Sunteți pe pagina 1din 19

Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.

Distributed Morphology1
Daniel Siddiqi*
Carleton University

Abstract
This paper provides an overview of Distributed Morphology (DM, Halle and Marantz 1993,
1994), a framework withing the Minimalist Program (Chomsky 1995) which incorporates both
morphology and syntax into one component of the grammar rather than postulating two separate
components (as Lexicalist Minimalism does). This paper will outline the major distinctive features
of DM, sketch some of the objections to its approach (and offer responses) and survey several of
the internal debates among its practitioners.

1. Introduction
The morphological component and the syntactic component of the grammar have much
in common. For one thing, they manipulate many of the same features: For example,
case is often considered a syntactic feature with a morphological realization. This is true
of tense, agreement, aspect, mood, voice, etc. etc. Furthermore, typologically speaking,
the responsibility of accounting for these formal features is distributed in different ways
across the syntaxes and the morphologies of languages: highly fusional languages like Latin
and Greek depend heavily on morphology, while more isolating languages like Mandarin
and English depend more on syntax. In polysynthetic languages such as Mohawk and
Swahili, the lines between syntax and morphology are completely blurred.
On top of realizing the same features, syntax and morphology2 have very similar struc-
tures. Both involve hierarchical structures created by adjoining two units to each other.
In fact, at first glance, a morphological tree and a syntactic tree look nearly identical to
one another (1).
(1) Syntactic and Morphological parallel structure:

a. Complex morphological structure


N
V –ion
N –ate
design
b. Complex syntactic structure

Dogs VP
chase cats

ª 2010 The Author


Journal Compilation ª 2010 Blackwell Publishing Ltd
Distributed Morphology 525

Of course, morphology and syntax are not identical to one another (otherwise, we would
not be inclined to call them different things!). One key difference is that morpheme
selection and morphological operations are often conditioned phonologically. This seems
to never be true of syntax. Another key difference is that morphological operations vary
in productivity within a speaker; again this is arguably untrue of syntax. Finally, syntax
has movement and other long distance relationships. This does not seem to be true of
morphology.
Because of these differences, while formal morphological theory has often been explic-
itly incorporated into formal phonological theory—Optimality Theory (OT: Prince and
Smolensky 1993) and Lexical Phonology and Morphology (LPM: Kiparsky 1982) are two
such examples – it has not often been incorporated into formal syntactic theory.
However, this incorporation into phonological theory ignores much of the similarities
between syntax and morphology. Very little of formal syntactic theory includes or even
accounts for morphology. In fact, a large portion of the body of syntactic theory relegates
morphology to a separate component called the Lexicon.
The purpose of this article is to provide a short sketch of a syntactic framework that
incorporates morphology. Distributed Morphology (DM: Halle and Marantz 1993, 1994)
is a framework within the Minimalist Program (Chomsky 1995) that incorporates all the
major syntactic elements of Minimalism but crucially rejects the idea that there is a sepa-
rate component of the grammar responsible for the morphology of a language (i.e. the
Lexicon) and instead distributes phenomena normally considered to be morphological in
nature throughout the entirety of the grammar (as opposed to locating it in one place).
This article describes the major tenets of the framework, its rationale, and some strong
criticisms of it with the purpose of familiarizing readers who are familiar with Lexicalist
Minimalist3 syntax with the DM framework.

2. Morphology Distributed
The Lexicon is typically considered to have several different functions. It is the storage
place of the formal meanings that the grammar manipulates. It is the storage place of the
arbitrary correspondence of those features to sounds. It is the component of the grammar
that builds complex words (i.e. compounding, affixation, and derivation). It is the locus
of morphological processes (such as inflection). In some theories, it is the location of
some phonological processes. It is the storage place of our real world knowledge of the
meanings and referents of words.
What makes DM crucially different from Lexicalist models of syntax (in particular
Lexicalist Minimalism) is that in DM, no one component accounts for all of these things.
Instead, these tasks are spread throughout the grammar. The formal meanings of the gram-
mar, called formal features, are the only things manipulated by the syntactic component of
the grammar and as such are part of the syntax. The arbitrary correspondences between
sound and meaning are called Vocabulary Items (VIs, rather than ‘lexemes’) and are stored
in a passive component of the grammar called the Vocabulary. VIs are recovered from the
Vocabulary and inserted into a derivation at PF (after the syntactic derivation, which is
why DM is called a Late Insertion model of the syntax). Similarly, the real word knowl-
edge we have about words (such as their referents, their special meanings, or their mem-
bership in idioms) is stored in another area called the Encyclopedia, which is not accessed
during a derivation (and some may even argue that it is completely extra-grammatical).
The operations typically assumed to be in the Lexicon are also distributed throughout
the grammar. Instead of being built by a dedicated component of the grammar, complex

ª 2010 The Author Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.x
Journal Compilation ª 2010 Blackwell Publishing Ltd
526 Daniel Siddiqi

words are ‘derived’ by the syntax. Most work in DM assumes that some processes in the
syntactic component are purely morphological in nature (such as head movement) and that
these processes occur after spell-out and before PF. Other theorists within DM (Julien
2002 being the most prominent) argue that there are no processes in the syntax that are
purely morphological. In addition, DM posits that there are morphological processes that
alter the phonological form after PF. Most of these, such as infix alignment, are considered
the realm of phonological theory and are usually outside the realm of DM. One notable
exception is readjustment rules, which are a key part of DM and occur after PF.

