Sunteți pe pagina 1din 97

Bracketing paradoxes in Italian

0. Introduction
0.1 Bracketing in linguistics 0.2 Bracketing paradoxes 0.3 Aim of the work 0.4 Methodology

1. Prefixal relational adjectives


1.1 Aim of the chapter 1.2 Binarism in word formation 1.3 Parasynthesis 1.4 Prefix-suffix combination 1.5 Intermediate steps 1.6 A constructionist approach 1.7 The role of prefixes 1.8 The role of adjectives 1.9 Novel PRAs 1.10 Word Internal Anaphora 1.11 Transcategorization 1.12 False heads 1.13 Alternation
1.13.1 Phrasal constituents 1.13.2 Ambiguity 1.13.3 Postnominal position 1.13.4 Non-relational usages 1.13.5 Factorization 1.13.6 Specialised meaning 1.13.7 Language-specific factors

1.14 Processing issues 1.15 Conclusions

2. Bracketing paradoxes in phrasal morphology


2.1 Aim of the chapter 2.2 Phrasal bracketing paradoxes 2.3 Language-specific factors 2.4 Phrases as lexical units 2.5 Semantic issues 2.6 Differences between NA and NPP constructions 2.7 Survey of phrasal nouns
2.7.1 Academic 2.7.2 Sport 2.7.3 Music 2.7.4 Locative/collective nouns

2.8 Adjectival phrasal BPs


2.8.1 Extension 2.8.2 Hyphenation 2.8.3 Factorization 2.8.4 APP costructions

2.9 Pragmatic constraints 2.10 A constructionist account 2.11 Processing Issues 2.12 Novel phrasal BPs 2.13 Conclusions

Final remarks

0. Introduction

0.1 Bracketing in linguistics


Bracketing has the purpose of making the structure of a complex form explicit. Be it a syntactic or a morphological item, bracketing allows to represent the hierachical configuration of its constituents, based both on linguistic analysis and on speakers' perception: expecially in morphology, occasionally the two plans suddenly diverge because of a reanalysis operated by a speaker who interprets a complex word as derived by different means than those which actually produced that result: famous is the case of [[hamburg]er] reinterpreted as [[ham][burger]]. In the domain of morphology, bracketing is a notational tool aiming at capturing the internal complexity of words, but it is not cross-linguistically applicable: nonconcatenative languages, i.e. those whose word formation does not involve stringing morphemes together, do not allow the recognition of single isolable units which unitarily contribute to the derivation process and thus cannot receive a standard bracketing. Arabic, for instance, employs triconsonantal roots which are modified through their vowel pattern, e.g. the root k-t-b derives in /kita:b/ 'book', /katab/ 'he wrote' etc.: it would not be possible to bracket such derived words isolating their constituents.

0.2 Bracketing paradoxes


What are bracketing paradoxes? In "bracketable" languages, a number of expressions have been labeled as paradoxical because for various reasons two incompatible bracketing were assumed. This was often ascribable to the mismatch occurred between two representational levels, namely morphosyntax and semantics (most of the times) or morphosyntax and phonology: let us consider the often-cited examples antifebbrile 'antifever', trasformational grammarian and unhappier: It. [anti[febbrile]] vs. [[antifebbr(e)]ile] 'antifever' [[trasformational] [grammarian]] vs. [[trasformational grammar]ian] [un[happier]] vs. [[unhapp(y/i)]er]

In [Prefix-Noun-Suffix]A relational adjectives like It. antifebbrile (Montermini 2008:205), although the semantic instruction should lead to a [[Pre][N]][Suf] bracketing, the Noun-Suffix string is often identical to the adjective already available in the lexicon, even though the affix has scope over the complex base (whether attested or not). The expression transformational grammarian does not have the same morphosyntactic properties of old grammarian, and two bracketings can be assumed, each meeting the demands either of semantics (in [[trasformational grammar]ian] the relation with the base transformational grammar is explicit) or of morphosyntax ([[trasformational] [grammarian]] reflects the structure of an adjective-noun pair and the fact that grammarian is already established in the English lexicon). The phonological restriction on the attachment of the English comparative suffix er (which may attach to monosyllabic adjectives and a small class of bisyllabic ones) is not compatible with the meaning of forms like unhappier, where the affixation of the base (adding one or more syllables) does not prevent it from conforming to the standard phonological pattern.

0.3 Aim of the work


The present work will deal with two kinds of bracketing paradoxes in Italian, namely those posed by prefixal relational adjectives (e.g. antifebbrile 'antifever') and phrasal constructions (e.g. chitarrista elettrico 'electric guitarist'). An analysis of their properties will be carried out and an attempt will be made to explain their "paradoxical" nature. A critical assessment of the literature on bracketing paradoxes will help focusing the theoretical puzzle and trying to define the limits of the phenomenon. The focus on Italian data will offer a rather different perspective than the English-based traditional approach. Rather than considering bracketing paradoxes as a "bug" in linguistic theory, the present work will try to offer a contribution to the explanation of their natural occurrence in everyday language. Although the analysis of bracketing paradoxes in Italian will not be strongly theoryoriented, the assumptions encountered in the generative-transformational literature will be generally discarded or ignored, and preference will be given to some tenets of the Construction Grammar framework, expecially insightful with respect to the boundary between syntax and the lexicon (see chapter 2).

0.4 Methodology
A methodological choice was drawing on data coming from the Web. As a very large source of linguistic data, the Web allows to tackle the problem of data sparseness, expecially in word formation studies where the focus of a study can be represented by rare patterns surrounding the core standard, and even hapax can provide interesting insights and shed light on the rest of the data. As pointed out by Baroni & Ueyama (2006:209), "the Web contains a huge quantity of textual data for an ever increasing number of languages, it contains many different genres and specialized texts, and, of course, it is a "renewable" source of language, as long as people post new data". Whenever sufficient data was available, the present work drawn on web corpora (in addition to La Repubblica corpus, Baroni et al. 2004). Web corpora have their own advantages and disadvantages. Issues related to size (crucial in the light of Zipfian properties of language) and speed of the building process speak in favour of them. Moreover, the Web contains spontaneous written texts which display characteristics of oral communication, and as its usage for archival purposes spreads, it is likely to contain more and more instances of traditional written genres. Major disadvantages are the "noise" (non-linguistic material, duplicated documents) and the control over content. As noticed by Baroni & Ueyama (2006:210), it is not correct to refer to such problems as problems of Web corpora: "rather, they are problems of large corpora built in short time and with little resources, and they emerge quickly with Web corpora since the Web makes it possible to build 'quick and dirty' large corpora". The corpus used in the work is ItWaC (Baroni et al. 2009), a 2 billion word corpus constructed crawling all .it web domains. ItWaC was part-of-speech tagged with TreeTagger. In order to enhance the understanding of how speakers cope with a range of constructions that lie at the heart of the phenomenon studied, native speakers were occasionally asked to provide judgments of sentences containing critical words or paraphrases for comprehension.

1. Prefixal relational adjectives

1.1 Aim of the chapter


Bracketing paradoxes can be found at word level. This chapter will narrow its scope by looking at prefixed relational adjectives: why can we find instances of bracketing paradoxes in this construction domain? How can we relate this to word formation theory? Are there constraints at play on such complex words' coinage and usage? Prefixal relational adjectives (henceforth PRAs) represent a productive word formation pattern in which discrepancies arise between compositional meaning and morphosyntactic structure. Let us consider the already mentioned Italian adjective antifebbrile: despite its semantic interpretation as a suffixally derived compound, the adjectival suffix is assumed to pertain to the nominal base rather than to the complex form, on the basis of the independent attestation of the embedded relational adjective febbrile. Further arguments come from allomorphic adjectivation, which sistematically overrules large scope affixation, expected on a semantic basis. it. antifebbrile lit. anti-fever.adjsuff [[anti] [febbr(e)]][ile] [anti] [[febbr(e)][ile]] lit. pre-alphabet.adjsuff [[pre] [alfabet(o)]][ico] [pre] [[alfabet(o)][ico]] 'antipyretic'

it. prealfabetico

'prealphabetic'

There has been little agreement on the morphological interpretation of the pattern, as well as on the general assumption that a mismatch between compositional meaning and morphosyntactic structure is at play. After reviewing some of the main theoretical issues involved in the attempts to describe and explain the phenomenon, and comparing them with corpus data from Italian, an account will be provided of some of the factors which can be held responsible for the pervasiveness of the pattern in World languages. A more in-depth corpus-based analysis of specific patterns in Italian will then try to answer some questions on prefixed relational adjectives' usage contexts and alternation with similar constructions.

A few preliminary observations on binarism will outline a brief picture of bracketing controversies in the word formation domain.

1.2 Binarism in word formation


At least since Aronoff (1976) and Booij (1977) a binary approach is generally defended in morphological processes. Complex words containing more than two cooccurring morphological elements are assumed to mirror binarism in syntactic structures. Such words are thus expected to correspond to either a [A+[B+C]] or a [[A+B]+C] structure. The Binary Branching Hypothesis states, in fact, that morphological structures are (regardless of their complexity) binary (see the figure below, taken from Scalise & Guevara 2005:163).

Such theoretical claim, however, cannot straightforwardly account for a range of morphological structures in which some evidence speaks against the possibility of a neat binary analysis. Predictably, such unclear cases pose particular bracketing difficulties. If one rejects the Separation Hypothesis (Beard 1995), claiming that morphosyntactic and morphophonological processes are two different analytic levels, and tries to outline non-mismatching representations, a range of phenomena appears in which assigning constituency is problematic. Guevara (2006:3) offers a review of such word formation cases. Rainer (1989) highlights that some kinds of coordinate compounds like anglo-italoamerican cannot hold a binary representation. If we move from compounding to derivation, however, exceptions become rarer. Synaffixes are conceived as complex affixal clusters (consisting of two or more analyzable formants operating simultaneously on the morphological base) which behave as single units both from a functional and a semantic point of view. The best-known

Italian synaffix is probably -istico, formally consisting of the two independent suffixes -ista and -ico. Let us consider the examples below: pensionistico 'related to pensions' paesaggistico 'related to landscape' infermieristico 'related to nurses' ?pensionista ?paesaggista ?infermierista

Apparent intermediate formations in -ista are either unattested (pensionista, infermierista) or devoid of any derivational relationship with the -ico adjectival forms: paesaggistico means 'related to landscape' (paesaggio), not to landscape architect (paesaggista). It seems reasonable to argue that sometimes the adjective in -istico receives a mixed reading (e.g. see It. giornalistico, both 'related to newspaper', giornale, and 'related to journalist', giornalista) As highlighted by Guevara (2006: 5), in fact, the affix-cluster -istico is "functioning as a single suffixing unit which builds binary structures, despite the fact that three recognizable morphological formants seem to be joined at once". The possibility of assigning a non-binary structure like N+Suffix1+Suffix2 is thus discarded. Major theoretical approaches to -istico-like adjectives have either recognised it as an independent complex suffix (licensing the binary N+[complex suffix] representation, see Bauer 1990), or maintained the status of the individual suffixes (see Scalise 1990) accepting the intermediate step in the derivation as a possible base even if unattested (bringing to a [[N+Suffix1]Suffix2] bracketing). Another major challenge to binarism in word formation is represented by parasynthesis, i.e. concomitant prefixation and suffixation. Some accounts of parasynthesis are ternary (e.g. see Serrano-Dolader 1999), while binary representations differ in the proposed hierarchical structure (summarising, [[AB]C] vs [A[BC]] bracketings, see for instance Corbin 1980 and Scalise 1994). Circumfixation (or discontinuous affixation) can superficially be regarded as the same phenomenon, i.e. the simultaneous attachment of two affixes, even though some approaches make a "principled distinction [...] based on a fundamental criterion: only in circumfixation do the interrupted parts form a single morphological constituent" (Guevara 2006: 8). As already mentioned, if one agrees with the Separation Hypothesis, two non-isomorphic representations can be stipulated which make these cases only apparent counterevidence to binarism (see figure below, ivi), being a) the morphosyntactic level and b) the morphophonological one.

Parasynthesis and discontinuous affixation will be addressed later with respect to their adequateness for explaining prefixal relational adjectives. Globally speaking, the mentioned word formation cases in which controversies arise about hierarchical structure do not pose a strong challenge to a binarist approach to morphology. Prefixed relational adjectives, as the next section will show, display a formal structure mismatching with the expected semantic representation. Binarism, however, is not challenged since the two representational levels are both binary. It will be suggested that the formal constituenthood could also be influenced by a general preference twards right-branching constructions.

1.3 Parasynthesis
The traditional approach to PRAs is a parasynthetic analysis, i.e. complex forms are conceived as the outcome of a derivational process based on the simultaneous application of a prefix and a suffix on a word base. Since none of the possible intermediate derivational steps (Pref+N or N+Suf) actually occurs, parasynthesis generally represents a challenge in a bracketing perspective: one cannot assign neither an [A[BC]] bracketing, nor an [[AB]C]. giallo barca > in+giall(o)+ire > im+barc(a)+are 'to yellow' 'to embark' *ingiallo, *giallire *imbarca, *barcare

Italian PRAs have received such treatment because of the form/meaning mismatch which did not allow them to project structural hierarchy onto morphosyntactic bracketing. It must be noticed, however, that PRAs differ from traditional parasynthetic cases in an important aspect: whereas their binary form are not attested (as can be seen from the above examples, they are often prefixed verbal constructions), PRAs' constituents are indeed attested, but since structural mismatch hampers a proper bracketing, a flat structure is stipulated.

