2:128-9, 144-9,156-7;
495,497 (Tupi-Guarani) 4:503-14;
Nies, Joyce, 3:359 (Urubu-Kaapor) 7:369-72;
Nijamvo language, 2:252 (Wai Wai) 4:80, 85,107,173,187-98,
Nimuendaju, Kurt, 1:4, 128, 280 217;
Ninarn language, 2:7,17 (Yagua)2:351-2,445-452
Noble, G. Kingsley, /: 16,472, 515; 3:360, noun phrases, 7:84-100, 167-74, 271-2,
363,365-8,371,375,390 368-72,411-2,607-8,84-100;
Nomatsiguenga language, /: 567-642 passim 2:118-44, 348-360, 34-7; 3:83-99,
nominalizations, 1:55, 73-4, 89-98, 163, 250-67; 4:80-96,287-322,493,513;
172-4, 251, 262-3, 277-80, 376-8, 536, (Apalai) 7:84-100, 607-8;
542-5; 2:142-4,280-1,291-2,335-7, (Canela-Kraho) 7:167-74;
349, 352-360, 449-50; 3:68-72, 89-90, (Guajajara)7:411-2;
92-9, 240, 263-7, 379; 4:12, 29,49, (Macushi) 3: 83-99;
76-7, 88-96,129, 183, 185-6, 319-22, (Paumari) 3:250-67;
494, 539-45,563, 594-5; (Pirahä) 7:271-2;
(Apalai) 7:55, 73-4, 89-98; (PreAndine Arawakan) 7:607-8;
(Arawakan) 7:536, 542-5; (Sanuma) 2:34-7,118-44;
(Canela-Kraho) 7:163, 172-4; (Tupf-Guarani)4:513;
(Macushi) 3:68-72, 89-90, 92-9; (Urubu-Kaapor) 7:368-72;
(Maipuran) 3: 379; (Wai Wai) 4:80-86;
(Paumari) 3:240, 263-7; (Warekena) 4:287-322;
(Pirahä) 7:251, 262-3, 277-80; (Yagua) 2:348-360
(Proto-Tupi-Guarani) 4:494, 539-45, 563, Nuevo San Martin (town, Peru), 4:444
Index 635
number, /: 16, 61, 107-9, 144,185-6, Orellana, Francisco de, /: 4, 8; 2:3
205-6, 392-4,416-7, 500, 515,578; Origüera, Toribio de, /: 8
2:17, 144, 197-8, 354-8, 377-8,445-6, Orinoco (river, Venuzuela), 2:17; 4: 227
453; 3: 89-91, 123, 254-9, 286-9; orthography, 7:471; 2:221,429-30; 3:3;
4:300-304; 4:579
(Apalai) 7:61, 107-9; OSV constituent order, 7:16-8, 326-31,
(Arawakan)/:500,515; 408-9,472,475-6, 479-81,485;
(Canela-Krahö) 1:144,185-6; 4:556-7
(Guajajara)7:416-7; Ott, Rebecca H., 3:359,430
(Macushi)3:90-l,123; Oett, Willis G., 3:359,430
(Paumari) 3:254-9, 286-9; OVS constituent order, 7:16-8, 33-4, 331,
(Sanuma) 2:17,144,197-8; 408-9,447,451,473,475,487-8,497,
(Urubu-Kaapor) 1:392-4; 499; 4:14, 27,445-6, 556-7
(Warekena) 4:300-304;
(Yagua) 2: 354-8, 377-8,445-6,453 Pacaas Novos language, 4:4
numerals, Pacaya language, 2:252
(Apalai) 7: 88; Paezan languages, 7:12
(Arawakan) 7:578; Pajonal Campa language, 7:567-642 passim
(Guajajara)/:412; Paläcio, Adair, 7:23
(Macushi) 3:89; Paleo-American languages, 7:12
(Paumari) 3:337; Palikur language, 7:469-566 passim; 3:13,
(PreAndine Arawakan) 1:578; 355-499 passim
(Proto-Tupi-Guarani) 4:494, 555, 602-3; Panare language, 4:3, 452
(Sanuma) 2:151; Pano, Eleazar, 2:459
(Urubu-Kaapor) 1:374-5; Panoan languages, 7:15-16; 2:8, 384-5;
(Warekena) 4:312-13; 3:3,372; 4: v, 3, 8,12,15,443,444,
(Yagua) 2:349,453 477-9
Papago language, 7:457,462
obligation marking (Wai Wai), 4:65 Para (state, Brazil), 7:33, 199; 3:6; 4:491,
oblique-topicalized construction in Tupi- 495
Guarani, 4:526-8, 556, 561, 575-6, 580, Paraguay (nation), 7:15; 3:360; 4: v, 491,
592 495