3. The Rejection of the Lexicon


While it may not seem that divergent from main stream Minimalism today, the rejection
of the Lexicalist Hypothesis was, at the time of DM’s invention, its most distinct feature
and its most controversial. Even today, when compared to other syntactic models such as
LFG (Lexical Functional Grammar, Bresnan 2001) or HPSG (Head-driven Phrase Struc-
ture Grammar, Pollard and Sag 1994), the rejection of the Lexicon is the marked feature
of DM.
Historically, the best representatives in the Lexicalist debate have been Chomsky
(1970)4 and Marantz (1997b). The debate centers on the distributional differences
between a gerund (growing) and a derived nominal (growth). Chomsky argues that these
differences arise because derived nominals are formed in the Lexicon, while the syntax
creates gerunds. Chomsky’s argument has two main features. First, gerunds are completely
productive, while derived nominals are not. For example, while both the barbarian’s
destroying of the village and the barbarian’s destruction of the village are available, there is no
derived nominal equivalent to the barbarian’s sacking of the village (cf. the barbarian’s sack-
tion ⁄  sack-th ⁄ sack-ity of the village). Second, while gerund phrases (i.e. John’s growing of
tomatoes) always have exactly the same meaning as their corresponding verb phrases (John
grows tomatoes), this is often not true of derived nominals, which can have idiosyncratic
meanings (#John’s growth of tomatoes). This distinction gave rise to the weak version of the
Lexicalist hypothesis that argues that inflection is part of the syntactic component of the
grammar, but derivation is a function of a separate component.5
Marantz (1997b) ultimately points out that the reason that Chomsky reaches his
conclusion was based in the transformational nature of standard theory at the time. The
fact that the derived nominal’s meaning is not transparently linked to its verbal equivalent
is only a problem for syntactic theory if you assume that the former is transformed or
derived directly from the latter. As today’s theories do not assume a common deep struc-
ture between the two forms, contemporary theory has no reason to argue that the mean-
ing of the verb phrase, the gerund, and the derived nominal should necessarily be the
same. Therefore, contemporary syntactic theory does not have the same issues with
accounting for both gerunds and derived nominalizations with the same mechanisms.
With the nominalization argument being more or less obsolete, Chomsky’s (1970)
remaining major argument for the independence of a lexical component comes from the
‘special’-ness of the word. In much of linguistic theory, the word has an importantly unique
role in the grammar. Again, in theories like LFG and HPSG, the grammar is driven by
words (to the point where you can argue that the syntactic component is truly minimal).
However, the word is a linguistic unit that is notoriously difficult to define specifically
or technically. Ignoring any orthographic definition, three possible definitions spring to
mind: (i) syntactic, (ii) semantic, and (iii) phonological. Unfortunately, these three differ-
ent definitions of the word seldom overlap perfectly. For example, can’t is only one

ª 2010 The Author Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.x
Journal Compilation ª 2010 Blackwell Publishing Ltd
Distributed Morphology 527

phonological word while it has two semantic meanings and represents two syntactic
nodes (at least T and Neg, perhaps also Mode). Conversely, buy the farm contains several
phonological and syntactic words, but has only one meaning. To sum, each of the
following can be considered one word or many from some metric: can’t, dump truck,
blackboard, the car, and the whole nine yards.
The Lexicalist hypothesis as put forth by Chomsky (1970) argues that there are several
features of words that grant them special status: words are the locus for special sound
processes, special structural process, and idiosyncratic non-compositional meaning.
Marantz (1997b) refutes each of these claims to special status. As to the first, Matantz
points out that the idea of a ‘phonological word’ as used by a formal phonological
theorist (i.e. a prosodic word – one domain for stress assignment) differs often from what
a syntactician might consider a word (i.e. a lexeme). Some syntactic heads are too small a
unit for lexical phonology, while others contain complex prosodic structure. Further-
more, if the prosodic word and the syntactic head did happen to coincide, Marantz
(1997b) argues that that does not entail that the formation of the prosodic word must
occur before the syntax component of the grammar (unless the Logical form of the utter-
ance was sensitive to the word’s prosody).
As to structural processes that are unique to words, beyond the gerund-nominalization
discussion, even a cursory glance at a morphological structure shows that it contains a
basic hierarchal structure that is very similar to syntactic structure which provides the basis
for Baker’s (1985, 1988) Mirror Principle6, which argues that cross-linguistically the syntac-
tic ordering of formal meanings (such as tense, aspect, voice, etc) is parallel to the order-
ing of affixes that realize those same meanings morphologically (such as in heavily
polysynthetic languages7), which means that the morphology of those languages must be
analyzed syntactically. This parallel seemingly contradicts the Lexical Integrity Principle
(Chomsky 1970; Jackendoff 1997; Bresnan and Mchombo 1995 and many others), which
argues that necessarily the syntax cannot see into the structure of the words.
Finally, as to the most consistent argument for a Lexicon, that the Lexicon is the loca-
tion of idiosyncratic sound ⁄ meaning correspondences. Chomsky (1970) assumes that
words are the largest linguistic unit with idiosyncratic meaning (morphemes being the
smallest). All linguistic units larger than the word, Chomsky argues, are constructed by
the syntax and their meanings are thusly predictable. Jackendoff (1996) argues that there
is crucially no empirical difference between the idiosyncratic meaning of words and the
idiosyncratic meanings of idiom phrases (such as kick the bucket). Marantz (and conse-
quently DM) follows Jackendoff and recognizes that idiosyncratic meanings exist at any
level. Thus, the term ‘idiom’ more accurately refers not only to complex phrases and to
complex words with non-compositional meanings, but also simple words (such as cat).
In many ways, DM is the result of accepting the arguments against a dedicated
morphological component and rejecting the Lexicon. DM is essentially Minimalist syntax
without a Lexicon.

4. DM’s Distinctive Features


Unsurprisingly, the rejection of the Lexicalist Hypothesis has several side effects on the
grammar proposed by DM. The vast majority of the differences between DM and Lexi-
calist Minimalism simply evolved out of the fact that DM now deals with morphology as
well as traditional syntax. The three biggest distinctive features of DM are typically
considered to be: (i) late-insertion, (ii) persistent syntactic structure within words, and
(iii) underspecification.

ª 2010 The Author Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.x
Journal Compilation ª 2010 Blackwell Publishing Ltd
528 Daniel Siddiqi

Unlike in Lexicalism, in DM, the syntax only manipulates formal morpho-syntactic


features in order to generate syntactic structures. These features are selected from an
abstract list of features and (either individually or in bundles) are the building blocks used
by the syntax. These features ultimately cannot be pronounced – we need phonology to
do that. The Late Insertion Hypothesis (Halle and Marantz 1994) argues that the phonol-
ogy that realizes these features is not inserted until PF, rather than being present through-
out the derivation.
In DM, at spell-out, terminal nodes are comprised only of bundles of interpretable
morphosyntactic features. It is only after all the syntactic processes (excluding those
between spell-out and LF) are finished that phonology enters the derivation from the
Vocabulary, a static list of VIs that provide the sounds needed to realize the interpretable
features contained in the terminal nodes of a derivation so that that derivation can be
pronounced. Thus, the Vocabulary is the inventory of signs available to the language,
which can be used to represent the featural nature of the syntax.
To see this at work, imagine a syntactic tree with a terminal node comprised of
the features [present], [singular], and [3rd person]. In English, the affix -s realizes that
combination of features. What this means is that there is a VI, -s, which is inserted
into that node at spell-out to realize that group of features with the overt phonology
⁄ z ⁄ (2).