Tollemache (1945:112, my translation) discusses Latin and Italian cases which "must be judged parasynthetic regarding meaning". He points out that cisalpinus [...] does not mean "an alpine who is beyond". [...] However, if we look at the form, we have to admit that cisalpinus is cis+alpinus, transpadanus trans+padanus; i.e. adjectives already existed before compounding; these compounds are thus logically parasynthets, whereas formally ordinary compounds. [Among Italian examples] there are cismontano, internazionale, transiberiana, ultramontano. Other analyses were based on the notion of circumfixation (or discontinuous affixation) formulated by Booij (1977:32), which may apply to cases where it is difficult to recognise consistent word formation steps (in a rule-based approach to morphology) and the solution can consist in stipulating that two affixes are actually two segments of the same affix: The one affix a rule hypothesis is the hypothesis that WF-rules add only one affix to a word. Generally, this hypothesis seems to be correct. The only counterexamples are such words as gebergte (mountain range) < berg (mountain) and geboomte (trees) < boom (tree). But ge-te can be considered a discontinuous affix because ge- and -te do not contribute individually to the meaning of the complex word. In the domain of PRAs prefix and suffix are thus conceived as a unit which applies to a noun base, so that the following examples would be analysed as follows: circumequatoriale: circum__ale + equatore antiparassitario: anti__ario + parassita antisismico: anti__ico + sisma The circumfixal approach suffers from similar weaknesses to the ones found in the parasynthetic analysis. As pointed out by Guevara (2006: 12), circumfixation and parasynthesis are superficially "the same phenomenon: simultaneous attachment of two morphological elements, one to the left and one to the right of the base". Only in circumfixation, however, "the interrupted parts form a single morphological constituent". The most important drawback is probably the fact that the semantics of

10

every would-be circumfix would be based on the exclusive contribution of the first segment (i.e. the prefix). This means that many circumfixes with the same "prefix" would share the same semantics, see for instance

anticlericale, antisolare, antiparassitario, antiberlusconiano, antisismico, antifungino, antigovernativo, anticompromissorio, antiaugusteo, antivaioloso, antiterrestre etc. and that the same "suffix" (final segment of the circumfix) would in turn correspond to different meanings:

adimensionale, anticlericale, circumequatoriale, cisretinale, codecisionale, denominale, disfunzionale, exconiugale, extracontrattuale, internazionale, pluridirezionale etc.

Such analysis also fails to consider synonymous constructions where no suffix is attached to the base.

1.4 Prefix-suffix combination


As noticed by Iacobini (2004), one major drawback of this approach is that this does not explain why the adjectival suffix of the parasynthetic adjective is always identical to that of the nonprefixed one: why, assumed that costituzionale and rivoluzionario are independently attested adjectives, are anticostituzionale or antirivoluzionario licensed unlike anticostituzionario1 or antirivoluzionale? Parasynthesis applied to PRAs is also inconsistent with the fact that the same prefix may combine with different suffixes, which makes such description antieconomic. I will now provide corpus-based evidence on prefix-suffix combination in Italian PRAs. There is no general agreement on the definition and enumeration of Italian prefixes (see Montermini 2008:25). Corpus data will be limited to the list of prefixes provided by Iacobini (2004). The table below summarizes the attested combinations (195) and provides a random sample for each one. It is apparent that a parasynthetic approach becomes antieconomic when dealing with a wide range of combinations, since it should maintain that each one corresponds to a single word formation rule (see Booij 1977's 1. Words preceded by are possible but unattested words.

11

"one affix a rule" hypothesis). It is worth noticing that a large number of prefix-suffix combinations also strongly challenges the circumfixal approach, which should stipulate a high number of affixes. Moreover, this evidence moves out the theoretical possibility that adjectival attestation could be explained in terms of prefixal selection (i.e., prefixes selecting specific adjectives according to their derivational affix). The table includes a selection of productive prefixes and denominal adjectival suffixes. Data come from complex queries performed on La Repubblica corpus (Baroni et al. 20042,) and ItWaC (Baroni et al. 2009), consulted online through Sketch Engine (Kilgarriff et al. 20043). Although such corpora are not morphologically tagged, individual queries were performed specifying POS, initial and final segment of the word. The forms obtained were then evaluated according to their correspondence to the morphological configuration searched for (i.e. words containing segments homophonous to prefixes and/or suffixes were discarded). Additional words were included if at least 10 hits were available on the web (via Google Search).

2. Available online at http://dev.sslmit.unibo.it/corpora/corpora.php 3. Available online at http://www.sketchengine.co.uk/

12

TABELLA

13

1.5 Intermediate steps


Other approaches to PRAs formation claimed that a ternary structure cannot account for them and that intermediate steps in the derivation must be assumed. According to Sgroi (2007), for instance, the adjective sottomarino 'submarine' is derived from an unattested compound sottomare 'under-sea'. Although such analysis does not treat sotto- as a prefix, it can be extended to PRAs in terms of an intermediate unattested derived word. The issue of why the suffixing form is identical to the one applying to the simplex word (sottomare+ino, mare+ino) remains unexplained. Corbin (1999) analyses PRAs' status along the debate about prefixal categorization. Maintaining that PRAs are converted prefixed nouns speaks against the "dominant conception" that prefixes have no categorizing capacity. She points out that adjectives formed via prefixation on nominal bases often display a suffix ending which has only a category-marking role. A semantic analysis is invoked in order to discard other possible structural analyses: so antigrippal 'antifever.Suff', which seems segmentable in anti+grippal, cannot be semantically explained starting from a structure combining these two constituents given that its sense can be compositional only in relation with antiand grippe (vaccin antigrippal 'antifever.Suff vaccine' has the same meaning and usages of vaccin antigrippe 'antifever vaccine') Corbin introduces the notion of "paradygm integrator" as a purely class-marking suffix with no constructional function in the complex word, and provides an account of PRAs inspired (as highlighted by Montermini 2008:203) to the Optimality Theory. According to such model this word formation pattern would be constrained by two partially conflicting conditions at play in the grammatical output: (i) a Semantic Conformity Constraint, on the basis of which the optimal form of an affixed word corresponds to the concatenation of every form contributing to the definition of its meaning; (ii) a Categorial Conformity Constraint, predicting that the optimal form of an adjective formed via affixation has a suffix-like ending. Let us consider the above mentioned examples antigrippeAdj and antigrippalAdj: Corbin states that while the former represents the "victory" of the Semantic Conformity Constraint, the latter represents that of the Categorial Conformity Constraint.

14

A question that arises naturally from such formulation is up to what extent one can predict the outcome of the interplay between constraints, and thus the morphosyntactic configuration of the word, on the basis of a range of factors: semantic, pragmatic, or related to usage contexts. What role do different factors play in the "victory" of either semantic or categorial constraints? The issue will be addressed later. According to Corbin (1980) seemingly parasynthetical PRAs can be analysed assuming a two-step derivation: the prefixation of the nominal base and the suffixation through copy of the relational adjectival suffix of the base noun. Corbin criticises the parasynthetical approach's assumption that no other word formation process can be described: this "dehierarchizing" (Corbin 1980: 190) view of morphology cannot effectively cope with the choice of the affixes (unless, as previously shown, by stipulating many specific word formation rules) and cannot explain why suffixed adjectives alternate with non-suffixed ones. As insightfully highlighted by Corbin (ibidem: 201, my translation), The adjective antialcoolique is the exact equivalent, on a semantic and syntactic basis, of the unattested adjective antialcool. This equivalence does not exist, in turn, between alcooliqueAdj and alcoolN. Corbin's solution saves both morphosyntactic hierarchy and offers an explanation for suffixal selection. According to the proposed pattern (Corbin 1980: 207) a copy of the nominal suffix is stipulated to occur facultatively in the adjectival derivative: this would explain why it is the same that would apply to the simplex base root. [Pref [X]]A/N +copy]A/N This notation can be applied to five cases of prefixal adjectives with different morphosyntactic and categorial properties. Let us consider the following Italian examples (extracted from ItWaC):
[Pref [X]]A [Pref [X]]N [Pref [X]+copy]A [Pref [N]+copy]N [Pref [NP]]A #669304525 Questi i farmaci antifebbre. Per combattere la febbre #1137876106 l' Artemisia annua come efficace antifebbre; un altro #38452428 dotazione personale di farmaci antifebbrili e antidolorifici #243768054 38ricorrere ad un antifebbrile per uso pediatrico #1076940707 antiparotite, antirosolia, antifebbre gialla. Rinvio per

15

Corbin's approach, as already mentioned, has the advantage of highlighting the nominal nature of prefixal adjectives and the fact that suffixation can be said optional. As noticed by Montermini (2008:204), however, it is redundant in the assumption that every suffix in the mentioned forms is a homophone suffixoid with categorial-only functions. It is more economical to assume that PRAs are based on the attested adjectives and that a mismatch occurs between structure and meaning. As can be seen from Italian data extracted from ItWaC, some examples offer counterevidence for a model of PRAs based on the copy of the suffix. Cases like the following cannot be straightforwardly ascribed to a mere process of suffixation, and probably involve the access to more complex paradigmatic relationships (see the following example for a case of "empty morpheme"): #85842728 obbiettivo l'insegnamento post-obbligatorio piuttosto che la gestione Corbin (1980:206) claims that her model can account for the impossibility of attaching a suffix to a NP / multiword expression, given that no affix can have previous relationships with it. Some Italian examples seem to contradict such generalization, see for instance Gaeta (2006) for a collection of affixed phrases (with varying degrees of lexicalization): doppiolavorista who is having a double job, stileliberista crawlswimmer etc. Moreover, the claim that suffixes have only categorial functions may not apply to some cases of suffix alternations. The issue will be addressed later.

1.6 A constructionist approach


The Costruction Grammar approach to word formation has the main merit of providing a theoretical and notation model which can fruitful deal with the issues related to intermediate steps. In particular, Booij (2007) suggest under this theoretical framework a thorough revision of debated word formation issues. The main point put forth by Booij is the adoption of a schema-based approach to morphological processes, rather than a purely rule-based. Word formation patterns are based on "abstract schemas that generalize over sets of existing complex words with a systematic correlation between form and meaning" (Booij 2007:34). Schemas are productive, since they specify how new complex words can be created, by describing partially-filled patterns, i.e. in which some spots are occupied by specified items and some can vary according to each instantiation. From a costructional viewpoint, it is possible to get rid of the notion of intermediate step. This allows the description of productive word formation

16

patterns to ignore the nonattestedness of items embedded in complex words. A classic problematic word like blue-eyed is dealt with by assuming that a unification between two schemas occurred, namely between the [A+A]A compounding schema and the denominal adjectival derivation in -ed, which are productive constructions in English: Compounding: [A+A]A Derivation: [N+ed]A [A + [N+ed]A]A having N with property A Although eyed is not attested (this specific issue was explained by Ackerman & Goldberg 1996 in terms of "Non Redundancy" constraints, since it would be unfelicitous from a pragmatic viewpoint to have a word describing the property of having a body part - modifying it would rather be informative: four-legged, right-handed etc.), the schema is nonetheless solid and describes a wide number of possible instantiations. The same approach allows a treatment of parasynthetic words which, although recognising the simultaneity of the affixation process (the concept was introduced by Darmesteter 1877:129) fully specifies the word internal underlying hierarchy: Booij (2005:38) describes the Dutch [on-A]A + [V-elijk]A = [on[[V-elijk]A]A unification pattern which, fed by deverbal adjectives like beschrijf-elijk 'describable' > on-beschrijfelijk undescribable represents the formation of words like ongelofelijk 'incredible' with no V-elijk base (gelofelijk). Gaeta (2006) offers an analysis of complex words like terzomondiale 'of the Third World' in which a conflation of two schemas is assumed in order to deal with the effects of paradigmatic relationships. Gaeta (2006:4) adopts a box notation representing each internal constituent (see figure below). [X X]N [[N]Nale]A [X [[X] ale]A]]A terzomondo 'Thirld World' mondiale 'mundial' terzomondiale 'of the Third World'

17

Gaeta claims that the same conflation process can be assumed with Italian parasynthetic verbs: [im[mettere]V]V to put in mettere to put [s[caricare]V]V to unload caricare to load Can the same constructionist approach be applied to PRAs? As a starting point, it is possible to state that a unification pattern could better deal with the status of the internal relational adjective, e.g. settimanale in infrasettimanale. The critical point to consider is the Pref+N/Pref+N+Suf alternation, which forces the description to opt for either one or the other. Gaeta claims that the prefixal+suffixal unification pattern in parasynthesis assigns a priority to the suffixal schema - "this is justified by the occurrence of word pairs displaying different prefixes [like] infornare/sfornare to put into/to take out of the oven"(Gaeta 2006:6). This claim is hard to extend to PRAs, since an alternation is found between synonymous suffixed and non-suffixed forms, and even (rare) word pairs displaying different suffixes are attested. If we take a look at the overall configuration of Italian prefixed adjectives, three kinds of words displaying different properties can be taken into account on a first approximation. The first one is represented by qualifying adjectives compositionally modified by prefixes, involving no scope ambiguity. ultraleggero semiserio stracontento 'ultralight' 'semiserious' 'very very happy'