Ocaina people, 3:362 Paraguay (river), 7:472; 3:360
Odmark, Mary Ann, 1:539, 553; 3:162 Parakanä language, 4:495
Oiampi language, 1:14; 2:8,9 Parkwoto people, 4:25
Oiapoque (river, French Guiana), /:471 Parara language, 2:252
Oldfield, Margery L., 3:6 parataxis, 7:37-41, 39,40, 52-3,137-9,
Old Guarani language, 4:494,495,496, 522, 208-13, 211 -3, 335-6, 622-5; 2:34-9,
547, 560, 572, 578 264-91; 3:31-5,173-5; 4:36-42,
Olson, Mike, 2:413 231-3
Omagua language and people, 7:8; 4:495 Paraua (river, Brazil), 7:326
operator-operand theory of constituent order, Parecis language, 7:469-566 passim, 638;
7:452 3:13, 355-499 passim
oral literature, Amazonian, 4:6 Paressi-Saraveca(n) languages, 7:471
Oregon Foundation, 3:355 Parima (river, Brazil), 2:17
636 Index
Parintintin language, 4:495-6, 501, 504, 509, (WaiWai)4:ll,47-50;
514, 518-9, 522, 540, 546, 555-6, 568, (Warekena) 4:239
572, 578,588-606 passim, 615 (Yagua) 2:277-87, 357
Parker, Stephen, 3:359,455 Patamona language, 3:9,23
Parque Nacional de Xingu (Brazil), 7: 472 Paumari language, 7:459,469-566 passim,
participants (in discourse), 7: 305-6,442, 487-9,515; 2:9; 3:4,9,10-12,
591-7,625-36,630-1; 2:62-3,253, 161-352,363,374; 4:15
361-369; 3:196-201 4:54-5,248,250, Payne, David L., 7:462-3,469-566 passim,
453,475-6; 568, 575, 579-80, 583-4, 589-91, 596,
particles, 7:70-1,88-9,111-20,188, 604, 607, 637; 2:8, 259-60, 299-302,
225-6, 236-8, 303-8, 397-8, 601-2; 331, 334, 354, 367, 384,422,445; 3: v, 10,
2:212-8,266,307-8,418-428; 3:67, 12,355-499; 4:11,13
111-12,133-9,339-45; 4:10,12, Payne, Doris L., 7:440-65,474,476, 560;
129-47,398-9,494,552-4; 2:7, 9, 249-474; 3: v, 4, 382
(Apalai) 7:70-1, 88-9,111-20; Payne, Judith, 7:469-566 passim, 568, 580,
(Arawakan) 7:601-2; 596,607,637; 2:296; 3:355,380,469
(Canela-Krahö) /: 188; Payne, Thomas E., 7:463; 2:7,249-474;
(Macushi) 3:67, 111-2, 133-4, 136-9; 3: v, 4; 4:476,478
(Paumari) 3:339-45; Peba language, 2:252; 4:15
(Pirahä) 1:225-6, 236-8, 303-5, 307-8; Pebas (city, Peru), 2:251, 317
(PreAndine Arawakan) 1:601-2; Peba-Yaguan languages, 7:440; 2:252
(Proto-Tupi-Guarani) 4:494, 552-4; Pedra Furada archaeological site, 4:6
(Sanuma)2:212-9; Pemon language, 3:9, 23
(Urubu-Kaapor)7:397-8; Pefia, Hilario,2:313,459
(Waiwai) 4:10,129-47; , Espinosa, 2:252
(Warekena)4:12; perfective aspect, 2:388, 93, 398-9;
(Yagua) 2:266, 307-8,418-428 3:117-8; 4:336-9,342-4,447-51
Paru (river, Brazil), 1.33 Perlmutter, David M, 7:499; 2:277
passives, 7:47-9, 53, 142, 186-7, 219, 341, person, 7:61, 107-9, 185-6, 297, 392-4,
394,428, 508-10, 536, 588-9, 611 -2, 416-7, 574, 585-6, 610, 618-9; 2:197;
634-5; 2:46-7, 198, 277-87, 357; 3:53-4, 123, 286-9; 4:120-21, 293,
3:37-40, 123-6, 180-4,289-306; 4.11, 322, 565-73, 588-92;
47-50, 239 (Apalai) 7:61, 107-9;
(Apalai) 7:47-9, 53; (Arawakan) 7:574, 585-6,610, 618-9;
(Arawakan) 7:508-10, 536, 588-9, (Canela-Krahö) 7:185-6;
611-2; (Guajajara) 7:416-7;