(2) Specification for -s


[present] -s
[singular] /-z/
[3rd person]

Another major distinctive feature of DM is what Haugen (2008) calls the ‘pervasive syn-
tax perspective’. In DM, the complex words are derived by the same processes that derive
the complex structure of a sentence. Essentially, a complex word is just a phrase. The
spell-out of a complex constituent as a ‘phrase’ or a ‘word’ is simply determined by the
nature of VIs in the structure. For example, both little red dog and authentication are com-
plex noun phrases (notated in DM as nP, as the dominating constituent is a projection of
the functional nominalizer n – read ‘little-n’, named in parallel to little v, which does the
same job for verbs). Authentication is a formed form the nominalizing head and the verb
phrase authenticate which itself is composed of a verbalizing head, -ate, and an adjective
phrase authentic8 (3).

(3) authentication

nP
vP -ion

AP -ate
authentic

ª 2010 The Author Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.x
Journal Compilation ª 2010 Blackwell Publishing Ltd
Distributed Morphology 529

Authentication is spelled out as one complex word rather than a phrase containing isolated
functional morphemes simply because the VIs that realize the included functional heads
are affixal in nature. Clitics have been treated historically the same way by syntactic
theory. Therefore, lexical decomposition and phrase structure are effectively the same
thing–which is why Harley and Noyer (1999) call this feature of DM ‘syntactic structure
all the way down’.
The last of the major distinctive features of DM is underspecification. Underspecifica-
tion is employed when a VI realizes the formal features of a terminal node. The VI’s
insertion is governed by the subset principle, which dictates that a VI has minimal specifi-
cations that a target node must satisfy but the node can contain features beyond those
specifications. For example, the English copula9 are is compatible with several different
combinations of tense and phi features including: 1st person plural present, 2nd singular
present, 2nd plural present, and 3rd plural present. This distribution is attributable to the
fact that are is specified only for the feature [present] which is a subset of all four environ-
ments.
As DM uses underspecification, it makes very different predictions about the behavior
and distribution of syntactic elements from those Lexicalist models whose lexical items
are (in contrast) fully specified as to what they realize.10 For example, an underspecifica-
tion model predicts the emergence of less marked, elsewhere forms, as they are the
underspecified form. An example in the realm of morphology is the use of they as a
singular impersonal pronoun. In syntax, the use of the unmarked copula in such forms as
How’s the kids?11

5. Insertion and the Subset Principle


In DM, the ‘words’ (i.e. VIs) enter the derivation at PF through a process called Inser-
tion. To see an example of this, consider the Spanish determiner system, which has a
maximally contrastive distinction for gender, number, and definiteness (4).
(4) Spanish Determiners.
el masculine singular definite
los masculine plural definite
la feminine singular definite
las feminine plural definite
un masculine singular indefinite
unos masculine plural indefinite
una feminine singular indefinite
unas feminine plural indefinite
For the purpose of this illustration, let us assume that the difference between the deter-
miners un and el is that el is specified for [def], while un is unspecified for definiteness.
Otherwise, el and un are identically specified for number and gender. Realizing terminal
nodes with these different phonological strings is accomplished through the competition
of the two items. In the Spanish example, when el and un compete for insertion into a
D terminal node specified for the feature [def], el wins the competition. This is because
el is better specified for the features present in the target node than is un. While both are
perfectly eligible for insertion (neither VI contains features which conflict with the node),
in a competition for insertion, the VI which is specified for the largest number of features
without being specified for any features that are not in the target node will win the
competition (in this case el) (5). This is called the Subset Principle.

ª 2010 The Author Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.x
Journal Compilation ª 2010 Blackwell Publishing Ltd
530 Daniel Siddiqi

(5) Competition
un12 underspecified
Syntactic el [def] INSERTED CANDIDATE
environment la [fem][def] overspecified (conflict)13
[definite] una [fem] overspecified (conflict)
[singular] los [def] [pl] overspecified (conflict)
[masculine] unos [pl] overspecified (conflict)

To put this principle in somewhat simpler terms, insertion is strictly governed by the two
following constraints: (i) The VI in the competition that has most features in common
with the target node without having any more will win the competition and be inserted.
(ii) The target node may have more features than the VI to be inserted, but the VI may
not have more features than the target node.
In other words, the features of the VI must be a subset of the features on the node
(though not necessarily a proper subset). When a VI is inserted into a node, all the formal
features of the node are removed from the derivation (even those not realized by the VI).
This removal of features is called Feature Discharge.
Some complications of this system involve a VI’s sensitivity to the features of nearby
nodes (usually constrained to c-commanding or c-commanded nodes). One such sensitivity
is called licensing, which allows insertion of certain VIs only if another nearby node meets
certain requirements. For example, certain affixes only go on certain classes of stems – in
English, Latinate affixes attach mostly to Latinate stems (in, and -ity are two such affixes
which cannot attach to native English words). The process of insertion is usually assumed
(after Bobaljik 2002) to be cyclic, allowing an affix to see a stem in such a way.
This sensitivity to neighboring nodes sometimes results in a VI secondarily realizing a
feature realized by another VI (often the case with agreement marking). In this case, this
secondary realization of features is called Secondary Exponence (see Embick 1997; Harley
and Noyer 2000 for good examples of the system).
This is easily shown again in the Spanish determiner system. The competition between
el and la is not settled by the featural content of the node they are being inserted into (by
hypothesis, the nodes realized by these different determiners are identical), rather it is
settled because la is a secondary exponent of the feature [fem], present on the noun14.
The gender feature is not in the determiner node and is not discharged by the insertion
of a determiner VI, but rather is in a node c-commanded by the determiner (the noun)
and is discharged by another VI (the VI that realizes the noun). Similarly, the difference
between el and los is that los is a secondary exponent of [pl], again not a feature that is in
D, but rather is located in another node c-commanded by the determiner (in this case
Num). For an example of this see (6, secondary exponents are marked with round
brackets)

ª 2010 The Author Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.x
Journal Compilation ª 2010 Blackwell Publishing Ltd
Distributed Morphology 531

(6) DP
Primary exponent of [pl]
[def] NumP
-s NP Primary exponent of [fem]
[pl]
mesa
[fem]
un underspecified
el [def] underspecified
la ([fem])[def] underspecified
una ([fem]) underspecified
las ([fem])[def] ([pl]) INSERTED CANDIDATE
unos ([pl]) underspecified

To complete our example of the Spanish determiner system, let us look at the
complete set of determiners along with a set of hypothetical featural specifications (7, sec-
ondary exponents still marked with round brackets).