18

These forms are grounded on a construction which allows a straightforward modification of the embedded adjective (be it complex or simplex). Below an outline of the progressive instantiation of the pattern: [ultra[A]A]A 'extremely A', [semi[A]A]A 'half A' and [stra[A]A]A 'exceedingly A' schemas were taken as examples:

When it comes to PRAs based on a headed construction, e.g. vicepresidenziale 'vicepresidential', microsismico 'microsismic', macroregionale 'macroregional', a [Pref[A]A]A analysis reveals its inadequacies. On the basis of the word meaning it is challenging to assume a stepwise derivation Adj > Pref+Adj based on the recognisable simplex adjective, namely presidenziale, sismico or regionale for the above mentioned examples. The prefixal subcategorization frames would be violated (Italian prefixes vice-, micro- and macro- attach, in fact, to nouns only), and no consistent semantics could be derived. Such adjectives are rather based on a Pref+N word, in our case vicepresidente 'vicepresident', microsisma 'microquake', macroregione 'macroregion'. What is at issue here, as already stated, is the formally easy bracketing Pref+Adj, at odds with the semantic structure of the word. A conflation schema could here account for the unification of [Pref[N]N]N and [[N]Nale/ico]A: [Pref[N]N]N [[N]N ale/ico]A [Pref[[N]N ale/ico]]A vicepresidente, microsisma, macroregione presidenziale, sismico, regionale vicepresidenziale, microsismico, macroregionale

19

With respect to this restricted class of PRAs, however, another argument can be put forth which does not apply to all PRAs. Hoeksema (1985) proposes that many derivational processes can be conceived as head operations, i.e. processes involving only the head of a word although having scope on the whole. In the words of Beard (1998), English derived verbs exhibit the effect of a head operation in maintaining their conjugations even when serving as a base in a derivate. The past tense of understand is understood, and that of overdrive is overdrove. This seems to indicate that although past tense has scope over the entire derivation (or compound) in these instances, morphology applies strictly to stand and drive, respectively; otherwise, we might expect the past tense to be *understanded and *overdrived. It seems fairly obvious to assume that it is not possible to extend the head-based approach to prototypical exocentric PRAs. Let us consider some examples: antifebbrile infrasettimanale postbellico 'antifever' 'midweek' 'postwar'

None of these adjectives can be traced back to a headed construction (although the prefixes attached occasionally enter headed words, see e.g. antieroe 'anti-hero'). They represent a class of PRAs in which the prefix has a relational function rather than a modifying one: antifebbrile, in its adjectival form, means 'related to something against fever', infrasettimanale 'in the middle of the week' etc. Whereas in headed PRAs the bracketing paradox could be motivated by a head operation, nothing seems to justify a corresponding bracketing for headless PRAs. We should remember that such words display an alternation with non suffixed forms: [Pref[N]N]A antifebbre vs vs [Pref[[N]N RAsuff]A]A antifebbrile

How can we explain that three kinds of prefixed adjectives with different semantic properties are susceptible of the same formal analysis? Rather than focussing on this question, it could be more fruitful to ask instead what factors could prevent this from happening.

20

A constructionist perspective allows us to hypothesize that an analogical extension occurred of the schemas involved along the just sketched continuum. If we exclude the ultraleggero kind of adjective, where the [Pref[A]A]A construction is unproblematic, and turn to the head-based adjectives vicepresidenziale, microsismico, macroregionale some factors can be held responsible for motivating the attraction towards the already available derived adjective: first of all, the possibility of the already introduced derivational head-operation. It should be added that it parallels other headbased morphological processes, like inflection. Consider for instance headed compound pluralization in Italian (see Stump 1991 for examples of diminutives' head pluralization): capostazione (left-headed, pl. capistazione) 'stationmaster' terremoto (right-headed, pl. terremoti) 'earthquake' cassapanca (dvandva, pl. cassepanche) 'blanket chest' Moreover, the meaning of the embedded adjective often has a closer semantic relationship with the final PRA (although this aspect largely depends on the individual prefixal semantics) if compared to exocentric PRAs. We can generalize that, unlike the latter, the meaning of the prefixal adjective is often in a hyponymic relation with the simplex adjective (sismico/microsismico, regionale/macroregionale etc.). We should also be aware that, unlike exocentric PRAs, such forms do not license an alternation with non suffixed forms, because their non affixed bases (always attested, as nouns: see e.g. microcredito 'microcredit') would not be categorially undetermined like exocentric ones (if we maintain that prefixes are not category-changing). A similarity lies, however, on the fact that both constructions, due to their prefixal subcategorization frame, do not allow the idiosyncratic meanings of their internal adjectives to emerge. In order to show this, let us have a look at the possible interpretations of prefixed words embedding adjectives ambiguous between a relational and a qualifying reading: if we consider the adjective febbrile 'feverish' [both literal and figurative], a non-attested construction like ultrafebbrile would arguably receive both a relational and a qualifying interpretation, namely either 'very feverish' or 'beyond fever', whereas the qualifying sense of febbrile could not be selected in (the unattested) microfebbrile or postfebbrile. In order to account for the pervasiveness of the [Pref[A]A]A pattern, we can hypothesize that this schema was strengthened along the continuum leading from ultraleggero to postbellico, so that Italian speakers can rely on a range of constructional idioms (in the sense of Jackendoff 2002, constructions with lexically empty slots) like those below:

21

[pre[[N]N RAsuff]A]A'before N' [post[[N]N RAsuff]A]A 'after N' [anti[[N]N RAsuff]A]A 'against N' [pro[[N]N RAsuff]A]A 'in favour of N' ... The meaning conveyed by different prefixes is in my opinion crucial in the interpretation of the complex word. As far as we know, PRAs are the outcome of a structurally underspecified pattern. This means that PRAs display a level of structural complexity which is inferior to what one would expect on a semantic basis. It is reasonable to hypothesize that such mismatch is due to economical factors: the language system does not need a tripartite Prefix - Noun - Affix construction where a less complex Prefix - Adjective string can accomplish the same communicative functions. What are at issue here are the reasons behind such underspecification: a first answer could be that the semantics of prefixes itself allows the word to express a certain meaning without enriching its morphosyntactic structure. The focus of the following section will be, in fact, providing a clearer picture about the role played by prefixes in the domain of PRAs.

1.7 The role of prefixes


From the observation of the patterns considered above, it is possible to sketch a structured continuum based both on prefixal subcategorization frames and the meaning expressed by each prefix. This picture can help identifying the reasons for the spreading of the constructionist schema. If we take two variables into account, namely selection of nouns vs adjectives and temporospatial vs negative/alterative meanings, a clearer picture emerges of the intersection between sets (see figure below). We can thus account for the different behaviour of vicepresidenziale-like vs postbellico-like constructions: although both are noun-based (the +A, ALT stracalorico, on the other hand, selects only adjectives even when attaching to relational/qualifying adjectives), only the latter can "loose" the adjectival suffix maintaining the same meaning (we will see later to what extent this can happen), since the former would turn in a perfectly transparent Pref+N, which is instead often the base upon which the word is formed (vicepresidenziale > vicepresidente). It should be noticed that in the case of suffixed / non-suffixed PRAs' alternation, there is evidence speaking in favour of both directions (i.e. non-suffixed forms acquiring the suffix or, rarer, suffixed forms losing it), see for

22

instance infrasettimana 'midweek', widely attested in web searches (more than 50000 Google hits) and totally unattested in the examined Italian dictionaries and corpora.

This outline is not claiming that each prefix can only fit one kind of construction. Many Italian prefixes do in fact enter more, due to their attachability to more than one lexical category. Let us consider the following examples containing the prefix ultramainly retrieved from La Repubblica corpus (Baroni et al. 2004): i) raffinatezza , con il lino ultradelicato , le tuniche bianche di lana leggerissima Televisore portatile ultrasottile : sar leggero , pieghevole , grande pi o meno le organizzazioni politiche ultraconservatrici , violano il vostro diritto , sancito il grande sogno di popolo ultramarino sparso tra " il Minho e Timor ". di natura e origine ultrasensoriale , possono essere decisivi per la vostra vita . il segreto della sua longevit, l'ultracentenario rispondeva che essa consisteva in

ii)

Some evidence, however, encourages the view that, albeit a common etymological origin, two homophonous ultra- prefixes are now active in Italian. It is apparent how (i) express an alterative meaning (involving the selection of a qualifying adjective, e.g. delicato in ultradelicato 'ultradelicate') whereas (ii) build a relational adjective on a noun base. Where the adjective is ambiguous between a relational and a qualifying meaning, two readings of the outcome emerge: (iii) represent instances of the homophonous adjective ultraeuropeo 'very european/europeist' or 'beyond Europe': iii) io sono ultraeuropeo! ma non sono a favore dell'europa governata da
[www.italiamac.it/forum/showthread.php?p=739740]

giunge a lambire anche aree produttive europee ed ultraeuropee. Esso


[www0.provincia.bologna.it/.../00ECO01R_Imprese_impr_circ_Imola.DOC]

Acknowledging this aspect could also shed light on some inconsistencies found in descriptions of contemporary Italian prefixal system. Iacobini (2004:111)'s analysis

23

of prefixes' base selection could probably benefit from more distinctions with respect to prefixes which would attach to either nouns, qualifying adjectives or relational adjectives, that is to say
aantico-/conextrainfrainteripomultineopluriretrosottosubsuperultra-

Although Iacobini (2004:113) recognises that in some circumstances "the prefix can express different values according to the interpretation of the base adjective (extraprovinciale 'very hick', 'extraprovincial'), it is never hypothesized that, expecially when dealing with distant meanings (in the case of extra-, an alterative vs a temporospatial one), two homophonous prefixes could be assumed. We will not address related diachronic issues: it seems however reasonable to hypothesize that it was indeed the semantic dissimilarity which favoured the cohabitation of the homophonous prefixes' couples. The point to focus seems, in fact, the role of prefixes rather than the "interpretation of the base adjective". If we assume that both noun-based and adjective-based constructions share the same formal properties, it is possible to rely on the disambiguation between homophonous prefixes only, e.g. between alterative extra- (extravergine, extragenuino etc.) and temporospatial extra- (extraparlamentare, extrauterino etc.) which allows the speaker to map each constituent differently, according to the prefixal subcategorization frame. The rise of such template in Italian is most probably the result of the extension of the [Pref[A]A]A schema which is unproblematic for the description of modified qualifying

24

adjectives, and partially motivated by a head operation in the case of headed nounbased PRAs. Such extension to headless bases is not blocked because it is generally unambiguous and perfectly fits into the Italian adjectival system. It must be noticed, however, that since Italian also licenses invariable adjectives (we will address later the issue of language-specificity), suffixed adjectives, as already mentioned, can turn into non-suffixed. Let us consider the following examples: preguerra 'prewar' intramuscolo 'intramuscular' infrasettimana 'midweek' These words are unattested in many dictionaries, but widely attested in the Web (more than 20,000 Google hits each) and corpora: rispolverato la sua tattica preguerra, che quella di tentare di
[La Repubblica Corpus]

Sono utilizzati anche coagulanti intramuscolo o endovena a base


[La Repubblica Corpus]

reduce da una goleada nel turno infrasettimana di luned scorso.