(Campa)7:634-5; (Macushi) 3:53-4,123;
(Canela-Krahö) 7:142,186-7; (Paumari) 3:286-9;
(Guajajara)7:428; (Pirahä) 7:297;
(Macushi) 3:37-40, 123-6; (PreAndine Arawakan) 7:574, 585-6, 610,
(PaumarO 3:180-84, 289-306; 618-9;
(Pirahä) 7:219; (Sanuma) 2:197;
(PreAndine Arawakan) 7:588-9, 611-2; (Tupf-Guarani) 4:565-73, 588-92;
(Sanuma)2:46-7, 198; (Urubu-Kaapor) 7:392-4
(Urubu-Kaapor)7:341,394; (Wai Wai) 4:120-21;
Index 637
The four volumes of this handbook bring together a body of descriptive and the-
oretical work on a group of languages virtually absent from current linguistic
discussion. The languages of the Amazonian region have been so little docu-
mented - particularly in respect of their grammatical structures - that even wide-
ranging comparative and typological studies often fail to take account of them. Yet
this group contains much interesting typological diversity and sometimes presents
unique structural characteristics, not attested in other areas of the world.
Volume 1. 1986. The language families represented include Carib, Tupian, Ge,
Peba-Yaguan, and Arawakan. Two typological studies present evidence of previ-
ously unattested combinations of properties, some of which are specifically pre-
dicted to be impossible under certain current theories of language universals.
Preface · Map of South America · Abbreviations · Desmond C. Derbyshire and
Geoffrey K. Pullum, Introduction · Part I: Grammatical Sketches · Outline of
Contents for Each Chapter in Part I · Edward and Sally Koehn, Apalai · Jack
and Jo Popjes, Canela-Krahö · Daniel L. Everett, Pirahä · James Kakumasu,
Urubu-Kaapor · Part II: Word Order and Typological Studies · Carl H. Har-
rison, Verb Prominence, Verb Initialness, Ergativity and Typological Disharmony
in Guajajara · Doris L. Payne, Basic Constituent Order in Yagua Clauses: Im-
plications for Word Order Universals · Part III: Comparative Arawakan Stud-
ies · Desmond C. Derbyshire, Comparative Survey of Morphology and Syntax in
Brazilian Arawakan · Mary Ruth Wise, Grammatical Characteristics of PreAndine
Arawakan Languages of Peru
Handbook of Amazonian Languages
Edited by Desmond C. Derbyshire and Geoffrey K. Pullum
Mouton de Gruyter · Berlin · New York
Volume 3. 1991. The languages investigated have interesting word order patterns
and case-marking systems: Macushi is (probably) OVS and consistently ergative;
Paumary is SVO with co-occurring ergative and accusative case-marking systems.
Preface · Map of South America · Abbreviations · Desmond C. Derbyshire and
Geoffrey K. Pullum, Introduction · Part I: Grammatical Sketches · Outline of
Contents for Grammatical Sketches · Miriam Abbott, Macushi · Shirley Chapman
and Desmond C. Derbyshire, Paumari · Part II: Maipuran (Arawakan) Classi-
fication · David L. Payne, A Classification of Maipuran (Arawakan) Languages
Based on Shared Lexical Retentions · Chris Barker and Geoffrey K. Pullum, Cu-
mulative Index to Volumes 1-3