(7) Spanish Determiners15


phonology feature specification
el D16, [def],
los D, [def], ([pl])
la D, [def], ([fem])
las D, [def], ([pl]),([fem])
un D
unos D, ([pl])
una D, ([fem])
unas D, ([pl]), ([fem])

With this set of featural specifications, we can now see that the subset principle predicts
which determiner will appear in which environment. For example, if the node for inser-
tion had the feature [def] and that node c-commanded [fem], the VI to win the competi-
tion for insertion would be la. The VIs for los, las, unos, and unas are all eliminated
because they realize a feature ([pl]) that is not contained in the target environment. Of
the candidates that remain, la is specified for the most of the relevant features (in this
manufactured case, all of them). For another example, for insertion into a determiner
node not specified for definiteness that is c-commanding only the plural feature, all but
un and unos are eliminated for containing too many features. The competition results in
insertion of unos because it is the best specified.
There are two more notes of importance on competition. The first is exemplified
by un in the aforementioned example. In DM, the elsewhere condition or ‘default’ in
any given morphological pattern of alternation is the VI with the fewest requirements.
This means that it can go anywhere but only surfaces when no other VI can be
inserted (because they have all been eliminated for overspecification). Another note of
importance is that it is possible for competition to arrive in a tie (when two different

ª 2010 The Author Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.x
Journal Compilation ª 2010 Blackwell Publishing Ltd
532 Daniel Siddiqi

VIs are each specified for one of two features which can co-occur, for example –
such a tie could happen in the aforementioned example between la and los if the VI
las did not exist). The system proposed for tie breaking proposed by Halle and Ma-
rantz (1993, 1994) is that competition was simply (but crucially) extrinsically ordered.
The winning VI in such a tie is simply the first in this ordering. However, an alter-
nate tie breaking system was proposed by Noyer (1997) that employed the Universal
Hierarchy of Features. In this tie-breaking system, VIs that realize features higher on
the hierarchy are preferred for insertion. For example, if number is higher on the
hierarchy than person, than in the los tie with la, los would win the competition
because its features are higher in the hierarchy. This second system seems to be pref-
erable because it appeals to a universal feature of UG rather than positing a language-
specific ranking of preference for what affixes to use.

6. Morphology in DM
With the aforementioned changes to the grammar, the job of the syntactic component in
DM is now to also do the morphology in the derivation. Combine this with the narrow-
ing of the scope of the ‘core syntax’ in formal syntactic theory, and suddenly DM needs
several syntactic operations whose main jobs are what we have historically considered
morphology (the majority of which occur after spell-out and before PF). The major
operations that are unique to the DM model of the syntax (besides insertion) are
morphological merger, fission, fusion, and readjustment rules.
Morphological merger is the operation through which syntactic structure is effec-
tively replaced with morphological structure. Marantz (1988) defines morphological
merger as ‘a relation between X and Y be[ing] replaced by ([or]expressed by) the
affixation of the lexical head of X to the lexical head of Y’ (Marantz 1988: 261).
Put simply, morphological merger is the process through which two (usually adjacent)
zero-level nodes (i.e. heads) come to occupy the same position, creating a complex zero-
level node. Morphological merger has been argued to be the process behind several
familiar operations including head movement, affixation, and affix lowering (see Embick
and Noyer 2001 for one very good example which describes English affix lowering and
Do-support in DM). Morphological merger is the process that creates the difference
between two heads ultimately forming one compositional unit (a complex word) and
two heads forming a complex syntactic constituent (a phrase) (see 8 for a contrast of
analytical tense and affixal tense in English).

(8) a. Analytical Tense in English


TP
did VP
chase

ª 2010 The Author Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.x
Journal Compilation ª 2010 Blackwell Publishing Ltd
Distributed Morphology 533

b. Affixal tense in English


TP

T VP

V –ed t
chase Application of
merger

The process of fusion is in many ways the extension of morphological merger. In


many cases, one lexeme (VI) primarily realizes more features than are present in just one
syntactic node. This is often the case with agreement marking of phi-features on the verb
(-s realizes not only present tense but third person and singular as well). After morpho-
logical merger has created a complex head, the application of fusion reduces the complex
head to a simplex one, containing all the features previously contained in the complex
head. Again, the Spanish determiner system provides a nice example of fusion. The deter-
miner del in Spanish is a blend (or portmanteau) of preposition de and the determiner el. A
DM account of this would argue that D and P have conjoined into a complex head
through morphological merger and then fused into one head containing the features of
both D and P. The VI del realizes those features (9).

(9) Fusion for del


PP

P DP
Spelled out
D NP form
Application [def]
of merger Sol
[masc]

Complex result of head


movement
P
application [P]
D P of fusion [D]
[def]

The opposite of the process of fusion is the process of fission (Halle 1997; Noyer
1997). As opposed to fusion (which takes two positions of exponence and reduces them
to one), fission takes the features in node and spreads them out over many nodes (which
are created in the process). Fission is most useful to separate from one another

ª 2010 The Author Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.x
Journal Compilation ª 2010 Blackwell Publishing Ltd
534 Daniel Siddiqi

phi-features which are generated in the syntax all in one head (AGR). In this way, a
language can have separate lexemes for different combinations of person, gender, and
number (10).
(10) Fission of phi-features

AgrS0 AgrS
[1st]
[masc] [1st]
[p1] [p1] [masc]

The last of these morphological operations is readjustment. Up until now, we have talked
exclusively about competition in choosing a VI based on featural content. However,
another aspect of morphology is allomorphy, which in DM can be expressed as one VI
having different phonological realizations (but still realizing the same feature combina-
tion). Because this involves the same feature combinations, this typically cannot be
handled by competition. Allomorphy comes in two relevant varieties. The first is called
phonologically conditioned-allomorphy. An example in English is the choice of ⁄ -t ⁄ or
⁄ -d ⁄ as the realization of -ed depending on the voicing of the stem final sound. In DM,
this is a phonological process that happens at PF outside of the syntax. The second is
when a different set of sounds realize a particular set of features because of the particular
stem that the affix is attaching to. An example of this in English is the use of obsolete -en
for the plural marker following child and ox.17 This is typically called morphologically
conditioned allomorphy and in DM is realized for affixes by secondary licensing (with
sensitivity to particular stems).
However, when the stem itself is what undergoes the allomorphy, the picture is
more complicated. In DM, the distinction between content word and functional word
is argued to be the presence or absence of a root (symbolized as ). A  is a formal
feature that links the node to extra-grammatical meaning that gets interpreted by the
Encyclopedia. Roots are present in all content words and absent from all functional
words. Furthermore, roots are generic. This effectively means that the syntax is blind
to any formal difference between cat and dog. It is this blindness that mandates the
use of readjustment. Typically, morphologically conditioned allomorphy is accounted
for by competition. However, because content words all realize the same feature (),
content words cannot compete with each other for insertion. Thus, morphologically
conditioned allomorphy of roots (such as the change from mouse to mice or -cieve to –
cept18) cannot be accounted for by competition. Instead, DM argues that after the VIs
have been inserted, certain conditions (such as being c-commanded by [pl] in the case
of mice) can license the application of a readjustment rule, which changes the phono-
logical form of the VI to another form (see Marantz 1997a,b).19
See the following paragraphs for the regular derivation of walked versus the irregular
derivation of slept.