[http://www.labolletta.com/Forums-file-viewtopic-t-6741-view-previous.html]

Let us recall Corbin (1999)'s constraints mentioned earlier: (i) SEMANTIC CONFORMITY CONSTRAINT: The optimal form of an affixed word corresponds to the concatenation of every form contributing to the definition of its meaning; (ii) CATEGORIAL CONFORMITY CONSTRAINT: The optimal form of an adjective formed via affixation has a suffix-like ending. We can draw the conclusion that in such forms the Semantic Conformity constraint is prevailing, since the affixal form lost is not necessary for the definition of its meaning. A complementary perspective can be offered referring to Lieber (2004:161)'s Redundancy Restriction which states as follows:

25

Affixes do not add semantic content that is already available within a base word (simplex or derived). If we adopt Lieber (2004)'s lexical semantic model, an account of many prefixal constructions can shed light on their core semantic structure, which does not rely on the presence of a suffix. The most prototypical "exocentric" PRAs are in fact those formed with a prefix which resembles a preposition (e.g. anti-, post- and sopra- share many features with the prepositions contro, dopo and sopra). As pointed out by Bierwisch (1988) prepositions are "bi-argumental categories", since they involve two arguments whose relation is specified through their semantics. In syntax, prepositional phrases are not thus autonomous. Likewise, in morphology this gives rise to exocentric words: the head position is, in fact, represented by the external argument, absent at the word level (although many Pref+N adjectives have been analysed as embedded constituents of compounds, like riscaldamento prepartita 'prematch warm-up': in this case the external argument is collocated in another node of the same word). Let us analyse a small sample of prepositional-like PRAs according to Lieber's theoretical approach. According to Lieber (2004), the semantic representation of a linguistic unit can be defined as consisting of a semantic/grammatical skeleton and a semantic/pragmatic body. The skeletal representation of the Italian prefix post-, for instance, involves two arguments, one of which cannot be coindexed (the Principle of Coindexation formalizes referential integration) with any internal argument: [+ Temp ([i ] [([j ])] post 'j happens after i'

A PRA like infrasettimanale 'midweek' conveys the meaning related to something temporally situated in the middle of a settimana 'week', indexed as i below. The adjective itself does not provide semantic clues about this entity, but a coindexation takes place between the adjectival suffix and the external argument (j). [dynamic([j ])] [+ Loc ([i ] [([j ])] ([+material ([i ])])] 'sit.ed in the middle of the week' ale infra settimana The same holds, for instance, for the following PRAs: subepidermico

26

[dynamic([j ])] [+ Loc ([i ] [([j ])] ([+material ([i ])])] ico sub epidermide extracomunitario [dynamic([j ])] [+ Loc ([i ] [([j ])] ([-material ([i ])])] ario extra comunit intracellulare [dynamic([j ])] [+ Loc ([i ] [([j ])] ([+material ([i ])])] are intra cellula

'situated under the epidermis'

'sit.ed outside of a community'

'situated inside a cell'

This representation makes explicit how the core semantics is determined by the argumental structure of the prefix. The adjectival suffix has a coindexation function which arguably depends on contextual factors, such as the discourse availability of the external argument: this explains why the non-suffixed form is probably perceived as more natural where the argument is easily retrievable and indexed, e.g. in a postnominal position, i.e. where the adjectival suffix is not necessary from a morphosyntactic viewpoint, that is to say for marking the grammatical relationship between the adjective and the external argument. In the analysis of the alternation of infrasettimanale and infrasettimana, a strong preference for the suffixed form in predicate function was found. Specific queries aiming at assessing the most attested form after copulas yielded the following results: "/sono infrasettimanale/i" 'is/are midweek(+suffix)' "/sono infrasettimana" 'is/are midweek' 1008 hits

86 hits

Another factor at play is probably the adjectival status itself, less prototypical in postnominal position (where its role can be a compound constituent). It must be noticed that the non-suffixed form can also be employed with other part of speech functions, being categorially underspecified. Infrasettimana, e.g., was adverbialized in a considerable number of hits (34 out of the first 100 Google hits). See for instance anche la festa della polizia la si faceva infrasettimana.
[http://www.fiammeblu.org/forum/pop_printer_friendly.asp?TOPIC_ID=1193]

27

Mi hanno invitato sull'isola infrasettimana... dal luned al venerd.


[http://community.fondali.it/forum/viewtopic.php?p=144562]

Ti consiglio di ritornarci, magari infrasettimana, perch nei week-end


[http://www.baltazar.it/ristoranti-santarcangelo-di-romagna/5876]

per fortuna noi possiamo andare infrasettimana e con la pioggia!


[http://briggis-recept-och-ideer.blogspot.com/2009/07/insalata-di-5-cereali.html]

per questo quando posso vado infrasettimana


[http://www.hwupgrade.it/forum/archive/index.php/t-1296693.html]

This aspect further provides evidence for the purely categorial function of the suffix. It is reasonable to assume that the form infrasettimana, originated from the suffixed form, has then spread towards usages other than adjectival. Another problem is posed for theories conceiving prefixes as heads, which should face the fact that the same prefix (with the same meaning) can give rise to more than one change in word-class (or that they do not cause word-class changing at all). Summarising, although the [Pref[A]A]A schema expanded itself, semantic constraints are however at play and favour the rise of alternative concurrent constructions, arguably preferred by speakers because of their simplicity (less formal units required) and their semantic transparency. Moreover, the non-suffixed forms can be employed as adverbs, for instance, with no risk of ambiguity. According to the mentioned approaches (Corbin 1999 and Lieber 2004) we can combine the claims that the non-suffixed form is the optimal output of the word formation process, and that the adjectival suffix does not add content already available within the base. The latter could be an obvious claim for many suffixes forming relational adjectives, since they only have a formal function. We will see later, however, that although the Italian adjectival suffixes involved often have a purely transpositional function (from noun to adjective), PRAs also draw on nonrelational suffixes/adjectives.

1.8 The role of adjectives


An aspect which deserves a closer look is the role of independent relational adjectives in the formation of paradoxical PRAs. It is acknowledged that relational adjectives result from the categorial transposition of nouns. As Bally (1944:97) observed, a relational adjective "substitutes nouns without changing any aspect of their value". Relational adjectives display a syntactic behaviour which set them apart from other types of adjectives (qualifying, descriptive etc.); the tests which usually apply are:

28

i) Prenominal position (for Romance languages) a. b. Mario un bravo studente 'Mario is a clever student' *Quella una marina onda 'That is a sea.Adj wave'

ii) Predicate in copular sentences a. L'energia che usiamo rinnovabile b. ?L'energia che usiamo solare4 iii) Gradability a. Mario ha dei problemi molto seri 'Mario has some very serious problems' b. *Mario ha dei problemi molto coniugali 'M. has some very marital problems' Unlike other adjectives, relational ones cannot modify a noun in a prenominal position (i.b), be normally used in copular sentences (ii.b); quantifiers are not accepted as RAs are not scalar (iii.b). As already shown above, Italian displays a quite clear-cut range of denominal adjectival suffixes. Some properties of these adjectives should however be taken into account: first, most suffixes do not only form relational adjectives; second, purely relational adjectives can develop idiosyncratic meanings partially shifting from the "transposed noun" reading. Italian suffix -oso can have both a qualifying and a relational value. As a matter of fact, some adjectives are only qualifying, some are only relational and some are ambiguous between either interpretation (see examples below from ItWaC). Qualifying-only #252260 gli studi medi presso una prestigiosa scuola privata e conobbe il Relational-only #63908630 replicazione delle cellule cancerose, da rallentare la progressione Ambiguous 4. Grammaticality judgments vary among speakers, since copular sentences are probably the context where relational adjectives are most acceptable among the ones employed in the tests. 'The energy we use is renewable' 'The energy we use is solar'

29

#1410230 solo le reazioni a livello di sistema nervoso autonomo o di mimica. #1920740 esprimermi al meglio. S, ero nervoso, e mi sono lasciato andare Relational adjectives can thus be homophones to qualifying adjectives, or qualifying meanings can derive from the former. Apart from ambiguous suffixes, in fact, it must be noticed that relational-only suffixes also produce idiosyncratic meanings which cannot be reduced to the grammatical transposition of the base noun. Let us consider a small sample of relational adjectives containing different suffixes: funzionale 'functional' solare 'solar' letterario 'literary' abortivo 'abortive' africano 'african' marino 'marine' etnico 'ethnic' planetario 'planetary' A random list of 600 attestations (75x8) of the lemmas (in their occurrence as adjectives) was extracted from ItWaC in order to have a general idea about their meaning in context. The data confirm the assumption that qualifying and relational meaning often cohabit.
Sample concordances #39587 processo di integrazione funzionale tra le rispettive e biblioteche. #232919 alfabetizzazione che sia funzionale al successivo orientamento #208458 che in Italia) si utilizzano i pannelli solari dieci volte pi di noi, #3960163 piaciuto molto quel clip, mi era sembrato molto solare , e mi #226305 di un patrimonio artistico e letterario , scientifico e tecnico, #521837682 Questo romanzo molto letterario ma anche un romanzo #75434299 a seconda dei soggetti della vicenda abortiva: la posizione #84771296 effetto della pillola del giorno dopo abortivo o meno? #14099 10% imprenditori di origine africana ed extracomunitaria che di #1152550 Fu un diplomatico africano a consegnare ai servizi italiani #514507 importanza dell' ambiente marino e le interazioni che si sviluppano Relational Qualifying 28 47

55

20

68

30

45

39 73

36 2

30

#1621459203 Il fango dell' Elba, infatti, marino ed simile, ricco com' di #618270 sulla razza o l' origine etnica , le religioni e le convinzioni personali #2334270 attributo di etnico: musica etnica, cibo etnico , artigianato etnico #843076 tutti i problemi su scala planetaria - si tratti della guerra, di #4329286 far avere una mentalit pi ampia, pi planetaria . - " Total 50 25

66 409 (68.16%)

9 191 (31.83%)

A more in-depth consideration of this feature raises some questions about the status of adjectival affixes in the forms we are now considering. How do ambiguous adjectives interact with prefixal constructions? What role do non-relational adjectives play in their morphosyntactic and semantic outcome? As the data above suggest, it is not uncommon to find instances of denominal adjectives in which the qualifying use prevails on the relational one. In the small sample considered, let us consider the adjective abortivo 'abortive': it is derived from the noun aborto 'abortion' and results in a relational meaning 'related to abortion' and a qualifying one 'causing abortion'. A Google query of the most productive prefixes attached to the lemma abortivo yielded the following hits. If we look at the word inside PRAs, however, it is quite reasonable to assume that no trace of the non-relational use remains, since the argument structure of the prefixal construction only involves the selection of the noun; the adjectival ending only has a categorial function. scatena tutto questo fanatismo antiabortivo che credo sia portato il vissuto emotivo preabortivo e l'atteggiamento delle utenti del sintomi da stress postabortivo: le americane nel 39% dei casi Deltacortene perch sono un soggetto poliabortivo. Insieme al Denominal adjectives which are both relational and qualifying find their noun base selected by the semantics of the prefix giving rise to paradoxical PRAs. From this observation, we could assume that in the formation of such words relational adjectives are selected regardless of their concurrent qualifying use. A closer look at adjectival forms in PRAs allowed to go beyond and see that the word formation process does not only rely on the retrieval of relational adjectives: other kinds of adjectives, in fact, can

31

become constituents of the word in presence of a (looser) relationship with a noun form. The following forms were encountered: #645716926 anno di corso, se non dallo stesso biennio postobbligatorio. #1476988475 dominati dal pathos di una sopravvivenza oltrevitale, ponendo #1791724564 di stanchezza; un effetto antiansioso, ossia riducono l'ansia #541399038 amministrazione. Il comportamento antidoveroso dell' agente The examples above are cases in which the adjectival forms are not relational. They are nonetheless exploited probably because of their recognisable noun base. An analysis of the first 100 ItWaC hits of each simplex adjective (obbligatorio, vitale etc.) was performed in order to ascertain their use, which resulted in a very poor index of relational use (less than 5%). These data shed light on Corbin (1980)'s claim that suffixes have only a categorymarking function: it seems confirmed by the fact that the morphological construction draws on adjectival forms whose idiosyncratic meaning is not relevant at all in the complex prefixal forms, and which enter the construction only because of their morphological relationship with a base noun.

1.9 Novel PRAs


Do schemas capture a productive shortcut word formation pattern? One of the main theoretical aims of the constructionist approach is a description aiming at both simplicity and psychological reality. Booij (2005:38) refers to the domain of cognitive psychology, where evidence was found (Anderson et al. 2004) of the tendency towards the combination of different cognitive tasks into one "that has the effect of both, and links this to the explanatory and predictive power of unificated templates, which allow the speaker to create complex words without unnecessary intermediate steps. A closer look at the strategies employed by language users in comprehension can further shed light on the nature of the constructions we have dealt with in the previous sections. In order to better understand how Italian native speakers comprehend and coin PRAs and thus compare the descriptive level with the actual usage of the templates introduced, a series of tests was performed aiming at evaluating whether and to what extent Italian language users can have access to the an active schema defining the structure of PRAs for comprehending novel instances of the construction and coining new ones on the base of a semantic instruction.

32

A set of 10 novel words was created to be tested for comprehension among 20 native speakers (mean age: 27, ranging from secondary school to postgraduate education). The words created had to meet the following requirements: (i) be unattested in the web; (ii) conform to the [Pref[A]A]A template; (iii) be plausible, both from a pragmatic and a semantic viewpoint. Both transparently suffixed and allomorphic adjectives were employed as constituents. These morphologically legal novel words were then put into sentences, and participants were asked to paraphrase the meaning of the PRA in context. One would predict that, if the template can be easily retrieved by speakers, they will be able to infer the paradigmatic relationships necessary for establishing the argumental connections between the nominal base of the adjective and the prefix, e.g. the alleged noun base of antiadottivo (adozione) and its relation with the prefix anti-. This does not mean that the parsing of the adjective should pose more difficulties than other forms, since the main advantage of the unified template should be that of being simpler up to the extent to which comprehension is not more demanding (unlike a theoretical representation which does not correspond to the user's strategies, albeit economical or simple). The novel words employed were the following: proinfantile, antiadottivo, intraarboreo, circumcanadese, multiabortivo, proscolastico, ipoterreno, oltreelettorale, infratraumatico, prodecisionale; They were built through a prefix attached to a relational adjective, e.g. pro+ infantile 'in favour of children, childhood'. Looking at the paraphrases provided by participants, it was clear that in most cases they were able to understand the intended meaning. They were able to individuate the noun base and derive the adjectival meaning on the basis of the prefixal semantics.
Sentence with novel PRA Paraphrase

Questa legge proinfantile protegge i diritti umani dei minori. "a/in favore/aiuto dell'infanzia" (95%) Le malattie intraarboree colpiscono spesso le querce. Questa legislazione antiadottiva ha fatto [...] Il viaggio circumcanadese di Gianni finito ieri. La paziente multiabortiva ha partortito un figlio. "(sviluppate) dentro l'albero/gli alberi"(80%) "contro/contraria all'adozione" (70%) "attorno/intorno al Canada" (85%) "che ha subito/avuto/praticato molti aborti/pi di un aborto" (90%)

33

I provvedimenti proscolastici del governo hanno [...]. L'inquinamento ipoterrestre molto pericoloso. L'entusiasmo oltreelettorale del candidato ha sopreso tutti.

"a/in favore/sostegno della scuola" (90%) "sottoterra/sotto la terra" (75%) "(che ) andato/durato oltre/aldil del/ dopo le elezioni/ l'elezione" (75%) "che favorisce/ a/in favore di (la/una) decisione/le decisioni " (75%) "caratterizzata da/che (le) ha provocato/ causato/procurato" molti traumi (80%)

Il suo atteggiamento prodecisionale ha aiutato l'azienda.

Maria ha avuto un'esperienza pluritraumatica.

This confirms the claim that the template is productive and can be retrieved by speakers in order to understand novel instances. Without entering the details of the diachronic processes related to the issue, it must be highlighted that many representatives of the pattern were inherited from Latin (Italian cisalpino derives from Latin cisalpinus, cf. Tollemache 1945). Words not derived from Latin, combined with the coinage of new (comprehensible) instantiations of the template, demonstrate that the pattern is productive in Italian.

1.10 Word Internal Anaphora


Another relevant argument is related to morphological constituents' reference. Since Postal (1969) the notion of anaphoric islandhood has applied to complex words: their internal constituents cannot be coreferential with other words. Lexical Integrity Hypothesis, in turn, states that syntax has access to the word as a whole, not to its composing parts. This set of principles would categorically rule out word internal anaphoras (WIA) like the following (the discussion will be limited to 'outbound' anaphora): *Mary hates guitariists because she doesn't like iti. As pointed out by Montermini (2006) in his detailed review, cross-linguistic evidence showed how the criteria invoked to categorially judge as impossible these constructions were actually too strong and that WIAs could be acceptable up to a certain extent. A wide range of facilitating conditions have been thus proposed that could account for the variation in the acceptability of WIAs, including both semantic-pragmatic constraints

34

and morphosyntactic ones. Although some scholars treat anaphors as a non-grammatical issue (Montermini cites Ward et al. 1991:450's claim that there is no principle of grammar that explicitly prevents word-internal antecedents for pronominal anaphors), there is evidence that speaks in favour of a role played by morphology. Montermini (2006:133) shows that full NPs are much more acceptable as anaphors than pronouns / zero forms: (i) Seppur non sia milaniese, per diversi motivi sono vicino a questa citti Though I am not a Milanese for many reasons I am close to this city (ii) *?Seppur non sia milaniese, per diversi motivi lei sono vicino Though I am not a Milanese for many reasons I am close to it How can WIA shed light on the formal properties of PRAs? In order to fully assess the role of affixation in such constructions, a fruitful approach may be represented by a comparison between affixed and non-affixed forms, evaluating their behaviour with respect to WIA. What is at issue here is whether there are differences in the degree of acceptability of WIA due to the morphosyntactic configuration of the PRA involved. As recalled by Montermini, an often underrated aspect of lexical semantics is the absence of any intrinsic referential power. Sproat (1988:294)'s claim that a noun such as dog [...] does not pick out any particular dog or any particular set of dogs. Only NPs have this property may thus be the reason why WIA is not licensed. In the words of Montermini (2006:129), The very reason why anaphoric expressions cannot be coreferential with a word included within a more complex lexical item is that, indeed, words are not per se referential entities. [...] Referential capacity is a matter of phrases (and NPs are the prototype of referential units), not of words, or, as we would say today, of lexemes. Of course a lexeme has a semantic representation, but it remains a virtual one, and can only be instantiated when a word is used in discourse via a syntactic construction. Since it is commonly assumed [...] that morphological rules operate on lexemes, i.e. on pre-syntactic units, there is no reason to think that the base of a complex word should maintain an autonomous referential power.

35

Montermini concludes that, since WIAs are actually attested, two theoretical paths can be followed. Either we accept that units larger than words (NPs) can enter word formation, or we abandon the claim that lexemes have no referential capacity within complex words. It is observed that WIAs prefer bases with a strong referential power, expecially proper names. Does this imply that such names are not lexemes? Montermini (2006)'s main insight is probably the observation that the same range of prefixes favouring WIAs (with proper names as bases) was also the same favouring the attachment of phrasal units and inflected nouns (among the examples, manifestazione anti-Bush anti-Bush demonstration, marcia anti moschea di Lodi anti Lodi mosque demonstration, prodotto antirughe anti wrinkles product). Although proper names can enter a wide range of prefixed constructions, in fact, only some prefixes (namely post- and anti-) exploit the referential power of the individuals involved. Montermini (2006:139) notices that the following examples differ with respect to the semantic contribution of the proper name to the complex word. (i) manifestazione anti-Bush anti-Bush demonstration (ii) imberlusconirsi to become like Berlusconi (iii) mussoliniano mussolinian Whereas (i) clearly denotes a demonstration against the individual called Bush, in (iiiii) "the meaning of the derived word is not constructed on the individual himself, but rather on some stereotyped characteristics of his person, ideology, activity, etc. Someone who has imberlusconito has not become Berlusconi himself, but has become similar to him, by adopting some peculiar features of his personality, by subscribing his policy, and so on". Montermini (2006:139) remarks that this asymmetry mirrors the different properties of "two sorts of prefixes: those whose behavior recalls that of suffixes and those whose behavior recalls that of compound words". Let us consider Italian PRAs: formally, they are not compound-like, but they rather involve affixed constituents like (iii). Generally speaking, a PRA like postberlusconiano shares formal and semantic characteristics with both Pref+N and N+Suf adjectives. Their WIA behaviour can arguably be considered very relevant in contributing to their description. Corpus data (specific queries were designed in order to find instances) and Google searches yielded the following results: non sono io ad essere antiberlusconianoi, luii che un corruttore.

36

[http://www.reportonline.it/2010040241951/politica/lequivoco-dellantiberlusconismo.html]

Gli antiberlusconianii vogliono le stesse cose di luii senza lui;


[http://indipendenza.lightbb.com/politica-italiana-f2/antiberlusconismo-t458.htm]

#1833996035 Adesso solo antiberlusconiiana ... pur essendo liberista come luii #993625452 basta essere antiberlusconiiani , vogliamo essere alternativi a luii #592997138 polemica anticristiania : una grande religionei e una grande #156560777 campagna antirutelliiana [...] in occas. della suai seconda rielezione #231715100 Non sono mai stata antiamericiana anzi, guardavo [...] questa nazionei #625660845 qui infatti il titolo sembra antifelliniiano per eccellenza: luii cos legato In order to gain further insights on acceptability, an informal survey was submitted to 40 native speakers of Italian (mean age: 32, ranging from secondary school to postgraduate education), asking them to give judgments about sentences containing WIAs with antecedents embedded in both prefixed and non-prefixed adjectives (Yes = acceptable sentence, No = unacceptable sentence). Two set of sentences were built, each one containing either the prefixed or the non-prefixed version of each adjective, so that none of the participants could judge two versions of the same sentence. The table below summarizes the results.
Sentence 1a. Maria berlusconiana perch un politico capace. 1b. Maria antiberlusconiana perch un politico incapace. 2a. L'America obamiana apprezza il suo carisma. 2b. L'America postobamiana apprezzer il suo operato. 3a. Maria era veltroniana quando era il sindaco di Roma. 3b. Maria era antiveltroniana quando era il sindaco di Roma. 4a. I tifosi juventini credono che la sua dirigenza sia la migliore. 4b. I tifosi antijuventini credono che la sua dirig. sia la peggiore. 5a. Dario felliniano: ama tutti i suoi film. 5b. Dario antifelliniano: odia tutti i suoi film. 6a. L'opera manzoniana stata influenzata dalla sua vita. Yes 4 11 6 14 1 9 1 8 6 9 3 16 8 13 2 19 9 19 11 12 6 17 No 0 1 1 4 0 2 0 1 2 5 0 ?

37

6b. La critica postmanzoniana stata influenzata dalla sua vita. 7a. I sonetti shakespeariani contribuirono alla sua fama. 7b. I critici proshakespeariani contribuirono alla sua fama. 8a. Le trasmissioni defilippiane sono seguite dai suoi fan. 8b. Le dichiarazioni antidefilippiane sono odiate dai suoi fan. 9a. La propaganda mussoliniana si fondava sui suoi discorsi. 9b. La propaganda promussoliniana si fondava sui suoi discorsi. 10a. La politica italiana essenziale per il suo sviluppo. 10b. La politica antiitaliana nociva per il suo sviluppo. Total (non-prefixed)

12 7 13 7 15 9 12 6 9 50

7 13 3 10 3 11 5 12 7 142

1 0 4 3 2 0 3 2 4 8 (4%) 27

(25%) (71%) 112 61

Total (prefixed)

(56%) (30.5%) (13.5%)

The theoretical relevance of the attestation and acceptability of WIAs in PRAs built on proper names is twofold. First, another argument in favour of a purely formal suffixation of PRAs is added to the debate. If we combine, in fact, the evidence presented by Montermini (2006) and coming from the above web and corpus data, it seems reasonable to assume that a clear trend is manifested: proper names embedded in PRAs tend to maintain their referential power although they are formally identical to a construction where the semantics is not "constructed on the individual himself" (Montermini 2006:139). Whereas the adjective berlusconiano arguably does not contain any active referential unit, antiberlusconiano does, licensing WIA: antiberlusconii antiberlusconiiano berlusconiiano Xi Xi X*i

On the other hand, it would be difficult to assume, as suggested by Montermini for the case of Pref+N constructions, that the second formal constituent of the word could be a phrase and not a lexeme.

38

The second, consequent conclusion which can be drawn is a general confirmation of the similarity between Pref+N and Pref+N+Suf constructions. These data could represent a first step in assessing the differences in their behaviour and/or the reasons of their coinage. The issue will be explored later.

1.11 Transcategorization
PRAs can be the source of transcategorized nouns. This means that some nouns can be the result of a categorial shift from adjective to noun, see for instance antifebbrile, both adjective meaning "against fever" and noun denoting a drug against fever. With respect to this process, differences emerge between suffixed and non-suffixed forms, since the transcategorization shift occurs in different manners according to the thematic relations underlying the source word and other functional properties. An outline will be offered of the main paths followed by Italian constructions. Let us consider the semantic representation of antifebbrile: [dynamic([j ])] [+ Loc ([i ] [([j ])] ([+material ([i ])])] ile anti febbre 'against fever'

As we have argued above, a coindexation occurs with an external argument in the utterance (this is reflected in number and gender agreement of the suffix). When a conventional association arises between a (set of) referent(s) and a property expressed by a PRA, however (the same happens for simplex adjectives e.g. la nazionale, lit. the national 'the national team') the adjective itself is employed as a noun belonging to that class of referents and connotated by the properties expressed in the PRAs. It is implied that such noun fills the external argument of the PRA: [[dynamic([j ])] [+ Loc ([i ] [([j ])] ([+material ([i ])])]]j ile anti febbre 'drug against fever'

If we limit ourselves to antifebbrile, a fairly productive class of transcategorized nouns can be individuated, i.e. that of drugs with the purpose of acting "against" a certain illness or undesired psychophysical state. A survey carried out on La Repubblica corpus and the web yielded several occurrences of this pattern: e grana medico-legale con un anticoncezionale intrauterino. Drug Administration come un antidepressivo e poi bloccata

39

xorubicina -- scrive -- un antitumorale sul quale per diale di coppie che usano un antifecondativo sicuro a letto e di inghiottire un antidolorifico . E fin quando necessari 6/10 milioni per l'antitetanico monodose ; La Spirulina un antivirale ma anche un tonico molto Lo Zirtec un antistaminico di nuova concezione Zonisamide, un antiepilettico. Le crisi epilettiche assumere un antiadiposo che agisce sulla tiroide L'uso degli antinfiammatori stato limitato sin dall'origine Note that transcategorization is generally conceived as a derivational process, a device "alternative to morphology-based word formation rules such as derivation and compounding, to create new words out of existing ones" (Jezek & Ramat 2009:398). It is reasonable to assume, however, that from a constructionist viewpoint the systematic occurrence of a template with certain formal, semantic and categorial properties can feed an independent constructional idiom, so that we can formalise one describing the properties of anti- drugs in Italian: [anti[[UNDESIRED PSYCHOPHYISICAL ENTITY/STATE/PROCESS]N RAsuff]N 'drug against N'