ª 2010 The Author Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.x
Journal Compilation ª 2010 Blackwell Publishing Ltd
Distributed Morphology 535

(11) a. Derivation of complex verb


TP
[past] vP
v

Merger results in complex head

v
[past]

b. Realization as walked
V

[past]
Insertion
/wak/ /-t/

c. Realization as slept
V

[past]
Insertion
/slip/ /-t/
Readjustment
/slεp-/
Readjustment rules are completely idiosyncratic, usually applying to only one VI,
though some researchers have argued that they can apply to classes of content words to
account for patterns irregular morphological change (see Embick and Noyer 2006;
Embick and Halle 2005 and others). For example, one readjustment rule could be said to
target the phonological class of verbs rhyming with drive (forming drove as the past tense),
explaining the semi-productivity of that rule. Because readjustment rules necessarily take
place after insertion of sounds, they are the only morphological operations in DM to
occur after PF, so it is arguable that they are in fact a phonological operation, though no

ª 2010 The Author Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.x
Journal Compilation ª 2010 Blackwell Publishing Ltd
536 Daniel Siddiqi

one has made that argument. If readjustment rules were to be considered a phonological
operation, like the phonologically conditioned voicing of the regular past tense -ed, they
would not actually be a part of the morphosyntactic component of the grammar, and
instead would be part of the phonological component, which most of contemporary
linguistic theory assumes used different principles20.

7. The Derivation of a Sentence in DM


Because the DM framework is a part of the Minimalist Program, the construction of a
sentential utterance in DM is very similar to that in Lexicalist MP.
(12) a. Minimalist construction of an utterance

Syntactic Operations
Spell-out

PF LF

b. Construction of an utterance in DM21

MORPHOSYNTACTIC
FEATURES:
Syntactic
[+N]
Operations
[+singular]
[3rd person]

Morphological
Operations

VOCABULARY
INSERTION: Phonological Form Logical Form
/kæt/
/–s/

ENCYCLOPEDIA: Conceptual
Non-linguistic Interface
Knowledge-- (Meaning)
Little furry thing,
likes to sleep on
my face.

The differences between the Lexicalist model and the DM model all arise out of the
abandonment of the Lexicon and the distribution of its duties across the syntax. The

ª 2010 The Author Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.x
Journal Compilation ª 2010 Blackwell Publishing Ltd
Distributed Morphology 537

numeration in DM only contains morphosyntactic features, no words and no phonology.


In addition to the normal syntactic processes, in DM, after spell-out and before PF there
are more syntactic processes that create the morphological structure. Just as in Lexicalist
Minimalism, the syntax discharges the uninterpretable syntactic features. However, in
DM, interpretable features are discharged from the derivation at PF through the process of
Vocabulary Insertion. Finally, in DM, the logical form and phonological form come back
together where they are interpreted by the Encyclopedia.

8. Internal Debates
Like many scientific frameworks, DM is not used exactly the same way by all researchers
who subscribe to it. There are some interesting internal debates which I will share here.
The first variation of DM comes in the form of Julien’s (2002) proposal of a fiercely
reductionist view of DM. Julien (2002) argues that word order and affix order are
entirely derived by normal syntactic operations, effectively eschewing all the operations in
DM between spell-out and PF. From a certain point of view, in traditional DM, the
Lexicon’s word building power has not been removed entirely from the grammar, it has
just been moved to being after the syntax. Julien’s proposal is exciting because it
completely captures the spirit of DM by eliminating not only the Lexicon, but also elimi-
nating any type of morphological process. However, having the base syntax generate
complex word structures involves a major revision to the assumed structures of the
syntax. This may be why Julien’s (2002) proposal has not caught on.
Another internal debate is over the nature of roots. The original proposal is that roots
() are generic. Pfau (2000, 2009) argued that by proposing that roots are specific to
certain concepts and to the VIs that realize them, we can account for a large amount of
speech error data. This means that the word cat realizes a specific root (CAT) that is
different from the root realized by the word dog (DOG). The gist of this proposal is that the
grammar is sensitive to the difference between cat and dog before PF, contrary to the origi-
nal proposal. A large proportion of the literature in DM seems to have assumed Pfau’s pro-
posal, using specific roots in derivations, effectively accepting this revision as main stream.
Yet, another debate arises from Pfau (2000) revision. Siddiqi (2006, 2009) proposes that
the revision of having specific roots allows content words to participate in competition (cat
and dog both compete to realize CAT and cat wins). Because content words can now com-
pete for insertion, the need for readjustment rules is evacuated from the grammar. Siddiqi
(2006, 2009) argues that DM should abandon readjustment rules entirely and proposes a
model of the framework without them. Similar to Julien’s proposal, Siddiqi’s reductionist
proposal is very much in line with the spirit of DM. However, while some researchers
have assumed Siddiqi’s model, readjustment rules are a contemporary realization of Word-
Formation Rules (Halle 1973), which have been around for nearly as long as there has
been formal morphology, and as such are firmly embedded in morphological tradition,
which may cause the revision to fail to gain traction.

9. Complexity of the Grammar


The major criticisms of DM come as a result of DM’s rejection of the Lexicalist Hypoth-
esis. The most cogent of these arguments comes from the complexity of the ‘morphol-
ogy’. Simply put: Syntactic structure is more complex than what we have historically
called morphological structure. Syntactic structures have long distance dependencies such
as the one created by a wh-word and its trace. This mandates the use of a grammar at