From a functional viewpoint, it should be noticed that the suffixed form has the advantage of having a morphological exponent on which inflection can be realized. Unlike with adjectives, in fact, Italian does not display productive patterns for generating invariable nouns. Second, it does not lead to ambiguities with headed prefixal constructions in anti- (e.g. antiromanzo 'antinovel', antieroe 'antihero' etc.). With respect to the last point, however, non-suffixed forms are limitedly favoured in constructions based on exocentric prefixes. Antifebbre is attested as a noun (as well as prefixed words with no available derived adjective, e.g. pregara), because although it has little inflectional power, the thematic relations introduced by anti- can be instantiated in a noun. On the contrary, alterative prefixes generate ambiguity with a headed structure modified by a prefix. Let us consider the transcategorization of multistratoAdj 'multilayered' > multistratoN 'multilayered wooden panel'. It is likely that since the prefix multi- can directly modify the internal constituent strato, the noun multistrato can be interpreted as being headed by strato. We know, instead, that a multistrato is not a strato, but a specific object with many layers. This is testified by the frequent occurrence (300+ Google hits) of a pluralized form which indeed denote the interpretation of strato as a head, i.e. multistrati:

40

Anche per i multistrati sono numerosi gli spessori ed i formati


[http://www.bonomipattini.com/]

This is prevented (with exocentric prefixes too) in transcategorized prefixal adjectives where the output word's gender and/or number are different from those of the internal noun constituent, e.g. multisala 'multiplex cinema', prevendita 'pre-sale', postpartita 'post-match TV programme' (all male nouns with female bases). A last remark on headed nouns based on exocentric prefixes, e.g. preesame 'pre-exam', prepartita 'prematch', antiromanzo 'antinovel', antieroe 'antihero' etc. A semantic analysis has to my knowledge never been attempted. My idea is that such constructions could be conceived as follows: the same constituent occupies both the argumental roles introduced by the prefixal semantics, i.e. we have a referent which is specified by a relation with the same class of referents it belongs to: an antiromanzo 'antinovel' is a novel which displays features going against the conventions related to them, like the unfolding of a sequential plot, a "novel against the novel": [+ Loc ([i ] [([i ])] ([-material ([i ])]i anti romanzo 'antinovel'

Likewise, a pre-esame 'pre-exam' is an exam to be taken before another exam, and an antieroe 'antihero' is a hero whose character is contrary to that of the archetypal hero. [+ Temp ([i ] [([i ])] ([-material ([i ])]i pre esame [+ Loc ([i ] [([i ])] ([-material ([i ])]i anti eroe 'pre-exam'

'antihero'

1.12 False heads


As already pointed out earlier, the rise of bracketing paradoxes in PRAs can partially be ascribed to head-based adjectival derivations (e.g. vicepresidente 'vicepresident' > vicepresidenziale 'vicepresidential'). This analysis cannot be applied to exocentric constructions, since they are headless by definition. There is, however, a morphological domain where operations occur on non-heads: inflection. Italian display a wide range of

41

cases where non-heads are inflected. For the present purposes, a brief account will be provided of headless prefixal forms. Montermini (2008:208) mentions the case of Italian prefixed nouns which are "occasionally pluralized" in the following way: sing. sottobicchiere under.glass sottobicchieri under.glasses 'coaster'

plur.

'coasters'

This exocentric construction, defining "[something] under a glass" is pluralized as if one of its constituents, the noun bicchiere, was the head. It is not surprising to see how this is related to grammatical factors. Montermini notices that this kind of pluralization occurs only when the grammatical gender of the complex noun is the same of the constituent (in this case, sottobicchiere is masculine as bicchiere). A masculine noun like antiforfora 'anti-dandruff' with the embedded feminine constituent forfora 'dandruff' cannot pluralize in *antiforfore. Sottoveste 'petticoat', instead, a feminine noun with the embedded feminine constituent veste 'dress', can pluralize in sottovesti. We will ignore words with embedded pluralized constituents (see for instance antirughe 'anti-wrinkles'). Traditional morphological theory ascribes to the head of a word the property of percolating its features (among which the lexical category) to the whole complex word. We can imagine that speakers assume such grammatical match (probably reinforced by the determiner employed) as a sign of the headedness of the word, and consequently pluralize it. Although in the majority of cases the semantics of the word is transparently based on the constituent as a non-head, this is nonetheless pluralized most of the times. A Google search of the pluralized forms of sottobicchiere 'coaster', yielded 62 hits of "i sottobicchiere" (clue that the semantic structure is however still active), and 2,820 hits for "i sottobicchieri". Italian also displays pluralization on "false heads" in the case of exocentric compounds. See for instance the pluralization of grattacielo 'skyscraper', asciugamano 'hand towel': grattacieli, asciugamani. It seems reasonable to assume that a similar process could occur in the derivation domain, in particular with prefixal constructions. A large body of literature (at least from Williams 1981) asserts that prefixes are not susceptible of having head properties. The most recalled criterion is the transcategorization property: unlike suffixes, which determine the lexical category of the word they are attached to, prefixes rarely do,

42

expecially in Romance and Germanic languages. Most orld languages display a tendence to mark morphological categories and values through suffixes rather than prefixes, and this is reflected on the possibility of heterocategorial combinations. As highlighted by Montermini (2008:210), in Italian and French suffixes can realize all six conceivable combinations between lexical categories, whereas prefixes can enter only three of them. According to Corbin (1999:80) a wide range of factors has led scholars to build a "dominating conception" according to which prefixes cannot be heads. Although non-Indoeuropean languages display class-changing prefixation, the prototypical role played by prefixes in Italian probably leads speakers to assume that they cannot be the morphological exponent of the word. Unlike with suffixed forms, the categorial assignment is thus unclear and other morphological means are required in order to make the word conform to its expected grammatical properties. In the same way as vicepresidente 'presidential' derives the adjective vicepresidenziale 'vicepresidential', a Pref+N exocentric prefixed adjective (be it attested or not) can be conceived as the base of a formally Pref+Adj adjective in which the suffixation scopes over the whole word: antifebbre > postguerra > internazione > antifebbrile 'antifever' postbellico 'postwar' internazionale 'international'

Let us now consider other kinds of evidence coming from morphological reanalysis. Corbin (1999)'s Categorial Conformity Constraint predicts that the optimal output in word formation exhibits a categorially consistent ending, i.e. for instance an adjective with an adjectival ending. This does not imply that such ending is ascribable to a morphological operation based on the compositional semantics of the word. The word antifascistaAdj, 'antifascist' for example, displays an adjectival ending which is reanalysed as the derivational and inflectional marker of the word, as attested by the forms antifasciste, 'antifascist.F.pl', antifascisti 'antifascist.M.pl'. Here forms found in La Repubblica corpus: fu uno dei pochi professionisti antifascisti che rifiutarono sempre delle sue idee antifasciste , anzi le manifesta in modo spericolato The suffix -ista is not transparently added to the noun base to which it is semantically related, i.e. antifascismo 'antifascism' (-ista and -ismo are mutually exclusive: we cannot determine which is the base of the other) which in turn is a case of categorial conformity:

43

the noun ending -ismo is consistent with the category of the word, although it originally belongs to an internal constituent. The semantic representation of the word displays a double role of the suffix, coindexed with both the internal and the external argument of the prefix: [dynamic([i ])] [+ Loc ([i ] [([j ])] ([-material ([i ])])] ismo anti fascismo Corpus evidence shows that this is reflected in pluralization: cio non tutti gli antifascismi sono di per s democratici e ha citato variegata galassia degli anticomunismi cattolici: Dossetti, La Pira We can generalise that the grammar tendentially requires a morphological exponent on which morphosyntactic properties can be realised. As long as this does not lead to parsability and/or comprehension problems (prefixal argumental structures play a major role in this), it is not necessary that the exponent is transparently external to the semantic item to which it applies, creating thus two analysis levels, the formal and the semantic one. The benefits for the language system are probably lying in an "economy" of suffixation, both for non class-changing and class-changing derivation. In the first case (fascismoN > antifascismoN or fascistaAdj > antifascistaAdj, although the latter should be in my opinion taken as a derivate of the former) avoiding a sequence of two identical morphemes (or with very similar functions, see Lieber 2004). In the second case, employing an already established relational adjective instead of creating a more complex formal structure requiring an external suffix. 'sth. opposite to fascism'

1.13 Alternation
Is the alternation between suffixed and non suffixed PRAs due to specific grammatical and/or contextual factors? The aim of this section will be outlining some of the factors at play in the alternation between antifebbre-like and antifebbrile-like constructions. If it is maintained that such forms are synonymous, some conditions might exist which favour one of the two forms in different contexts. Corpus and web data was extracted containing alternating forms. We will now review the main remarks drawn upon the observation of how the usage of suffixed and non suffixed PRAs can be ascribed to semantic, syntactic or morphological factors.

44

1.13.1 Phrasal constituents One of the conditions which leads to the usage of a non-suffixed form is arguably the exigence of modifying the internal noun base, i.e. of obtaining an adjective with an embedded phrase. As we can see in the following examples extracted from the web, the same function is performed either by a noun (on which the relational adjective is based) or by a phrase: incentrato sulle manifestazioni antibelliche tenute nel 1968
[http://www.movieplayer.it/personaggi/3378/liev-schreiber/filmografia/]

E poco si sa del movimento antiguerra in Iraq che sta prendendo piede


[http://www.teatrostabiletorino.it/pressdata/031250.pdf]

adesione a quel banale razionalismo postconciliare che ritiene degno


[http://www.ratzinger.us/modules.php?name=News&file=article&sid=223]

Nella elaborazione post-concilio di Nicea, il Credo diventa


[http://mysterium.blogosfere.it/2006/03/il-concilio-di.html]

Inutile promettere posti di lavoro preelettorali - ha rimarcato Stanzione


[http://www.dentrosalerno.it/web/2009/05/29]

posti di lavoro fisso preelezioni regionali ...rischiamo di parlarci


[http://retenazionaleeducatori.forumup.it/post-7932-retenazionaleeducatori.html]

In the latter forms we have an adjective (e.g. regionali) or a prepositional phrase (e.g. in Iraq) attached to the noun base of the alternating suffixed form, which further specify the meaning and the reference of the noun. They cannot arguably perform this function if the noun is embedded into an adjective, see for instance le manifestazioni antibelliche in Iraq lit. 'demonstrations antiwar.suff in Iraq' or sondaggi preelettorali regionali, lit. 'polls preelection.suff regional' in which the modifier is interpreted as attaching to the noun modified by the PRA: le manifestazioni antibelliche in Iraq are not demonstrations against the war carried out in Iraq, but rather anti-war demonstrations carried out in Iraq, etc. An attempt to find instances of this hypothetical pattern (based on post-, pre-, anti- and pro- prefixes) was in fact unsuccessful, apart from one exception: trovare sfogo nell'ambito prescolastico dell'obbligo.
[http://it.groups.yahoo.com/group/retescuole/message/3]

45

In the context encountered, the ambito prescolastico dell'obbligo, lit 'sphere preschool.suff of the obligation' is only conceivable as the pre-compulsory education sphere, thus representing a violation of the above generalization (and maybe of the Lexical Integrity Hypothesis to a certain extent) because the PP dell'obbligo is bound to one of the internal constituents of prescolastico, namely scuola. The data examined show, however, that the generalisation is solid. Finally, let us consider the following form: collaborazione stretta con la Brigata anticriminale francese.
[La Repubblica corpus]

We can add that among the examined instances of PRAs, some seemed to display the suffixed form for the complementary reason, i.e. avoiding in a Noun - PRA - Adjective sequence triggering the interpretation attaching the final adjective to the noun base of the PRA: as we have seen above, it is generally not licensed where the PRA is suffixed (the last example addresses a french anticrime brigade, not a brigade against french crime). 1.13.2 Ambiguity Iacobini (1992) states that a condition on the formation of Pref+Adj PRAs (and not Pref+Noun) is the independent attestation of the relational adjective. This could be a sufficient condition but not a necessary one: since many attested Pref+N adjectives are based on nouns with independently attested relational adjectives, it is not straightforward that, where the relational adjective exists, it must be included in the PRA based on the noun. Montermini (2008:205) mentions anticrimine 'anticrime', antigravit 'antigravity' (the respective relational adjectives criminale 'criminal' and gravitazionale 'gravitational' are available in the lexicon). It can be hypothesized that one of the factors involved is the potential ambiguity triggered by constituents, i.e. a potential constituent is avoided and another is preferred because the former could generate ambiguity in the interpretation of the PRA. Among the examined instances of scarcely (or un-) attested suffixed PRAs, the candidate forms in -ale raised the issue whether the nouns undergoing trascategorization ("the categorial shift of a lexical item with no superficial marking, resulting from its employment in a new (morpho)syntactic environment" as defined in Jezek & Ramat 2009:391) can yield such effect. Criminale 'criminal' is both the relational adjective derived from crimine 'crime' and a trancategorized noun meaning 'crook'. Due to prefix subcategorization, the form

46

could be interpreted as a noun and thus mislead the interpretation, e.g. in the case of anticriminale "against the crime" vs "against (a) crook(s)". On the other hand, in some cases the non-suffixed form can be ambiguous between two parts of speech, e.g. in adverb - adjective couples like oltreoceano - oltreoceanico 'overseas', where oltreoceano can be either an adverb or an adjective (this case is not due to different parsing options, but rather to an undefined categorial status). che rimbalzavano oltreoceano trasmettendo l' immagine perch non inventarne uno oltreoceanico tra la Toscana Summarising, due to the part-of-speech ambiguity that in some circumstances can arise, it is reasonable to assume that the unambiguous form is usually preferred and the ambiguous one is either unattested or less frequent. 1.13.3 Postnominal position Montermini (2005:90) suggests that postnominal [anti-X]A constructions can be regarded in many cases as constituents of a compound: in this perspective panno antifebbre 'antifever cloth' could be conceived not only as a noun phrase ([pannoN antifebbreA]NP), but also as a compound ([panno antifebbre]N) with three internal constituents (this kind of compound is widespread in Germanic languages). Does the postnominal position favour the non-suffixed form? In order to investigate whether such factor has a consistent effect, a series of web searches was performed on Google. The aim was calculating if we have a clear predominance of postnominal contexts where non-suffixed forms are attested. 10 words were selected pertaining to couples of suffixed/non-suffixed PRAs (the first 50 relevant hits were considered for each of them). A reliable analysis should, in fact, consider PRAs with a concurrent suffixed form: where such condition does not exist, we cannot have clear indications because they would represent the only lexical item available to express a certain meaning. The following words were selected: antimissile (with concurrent suffixed antimissilistico) 'anti-missile' antimafia (antimafioso) 'anti-mafia' antigoverno (antigovernativo) 'anti-government' antiberlusconi (antiberlusconiano) 'anti-Berlusconi' antiterrorismo (antiterroristico) 'anti-terrorism'

47

prematrimonio (prematrimoniale) 'pre-marriage' precongresso (precongressuale) 'pre-congress' postelezioni (postelettorale) 'post-elections' postsisma (postsismico) 'post-earthquake' postolimpiadi (postolimpico) 'post-olympic' Non postnominal 119 (23,8%)

Postnominal Pref+N 381 (76,2%)

The results indicate a strong correlation between Pref+N structure and postnominal position. It seems that they are largely employed to form compound-like structures. As we will see in the next section, combining this aspect with contextual usages where the suffixed form is preferred can provide a clearer picture of the alternation. 1.13.4 Non-relational usages Although the object of this chapter was defined relational adjective, there is evidence of usages which are "often closer to qualifying adjectives" (Montermini 2005:90). Are suffixed ore non-suffixed forms more inclined to a qualifying/gradable usage? The issue was investigated through the analysis of anti- PRAs in contexts where they are employed as predicates, coordinated with other adjectives, modified by adverbs, and available in comparative forms. The queries were all performed on La Repubblica corpus. Many instances were found of clearly non-relational usages (226 in total): non vuole apparire meno anti-israeliano di quanto non siano A leggerlo , i passaggi pi anti-Bossi sono quelli in cui si de pu assumere atteggiamenti troppo anti-serbi senza rischiare un film molto anti-femminista, lui ha paura delle Pref+N Nonrelational 62 (27.4%) Pref+Adj 164 (72.6%)

48

The results show a clear tendency. Suffixed forms are preferred when PRAs have a qualifying function. Moreover, it must be noticed that many Pref+N constructions employed as non-relational adjectives are not ascribable to an alternant couple, but are rather probably coined on the spot for specific contextual communicative reasons, with no parallel suffixed form already established in the lexicon, e.g. che fra i Democratici il pi anti-Cossiga. D' Alema per adesso ro, cio Juve, e quindi molto anti-Inter. Anche se stasera a Rav In Gaeta & Ricca (2009)'s terms, such forms would be labeled [-lexical], since they do not enter the lexicon albeit based on a productive word formation pattern. In order to confirm this, a specific query was performed on a widely attested couple of adjectives, namely antiberlusconi and antiberlusconiano 'anti-Berlusconi', extracting all the instances cooccurring with predicate verbs, adverbs and coordinated forms. The results clearly show that the suffixed form is the only one employed in such contexts. No instance of the non-suffixed form was in fact found in La Repubblica corpus (against 13 suffixed). Maybe more interestingly, the result was strongly confirmed by a series of Google searches, yielding a vast majority of hits for the suffixed form: " antiberlusconi" (330 hits, 7%) " antiberlusconiano" (5,020 hits, 93%) "troppo antiberlusconi" (7 hits, 1%) "troppo antiberlusconiano" (468 hits, 99%) "pi antiberlusconi" (48 hits, 6%) "pi antiberlusconiano" (698 hits, 94%) "meno antiberlusconi" (1 hit) "meno antiberlusconiano" (100+ hits, 100%) "fortemente antiberlusconi" (5 hits) "fortemente antiberlusconiano" (245 hits, 98%) 'is anti-Berlusconi'

'too anti-Berlusconi'

'more anti-Berlusconi'

'less anti-Berlusconi'

'strongly anti-Berlusconi'

It should also be considered that the input for the suffixed form was the singular masculine (one out of four inflectional possibilities); searches including the other possible desinences would turn in stronger results. The general trend highlighted by

49

corpus data encourages the view that suffixed forms are preferred in non-relational usages because some contexts, for instance predicative usages, require more grammatical information to be expressed. In addition to this, the evidence that nonsuffixed constructions are preferred in compound-like complex words is strengthened, since modifiers in postnominal position typically accomplish a non-gradable function. The results produced by the investigation of the postnominal context and nonrelational usages shed new light on the "division of labour" at play between suffixed and non-suffixed forms. Although such forms cohabit and can reasonably be considered synonymous in most cases, corpus and web data show that their morphological structure has a consistent effect on some of the contexts in which they are employed. 1.13.5 Factorization Prefixal constructions can be factorized, i.e. if two coordinated words contain prefixes attached to the same lexical item, it can be omitted once avoiding a double, see for instance La radioterapia, pre- o post-operatoria, daltra parte
[http://www.aimac.it/stampa_tv_web.php?id=k6SUmA==]

The adjective operatoria is omitted in the first of the two disjuncted forms (preoperatoria). Does factorization favour one of the two concurrent forms? Being it a formal operation unrelated to meaning, we can expect that no relevant difference can be encountered. In order to ascertain this, specific web searches (La Repubblica corpus did not provide sufficient data) were performed aiming at finding several instances of conjuncted/disjuncted prefixal forms and assessing whether either the suffixed or the non-suffixed forms show a significantly higher presence. The first 100 relevant hits were considered for each search: "pre e post", "post e pre", "anti e pro", "pro e anti", "pre o post", "post o pre", "anti o pro", "pro o anti". Here a sample of the hits Pre o post-conciliari che siano, il Papa non ha
[http://paparatzinger2-blograffaella.blogspot.com/]

possono avere un'azione anti o pro-infiammatoria. Il dosaggio


[http://www.omegor.it/ricerca/omega3-sistema-immunitario.html]

foglio di adozione e controlli post e pre adozione


[www.ilcercapadrone.it/gatti16.htm]

un decreto anti o pro crisi? La maggior parte

50

[www.mininterno.net/fmess.asp?idt=2964]

The table below summarizes the results: +N +Adj

Factorized 445 (56%) 355 (44%) Prefixes The slight difference in percentage confirms the prediction that both constructions can be factorized. It is important to stress that, apart from the difference with non-suffixed forms, the possibility of factorizing suffixed ones is itself another piece of evidence on their morphological status. The formal segmentation of the word shows that language users perceive it as a combination of a prefix plus an adjective, i.e. in the same way as words which do not pose bracketing paradoxes. 1.13.6 Specialised meaning During the perusal of PRAs some forms were encountered which cannot be labeled as synonymous, although they exhibit the same structure of the synonymous couples. Multirazza 'multi-race' and multirazziale 'multi-racial' select two meanings of the base noun razza 'race', respectively the one related to the zootechnical field (animal breed) and the one related to the general concept of ethnic group: l'addestramento di cani multirazza e meticci fondato
[http://www.cani.com/allevamenti-di-cani/]

il cuore alla societ multietnica, multirazziale e multireligiosa


[http://archiviostorico.corriere.it/2003/giugno/19/]

The data observed demonstrate that some cases of alternation are semantically marked: although the derivational possibilities of a noun with two or more meanings could lead to both a suffixed and a non-suffixed form (e.g. l'addestramento di cani *multirazziali; societ *multirazza), each form is assigned a unique meaning, related to one of the word senses. In other cases, where no clear distinction between senses can be established, a neat divergence can be found in the usage contexts from a semantic viewpoint. In order to have clear indications on the context, the "Word Sketch" function of the Sketch Engine platform (Kilgarriff et al. 20045) was exploited. It provides

51

a corpus-based automatic summary of a lemma's grammatical and collocational behaviour. Analysing the word sketches (from the ItWaC corpus) of several couples of PRAs, it was possible to ascertain what words systematically co-occurred and thus get a picture of the meaning expressed by each form. Let us consider the word sketches of the multifunzione / multifunzionale couple. The four most frequent co-occurring lexical items were selected: multifunzione display 39 apparecchio 23 schermo 10 fotocopiatrice 8 multifunzionale sportello 127 agricoltura 115 ruolo 84 polo 25

In this case it is clear that multifunzione is restricted to the domain of concrete objects ('display', 'appareil', 'monitor', 'photocopier') provided with multiple functionalities and features (e.g. a fotocopiatrice multifunzione 'multifunctional photocopier' can scan, print etc.). Multifunzionale, on the other hand, seems to be rooted on the more abstract meaning of funzione, i.e. that of "purpose" and "service", so that nouns modified by it are generally characterized by a wide range of purposes and/or services: a ruolo multifunzionale 'multifunctional role' is characterized by many contemporaneous functions. Where no sufficient data was available on ItWaC, collocational information was searched through WebCorp (Kehoe & Renouf 20026). According to WebCorp, for instance, the most plausible collocation candidates for multirazza and multirazziale are respectively allevamento 'farm' and societ 'society'. Other couples of PRAs were evaluated according to their collocational behaviour, and for some of them a clear trend emerged towards the assignment of specialised meanings (associated to different senses of the noun base) to each form. 1.13.7 Language-specific factors An investigation on the alternation cannot disregard language-specific features which help better focus the phenomenon within the grammatical constraints posed by each language. Montermini (2005) looks at the attestation of suffixed and non5. Available online at http://www.sketchengine.co.uk/ 6. Available online at http://www.webcorp.org.uk

52

suffixed anti- adjectives. It is suggested that the possibility of having invariable adjectives like antifebbre is highly correlated with the language-specific presence of simplex invariable adjectives. As highlighted, the morphological process which [licenses] this kind of adjectives should be (almost) exclusively active in languages which independently stipulate the existence of invariable adjectives, as in the case of Italian. On the contrary, in languages where [it] is rarer if not impossible, the raising of this kind of derivates should be highly disfavoured (Montermini 2005:92, my translation). Montermini (2005) carries out his case study considering English, Italian and Russian data. These languages differ in the above mentioned property, since English licenses invariable adjectives (they are rather obligatorily invariable), Italian licenses both variable and invariable (see e.g. alto/a/i/e 'high' and viola 'violet') and Russian only admits variable adjectives. The prediction that each of these languages should behave differently is confirmed by the attestation of non-suffixed anti- adjectives (see the figure below, translated from Montermini 2005:94). Pref+N English Italian Russian 200 (73%) 84 (43,7%) Pref+Adj 72 (26,5%) Total 272

108 (56,3%) 192 89 (100%) 89

As expected, Russian did not display any invariable adjective among the selected formations in anti-. I performed a survey of Russian prefixed adjectives looking at other prefixes, with the help of a native speaker. The results obtained in a small random sample (100 hits) confirm Montermini (2005)'s conclusions, since no adjective was found with no suffixal marking. #458489 - , post.soviet.Suff 'postsoviet' In order to test the language-specificity of the alternation, I performed a similar analysis on two web corpora of the Dutch language (Kilgarriff et al. 2009a) and Portuguese language (Kilgarriff et al. 2009b), aided by two native speakers. A query

53

was performed in order to extract all anti- adjectives. The aim was assessing whether the morphological configuration of PRAs in Dutch and Portuguese was influenced by their adjectival morphology. See for instance Portuguese: Especialistas em luta antiterrorista e no complexo puzzle de sistemas electrnicos de defesa antimssil, em uso na aviao anticorpos radioactivos anti-cancro (ou seja, dirigidos como os tempos da luta anti-corporativa, defendendo o protagonismo devido aos protestos antiguerra, motivou crticas dos donos iedereen is eigenlijk in een soort anti-Bush sfeer hier, waardoor voorspellen dat een grootscheepse anti-terrorisme campagne Ze infiltreerden er in terroristische anti-Cubaanse organisaties Europese hoofdsteden in ruil voor anti-Russische retoriek de

Dutch:

Among the adjectives prefixed with anti-, a count was made of formally Pref+N and Pref+Adj adjectives. The results are below: Pref+N Dutch Portuguese 64 (36.2%) 74 (32.2%) Pref+Adj total

113 (63.8%) 177 156 (67.8%) 230

Portuguese adjectives were mostly of the Pref+Adj type (67.8%). This is consistent with the grammatical constraints of the language, which require, as in the case of Italian, the marking of both gender and number but license invariable adjectives. Dutch data displayed a slightly different trend: the percentage of Pref+Adj adjectives is quite high (63.8%) even though the inflectional morphology requires marking in a limited range of cases, i.e. only in the singular form introduced by indefinite articles. A necessary remark is that Dutch probably displays a less clear borderline between derivation and compounding: as a matter of fact, many anti- constructions were embedded into compounds: Wat nu aan de gang is met de antidiscriminatiewetgeving, doorbreekt dit principe. 'antidiscrimination legislation'

54

This may lead to the conclusion that in Dutch the same modifier function is realized through both adjectives and compounds' constituents (many instances of adjectives have, however, a rigid prenominal position), so that the final number of Pref+N adjectives is inferior to the expectations.