ª 2010 The Author Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.x
Journal Compilation ª 2010 Blackwell Publishing Ltd
538 Daniel Siddiqi

least as complex as a context-free grammar. However, the argument here is that morpho-
logical structure, particularly that which results from affixation, is much simpler – perhaps
even simple enough to be captured by a finite state grammar (see Beesley and Karttunen
2000). If the central claim of DM, that syntax and morphology are the same mechanism,
is true, we should expect to see words with the same long distance dependencies that we
see in phrases (or sentences). The claim of this argument is that we do not.22
There are two responses to this very good criticism. The first is to reject the claim that
there are no long distance dependencies at the level of the word. One counter example
comes through the combination of compounding and affixation noticed by Morris Halle
(and reported in Lasnik 2000). Halle noticed that center embedding occurs in the phrase
anti-missile missile (a missile designed to shoot down an anti-missile missile would be called
an anti-anti-missile missile missile and so on). Similar to that is the English voice and aspect
markers which contain a functional word and an affix attached to the head directly
governed by the functional word (be+ing, be+en, and have+en) (Chomsky 1957). Further-
more, any circumfix would necessarily entail a long distance relationship within the word.
However, these responses have their own criticisms. The first example involves two
different morphological processes (affixation and compounding) whose meanings are
dependent, but the processes likely are not. Similarly, the Chomskyan account of English
auxiliaries are necessarily an interface of syntax and morphology. Finally, while many
formal morphologists would disagree, the existence of circumfixes could conceivably be
rejected as the combination of a prefix and a suffix.
Another response to the criticism that there are no long distance dependencies at the
word level is to reject that the word is in fact a domain. From the DM point of view,
there is no domain that occurs between the morpheme and the phrase – that our percep-
tion of a word is not actually indicative of any actual domain23. Having rejected the word
as a domain, there cannot be a contrast between word level operations and phrase level
operations. Simply put, there are long distance dependencies in very large words, we just
happen to call them phrases.

10. Productivity
Another concern with DM is that it ultimately rejects morphology as a separate compo-
nent of the grammar. A problem with this rejection is that morphology contains a feature
that is not present in the syntactic or (arguably) phonological components: productivity.
Morphological productivity is the name for the fact that we prefer some morphological
processes or affixes more than we prefer others (or that some of these can appear in some
environments but are blocked in others). Aronoff and Fudeman (2004) have a fun exam-
ple of morphological productivity using the Harry Potter example of the word muggle.
English gives us several choices for the nominalizer that would describe the state of being
a ‘muggle’: -ity, -ness, -hood, or -dom. Mugglity seems to be easily eliminated because
muggle is not Latinate (or doesn’t sound Latinate). Muggleness is the default form as –ness–
is the default English nominalizer here. However, mugglehood and muggledom are also
acceptable to varying degrees because they seem more fantastic.
The choice of which of the aforementioned affixes to use is ultimately constrained by
two factors: (i) that the word does not appear to belong to an (extra-grammatical) group-
ing of words (words borrowed from Latin) and (ii) the less productive forms are accept-
able because they sound fancy. Ultimately, the choice of which affix to use for novel
forms is often not predictable! This is because word-formation, much more than syntax or
phonology, is a creative process.

ª 2010 The Author Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.x
Journal Compilation ª 2010 Blackwell Publishing Ltd
Distributed Morphology 539

This poses a problem for DM for several reasons. The first is that the tenets of formal
syntactic theory are based upon binary grammaticality, which productivity does not abide
by. The second is that it is very difficult to formalize productivity. The default argument
for DM is that less productive forms like -ity, -dom, and -hood have a specific list of stems
or classes of stems that condition their insertion (and that -ness goes everywhere else).
This argument works well for older forms, but struggles to account for novel forms.
However, this is not a problem unique to DM. This is ultimately a problem to any com-
pletely formalized model of grammar because it involves unpredictability. It may also just
exist outside the realm of synchronic grammar altogether and instead be part of diachrony.

11. Memorized Forms


One of the central arguments for the aggressive lexical decomposition stance of DM
comes from Marantz (1997a) against a Lexicalist position that some complex words
derived by the Lexicon are stored whole, creating new lexemes (that in fact, derivation is
the process of deriving one lexeme from another). The argument against storage of fully
formed words in a Lexicon comes from the necessarily complex meaning of complex
words and the necessarily simplex meaning of simplex words. Marantz uses the example
of transmission and argues that it cannot have the same meaning as kill.
The crux of the claim here is that, even if transmission refers to just one referent (a car
part), it is necessarily composed of trans-, -mit, and -ion (even if we do not know the
meaning of the component lexemes). This is arguably similar to our ability to use a phra-
sal idiom, such as the whole kit and kaboodle, whose pieces we do not know the meaning
of 24. According to the tenets of DM, transmission is absolutely not stored as one piece
anywhere other than the Encyclopedia, and at no point in a derivation are its component
formal syntactic features actually expressed as transmission until after spell-out.
This seems somewhat counterintuitive for several reasons. It is difficult to define which
complex words that are borrowed into a language are analyzed as complex and which are
analyzed as simplex: mountain, maintain, and pertain are all Latinate, but does that mean that
they are all also complex?25 Also, some words’ structures are not readily obvious: exasperate
and establish come to mind as words that seem to be complex but what the stems and
affixes are is not obvious either morphologically or semantically26. Finally, the presence of
one morpheme in several different words is not always clear: Do instruct, construct, obstruct,
and destroy all contain the same root?27 Do transfer, translate, and ferry?28
These issues ultimately arise because of the borrowing of complex words into a
language. English is especially prone to this issue because English borrows not only com-
plex words but stems and affixes as well29. These effects do not typically happen with the
more productive native Germanic affixes. A Lexicalist model would almost certainly
argue that some (if not all) derivationally complex words are stored whole in the Lexicon
no matter how they arrived (through borrowing, memorizing, or active derivation).
Borrowed complex words are troublesome for DM because DM must assume that it is
the component parts that are borrowed and stored, not the entire word.

12. Final Thoughts


For much of the work carried out within syntactic theory, the choice between Lexicalist
Minimalism and DM does not make all that much of a difference. However, for the
morphologist, DM is very exciting. DM allows for the exploration of phenomena that
exist on the border between the domain of the word and the domain of the sentence. In

ª 2010 The Author Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.x
Journal Compilation ª 2010 Blackwell Publishing Ltd
540 Daniel Siddiqi

addition to dealing with traditional morphological issues such as morpheme selection, dis-
tribution, and alignment (Noyer 1997; Haugen 2008), researchers within DM have
worked issues traditionally within syntax such as argument selection (Harley and Noyer
2000; Siddiqi 2009), affix lowering and do-support (Bobaljik 1994; Embick and Noyer
2001), and the distribution of verbal particles (Harley 2007). Because DM approaches
syntax and morphology (and their interaction) in a novel way, DM is a both a comple-
ment to existing formal theory and an avenue for new research.

Short Biography
Daniel Siddiqi is an Assistant Professor at Carleton University and earned his PhD from
the University of Arizona in 2006. Dan’s main research interests are morphology, syntax,
and formal linguistic theory, and he works mainly within the framework of DM.
Recently, Dan published Syntax within the Word, which approaches argument selection
and stem allomorphy through DM and proposes revisions to the framework. He lives
with his wife Julie in Ottawa, Ontario.

Acknowledgement
I would like to thank Ash Asudeh, Ida Toivonen, Rochelle Lieber, David Basilico, Jeff
Punske, and Jaime Parchment for their help on this paper. Of course, all mistakes herein
are entirely my own.

Notes
* Correspondence address: Daniel Siddiqi, Carleton University, Ottawa, ON, Canada. Email: daniel_siddiqi@
carleton.ca

1
Much of the original proposals to the theory presented here were originally presented in Halle and Marantz
(1993, 1994) and were previously summarized in Harley and Noyer (1999) and Siddiqi (2009).
2
I am assuming a model of morphology that assumes morphemes as the building blocks of words – a model which
is often called item-and-arrangement morphology. Two competing ways of handling morphology, item-and-process
morphology and paradigmatic morphology, make very different assumptions about the nature of words and are
completely incompatible with Distributed Morphology. Please see Aronoff and Fudeman (2004) for a detailed
discussion of these different ways of approaching morphology.
3
Distributed Morphology is a direct response to the Lexicalist tradition within Standard Theory and its descen-
dents (up to and including the Minimalist Program). Given this background, this paper, when referring to Lexical-
ism, unless otherwise stated, is specifically referring to Lexicalist Minimalism. However, Lexicalist Minimalism does
not have an explicit unified model of the Lexicon, so generally we have to fall back on ideas throughout the
Lexicalist tradition. It is not my intent to make it seem like there is consistent agreement among Lexicalist
morphologists as to the nature of the Lexicon.
4
Chomsky (1970) is by no means the only argument for treating morphology as a separate component of the
grammar. This argument is made quite often. Halle (1973) was the first to really articulate the Lexicon. See also
Aronoff (1994).
5
The weak version of the Lexicalist hypothesis is still maintained by mainstream Minimalism and is the variety of
the Lexicalist hypothesis that DM specifically disputes. The strong Lexicalist hypothesis, which argues that deriva-
tion and inflection are both separate from the syntax is assumed by theories like LFG and HPSG. These strong
versions of the Lexicalist hypothesis are quite different from the weak version of Minimalism, making the arguments
that DM usually makes to reject the lexicon less apt. Please see Scalise and Guevara (2005) for an excellent descrip-
tion of the development of the lexicalist hypothesis.
6
The resulting debate between supporters of these two principles comprises a major part of the literature of formal
syntactic theory and, unfortunately, is much larger than the scope of this paper.
7
Polysynthetic languages (such as Navajo or Swahili) are strong examples of languages realizing morphologically
what better studied languages typically realize syntactically. The order of morphemes in these languages are direct

ª 2010 The Author Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.x
Journal Compilation ª 2010 Blackwell Publishing Ltd
Distributed Morphology 541

parallels to accepted verb phrase syntactic structure and (easily arguably) cannot be considered memorized forms.
Please see Baker’s (2002) Atoms of Language for an excellent survey of contemporary ideas of polysynthetic languages.
8
Authentic itself is likely multi-morphemic, with -ic realizing a.
9
…and its passive voice marker… and its progressive aspect marker…
10
Not necessarily a prediction, but a selling point of underspecification models is that they include a much smaller
number of lexemes (or VIs) since a VI can be used in many different environments rather than a specific one. For
example, a highly specified model may need to hypothesize at least two different lexemes (but in reality probably
four) for the word are, while DM needs only one, which is so underspecified to be the elsewhere form.
11
A note about underspecification, AVM (attribute value matrix) models of syntax such as HPSG and LFG are
completely compatible with and often use underspecification. Underspecification is not unique to DM, but does
separate it from Lexicalist Minimalism.
12
As the elsewhere candidate, un is specified for no features relevant to this competition.
13
Whether or not there is a feature conflict is debatable. Masculine and singular might not actually be in the
syntactic environment as both can be treated as the lack of the marked feature (feminine and plural). I assume here
that the VIs are underspecified, but the syntax is not.
14
For simplicity, I am assuming in this rough sketch that there is no feature in the VIs corresponding to masculine
gender (in Spanish, which lacks a three-way gender distinction), rather that masculine gender is the absence of
[fem]. It can be easily argued that all the masculine forms contain a feature [masc] in Spanish, though the system
doesn’t need it and such an analysis would be less illustrative of competition in DM.
15
This rough analysis of Spanish determiners assumes that the determiners themselves cannot be decomposed.
There is a possible analysis where the -s that consistently means plural, the -a that consistently means feminine, the
-l that consistently means definite, and the -n that consistently means indefinite can all be pulled apart into separate
morphemes. While this is an exciting possible analysis, it is just outside the scope of this paper.
16
D here is used as short hand for the features in a feature bundle normally composing a determiner that are not
relevant to the discussion at hand.
17
It is also present in bretheren. Curiously, children is, historically speaking, double marked for plurality, as -r- is a
realization of plural.
18
cf receive vs reception.
19
Siddiqi (2009) points out that one strength of DM is that it predicts both the application of a readjustment rule
and the application of a secondarily licensed affix. (Table from Siddiqi 2009)
regular root and regular affix walk -ed
regular root and irregular affix hit -ø
irregular root and regular affix slep -t
irregular root and irregular affix mice -ø
20
Notably excluding Optimality Theory, which does not posit separate syntactic and phonological components.
21
Thanks to Meg O’Donnell for original design of this chart.
22
This argument was made to me by Ash Asudeh in response to a talk given in October 2008.
23
Similar to our perception of centrifugal force.
24
Thank you to Heidi Harley for this fun example.
25
Nope. Mountain is simplex, perhaps explaining why the -tain is pronounced differently.
26
A fun example I have of this is the phrase wedding reception, which I didn’t realize derived from receive until
I stood in a receiving line years after I myself had been married.
27
Yep.
28
No, but not in the way that you may think. Translate and transfer both have the same stem in Latin (-lat is an alter-
nate form of -fer), but it is hard to argue that they are the same stem in English. The stem fer- in ferry, on the other hand,
is Germanic and ultimately from a different PIE origin from the Latin -fer despite their similar meanings.
29
And uses those stems and affixes prolifically. Millward (1996) claims that more complex words created from
Greek and Latin stems and affixes exist in English today than existed in Classical Greek and Latin, the languages
they were borrowed from.

Works Cited
Aronoff, Mark. 1994. Morphology by itself. Cambridge: MIT Press.
——, and Kirsten Fudeman. 2004. What is morphology? Oxford: Blackwell.
Baker, Mark. 1985. The Mirror Principle and morphosyntactic explanation. Linguistic Inquiry 16. 373–415.
——. 1988. Incorporation: a theory of grammatical function changing. Chicago: University of Chicago Press.
——. 2002. The atoms of language. Oxford: Oxford University Press.
Beesley, Kenneth, and Lauri Karttunen. 2000. Finite-state non-concatenative morphotactics. Finite-state phonology:
proceedings of the 5th workshop of the ACL Special Interest Group in Computational Phonology (SIGPHON),
ed. by J. Eisner, L. Karttunen and A. Thériault, 1–12. Cambridge: Cambridge University Press.

ª 2010 The Author Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.x
Journal Compilation ª 2010 Blackwell Publishing Ltd
542 Daniel Siddiqi

Bobaljik, Jonathan. 1994. What does adjacency do? MIT working papers in Linguistics 22: the morphology-syntax
connection, ed by H. Harley and C. Phillips, 1–32. Cambridge: MIT Press.
——. 2002. A-chains at the PF-interface: copies and ‘covert’ movement. Natural Language and Linguistic Theory
20(2). 197–267.
Bresnan, Joan. 2001. Lexical Functional Syntax. Oxford: Blackwell.
——, and Sam Mchombo. 1995. The lexical integrity principle: evidence from Bantu. Natural Language and
Linguistic Theory, 13. 181–254.
Chomsky, Noam. 1957. Syntactic structures. The Hague: Mouton.
——. 1970. Remarks on nominalization. Readings in English Transformational Grammar, ed. by R. Jacobs and
P. Rosenbaum, 184–221. Ginn: Waltham.
——. 1995. The minimalist program. Cambridge: MIT Press.
Embick, David. 1997. Voice and the interfaces of syntax. PhD. Dissertation, Pennsylvania: University of Pennsylvania.
——, and Morris Halle. 2005. On the status of stems in morphological theory. Proceedings of going romance
2003, ed. by T. Geerts and H. Jacobs, 59–88. Amsterdam: John Benjamins.
——, and Rolf Noyer. 2001. Movement operations after syntax. Linguistic Inquiry 32(4). 555–95.
——, and ——. 2006. Distributed morphology and the syntax ⁄ morphology interface. Oxford handbook of linguis-
tic interfaces, ed. by G. Ramchand and C. Reiss, 289–324. Oxford: Oxford University Press.
Halle, Morris. 1973. Prolegomena to a theory of word formation. Linguistic Inquiry 4. 3–16.
——. 1997. Distributed morphology: impoverishment and fission. MIT working papers in linguistics: PF: papers at the
interface, ed. by B. Bruening, Y. Kang and M. McGinnis, 425–49. Cambridge: Department of Linguistics, MIT.
——, and Alec Marantz. 1993. Distributed morphology and the pieces of inflection. The view from building 20,
ed. by S. Keyser and K. Hale, 111–76. Cambridge: MIT Press.
——, and ——. 1994. Some key features of distributed morphology. MIT working papers in linguistics 21: papers
on phonology and morphology, ed. by A. Carnie, H. Harley and T. Bures, 275–88. Cambridge: MIT Working
Papers in Linguistics.
Harley, Heidi. 2007. The bipartite structure of verbs cross-linguistically, or why Mary can’t exhibit John her paint-
ings. Write-up of a talk given at the 2007 ABRALIN Congres in Belo Horizonte, Brazil, March 2007
——, and Rolf Noyer. 1999. State-of-the-article: distributed morphology. Glot International 4(4). 3–9.
——, and ——. 2000. Formal versus encyclopedic properties of vocabulary: evidence from nominalisations. The
Lexicon-encyclopedia interface, ed. by Bert Peeters, 349–74. Amsterdam: Elsevier Press.
Haugen, Jason. 2008. Morphology at the interfaces: reduplication and noun incorporation in Uto-Aztecan. Linguis-
tik Aktuell ⁄ Linguistics Today 117. Amsterdam: John Benjamins.
Jackendoff, Raymond. 1996. Semantics and cognition, handbook of semantics, ed. by S. Lappin, 539–59. Oxford:
Blackwell.
——. 1997. The architecture of the language faculty. Cambridge: MIT Press.
Julien, Marit. 2002. Syntactic heads and word formation, Oxford: Oxford University Press.
Kiparsky, Paul. 1982. Lexical phonology and morphology. Linguistics in the morning calm, ed. I. S. Yang, 3–91.
Seoul: Hanshin.
Lasnik, Howard. 2000. Syntactic structures revisited: contemporary lectures on classic transformational theory. Cam-
bridge: MIT Press.
Marantz, Alec. 1988. Clitics, morphological merger, and the mapping to phonological structure. Theoretical mor-
phology, ed. by M. Hammond and M. Noonan, 253–70. New York: Academic Press.
——. 1997a. Stem suppletion, or the arbitrariness of the sign. ‘Talk given at the Universite’ de Paris VIII.
——. 1997b. No escape from syntax: don’t try morphological analysis in the privacy of your own lexicon. Proceed-
ings of the 21st annual penn linguistics colloquium (U. Penn Working Papers in Linguistics 4:2), ed. by A. Dimi-
triadis, L. Siegel, C. Surek-Clark and A. Williams, 201–25. Philadelphia: University of Pennsylvania.
Millward, C. M. 1996. A biography of the english language, 2nd edn. Fort Worth: Harcourt Brace.
Noyer, Rolf. 1997. Features, positions and affixes in autonomous morphological structure. New York: Garland.
Pfau, Roland. 2000. Features and categories in language production. Ph.D dissertation, Pennsylvania: University of
Frankfurt.
——. 2009. Grammar as Processor: A Distributed Morphology account of spontaneous speech errors. Amsterdam:
Benjamins.
Pollard, Carl, and Ivan A. Sag. 1994. Head-Driven Phrase Structure Grammar. Chicago: University of Chicago Press.
Prince, Alan, and Paul Smolensky. 1993. Optimality theory: constraint interaction in generative grammar. Rutgers
University Center for Cognitive Science Technical Report.
Scalise, Sergio and Emiliano Guevara. 2005. The lexicalist approach to word-formation and the notion of the lexi-
con. Handbook of word-formation, ed. by P. Štekauerand and Lieber R, 147–187. Netherlands: Springer.
Siddiqi, Daniel. 2006. Minimize exponence: economy effects on a model of the morphosyntactic component of the
grammar. PhD dissertation, Arizona: University of Arizona.
——. 2009. Syntax within the word: economy, allomorphy, and argument selection in Distributed Morphology.
Amsterdam: John Benjamins.

ª 2010 The Author Language and Linguistics Compass 4/7 (2010): 524–542, 10.1111/j.1749-818x.2010.00212.x
Journal Compilation ª 2010 Blackwell Publishing Ltd

S-ar putea să vă placă și