1.14 Processing issues


As pointed out by Bertinetto (1995:7), it is important to ackwnowledge that compositional and non-compositional patterns in morphology are related to both production and comprehension, so speakers "make use of both compositional and decompositional procedures". A theory of word formation must thus cope with both structural issues and processing mechanisms. The cognitive operations performed by the speakers can shed new light on grammatical idiosincracy and/or inconsistency and help unveiling some of the factors involved in the structural makeup of words. I will now provide a brief review of some processing issues related to derived words, which will contribute to motivate discrepancies in the meaning/structure interface of PRAs. It seems reasonable to assume that bracketing paradoxes in PRAs can be related to general properties of prefixation and suffixation in world languages. On the basis of Greenberg (1966)'s observations on prefixing and/or suffixing languages ("There is a distinct predominance of suffixing. Exclusively suffixing languages are fairly common, while exclusively prefixing languages are quite rare"), Hall (2000) reports that the presence of suffixation strongly exceeds that of prefixation across a wide sample of world languages. This has led scholars to consider the suffix class prototypical in the domain of affixes and to speak of a "suffixing preference". A number of arguments have been proposed in order to account for such asymmetry. Cutler et al. (1985) argues that the overall preference for suffix morphology results from the way language is processed by its users: word onsets are more psychologically salient and stems are more likely to be processed before affixes. Could this elements speak in favour of structural mismatches in PRAs? In a purely formal perspective, the bracketing paradox issue could be analysed in terms of a competition between prefix and suffix towards the attachment to the innermost constituent: Prefix - [Noun - Suffix] vs [Prefix - Noun] - Suffix. Prefixes, however, do not produce formal changes on their bases, unlike suffixes. This suggests that the morphosyntactic formal properties of the output word can be affected by such morphological asymmetry, ovveruling compositional semantics in the word formation process.

55

Another piece of evidence related to the processing of PRAs comes from so-called "prefix-stripping" models. Studies on visual and auditory recognition of prefixed words (see Taft & Foster 1975) have claimed that they are analyzed into their constituent morphemes before the occurring of lexical access, so that a PRA like antifebbrile would have its adjectival root febbrile accessed before the lexical access to the whole word. Establishing direct causal relations between processing strategies and attested morphological structures is far from unproblematic: however, the prefix-stripping model can shed light on the fact that bracketing paradoxes in PRAs allow the access to an adjectival form which is already in the lexicon. A non-paradoxical form would in turn require the access to the lexical item as a whole. Again, as long as this does not pose comprehension problems to the language user, it is reasonable to assume that a form containing independently accessible words is preferred. In addition to this, a paradoxical form probably allows a faster recognition of the noun base because of its paradigmatic relationships with the adjectival form embedded. A further issue involved in the processing of PRAs is the role of hierarchical morphological structure. It might be argued that PRAs' special morphosyntactic configuration is somehow easier to process than hypothetical concurrent forms. The bracketing paradox involved concerns the analysis of a PRA according to a leftbranching (expected on a semantic basis) vs a right-branching structure, actually observed, and the only one which can account for allomorphic adjectival forms, as can be seen below.

Libben (1994) assessed the role of morphological structure in lexical processing. A patient who showed sensitivity to the formal complexity of multimorphemic strings was tested in a naming latency experiment with affixed nonsense roots that were constructed so that the resulting complex nonwords were either left-, right-branching or illegal (see figure below, from Libben 1994:52)

56

Libben argues that due to his deficit-related diminished resources interacting with the computational demands of normal lexical processing "he might show effects of differences in hierarchical structure that could not be detected in the investigation of normal subjects" (Libben 1994:51). Among the morphologically legal nonwords, the subject showed more difficulty with left-branching structures than with right-branching ones. As PRAs show a right-branching structure where semantics demands a leftbranching one, it seems plausible to imagine that (although the present evidence was obtained through verbal, not adjectival stimuli) the ease of the processing can be related to the preference towards a semantically divergent morphology.

1.15 Conclusions
The previous sections have tried to outline an analysis of the phenomenon of bracketing paradoxes in the domain of prefixed relational adjectives. Unlike many scholars treating bracketing paradoxes as a "bug" in morphological theory, I have tried to offer a broad characterization of its instantiation in Italian taking into account the multifaceted nature of word formation, without any claim of exhaustiveness. It was apparent from the discussion that data did not allow to establish clear-cut boundaries. Corpus and web data, however, combined with the review of some of the main word formation models, have contributed to shed light on some aspects related to the presence of BPs in PRAs, as well as on their morphological, syntactic and semantic properties. The combinatorial possibilities of prefixes and suffixes have shown major drawbacks in theoretical models conceiving PRAs as a special word formation phenomenon to be assimilated to parasynthesis. The forms addressed in the chapter display a structure which does not challenge binarism, although their structure is unexpected on a semantic basis. The categorial-only function of the suffix (endorsed by Corbin 1999) has been unveiled providing arguments related to the semantics of PRAs, looking at Italian data in which non-relational suffixes were employed as "paradigm integrators", and observing how speakers were able to map PRA-internal anaphors in an acceptability test.

57

Thanks to comprehension data elicited from native speakers, it was also possible to observe how novel instantiations of the described constructionist template were easily integrated. The alternation of suffixed vs non-suffixed forms allowed to better focus their common nominal nature, and provide some generalisations about the factors favouring either form and their respective role in the lexicon. A rapid treatment of issues related to language processing has helped us establish a bridge between attested structural morphological properties and the way speakers naturally process complex words: although no direct conclusion can be drawn, both the preference for multimorphemic right-branching words and prefix-stripping theories of lexical processing can provide clues about the potentials reasons for some crosslinguistic aspects of lexicon's structure. Bracketing paradoxes in the domain of PRAs can thus be conceived as a natural state of affairs in Italian. On the basis of a range of structural and semantic properties they fit perfectly in the Italian adjectival system and represent a productive word formation pattern.

58

2. Bracketing paradoxes in phrasal morphology

2.1 Aim of the chapter


This chapter will deal with phrasal bracketing paradoxes (henceforth phrasal BPs), i.e. those phrasal constructions in which a mismatch occurs between morphosyntactic structure and semantic reading. As can be seen from the English electric guitarist, some complex expressions appear to license two competing accounts of their morphosyntactic structure, each of them capturing just one level of representation, namely either grammatical (i) or semantic (ii): (i). [electric] [[guitar][-ist]] (ii). [[electric] [guitar]] [-ist] The research to date has tended to focus on the solution of the theoretical puzzle rather than on its description as the result of a productive pattern in World languages. Far too little attention has been paid to the role of BPs in the lexicon/syntax interface. The majority of studies on phrasal BPs were skewed towards English data: as we will show, language-specific issues will shed light on phrasal BPs' formal properties. The morphosyntactic and semantic properties of phrasal BPs raise interesting questions about the role of phrases in word formation and enhance our understanding of the blurred boundary between the notions of word, phrase, and lexical unit. Previous studies on phrasal bracketing paradoxes focussed on personal nouns (at least since Spencer 1988), i.e. expressions denoting the actors of specific (often professional) activities e.g. modern linguist, transformational grammarian, baroque flautist etc. They have not dealt with the phrase-based derivation of adjectives e.g. energetico solare 'solar energetic', scolastico dell'obbligo 'related to compulsory education' etc. In order to contribute to fill this gap, a wide documentation of phrasal BP constructions in Italian will be provided: both from the domain of personal nouns and from that of phrase-based adjectives. After a survey of the relevant literature and its critical assessment with respect to Italian data, a descriptive section will highlight the formal and semantic properties of phrasal BPs in Italian. Some space will then be devoted to the assessment of how language users cope with this phenomenon in comprehension and production.

59

2.2 Phrasal bracketing paradoxes As noticed by Scalise (1994:250, my translation) a nuclear physicist is not a "physicist who is nuclear": the semantic constituent "nuclear physics" does not correspond to any morphological constituent. We can observe that phrasal nouns are actually NPs whose interpretation is not the same of head+modifier NPs. In addition to adjective + noun NPs, Romance languages display a set of phrasal BPs involving a PP instead of an adjective, e.g. It. calciatore a cinque lit. soccer.player to five 'futsal player', tennista da tavolo 'table tennis player'. Kerleroux (2007) addresses the semantics of pcheur sous la glace 'fisherman under the ice' and similar French constructions. Referring to a picture seen in Le Monde with the caption "Pcheurs sous la glace", she reports she "looked at the picture, trying to discern, under the ice, any fisherman. But the fishermen were on the ice, where they were putting a wire and a hook". At first glance, in both constructions it is hard to interpret the APs and PPs as attaching to the complex noun. A major challenge in the relation between modifier PPs and APs is that posed by Lexical Integrity issues. In a nutshell, the LIH (Bresnan & Mchombo 1995) assumes that the internal structure of words is not accessible to syntax processes. This implies that in an AP/PP + complex noun NP the former cannot "see" the word's internal components and have a semantically restricted scope over any of them. Let us consider the following examples: good truckdriver skilled designer In the above phrases the adjective can only have its scope on the complex noun, so a good truckdriver is not a driver of good trucks and a skilled designer is obviously a designer who is skilled. Unlike the examples mentioned, the structure of some phrases appears to cast doubt on the validity of the LIH. Booij (2009:146) discusses Dutch NPs in which "[...] the adjective may have scope over the first constituent of the complex word only, rather than over the complex word as a whole": a [[A]A[Nsuffix]N]NP wetenschappelijke onderzoeker ('scientific researcher') is not a researcher who is scientific, but rather someone who carries out scientific research". Likewise, Kerleroux (2007:141), assuming that "the syntax neither manipulates nor has access to the internal form of words" (Anderson 1992:84), wonders whether a syntactic

60

relationship can hold between a constituent (the PP sous la glace) and the internal part of a word, i.e. the verbal base pcher 'to fish' from pcheur 'fisherman'. One of the first attempts to offer a formal representation of this kind of constructions, capable of reconciling the demands from morphology and semantics, is that provided by Williams (1981). The notion of Lexical Relatedness stipulates that two items can be related if they only differ in one head position (see figure below). The Lexical Relatedness principle thus accounts for the conceptual link between nuclear physicist and nuclear physics despite the absence of any derivational clue.

The long-lasting attempt to solve the BP has turned attention away from the issue of explaining the nature of this construction. The purely formal treatment of BPs was based either on postulated additional representational levels (Williams' Lexical Relatedness) or on complex transformational mechanisms "repairing" the surface structure, mainly by means of manipulating the internal word structure in order to guarantee the respect of selectional restrictions, e.g. Pesetsky's (1985) Affix Raising. Approaches of this kind carry with them various limitations when dealing with nonconcatenative morphology, where no clear segmentation can be made and it is problematic to identify morphemes in order to specify their hierarchical position and postulate their internal movement. Consider the Italian cases below, which display an increasing loss of transparency (including suppletive formations): Transparent Back-formation Suppletive chitarra elettrica chitarrista elettrico 'electric guitarist' 'cognitive linguist' 'labour lawyer'

linguistica cognitiva linguista cognitivo diritto del lavoro giurista del lavoro

A word like chitarrista is transparently derived from the noun chitarra, so no problems are posed for theories involving operations on the affix -ista. On the other hand, the relationship between linguista ''linguist' and linguistica 'linguistics' is not so transparent and thus discourages a similar treatment.

61

One major step forward has arisen out of Spencer's (1988) insight into the prominent role of the paradigmatic dimension. In Spencer's view, the key issue is represented by the global word formation process which is not restricted to the word itself, but connected to the availability of other related lexical units which license the BP phrase by means of proportional analogy.

Spencer points out that the phrasal expression on which the final construction is based must be lexicalized, i.e. stored in the lexicon. In our example, we could say that unlike electric guitar, an unlexicalized phrase like big guitar could not lead to big guitarist (in the sense of 'player of big guitars'). Likewise, it is stressed that "the second prerequisite is that the meaning of the individual components be identical". That is, the sense of guitar in guitarist and electric guitar must be the same.

2.3 Language-specific factors


Looking at data on BPs coming from languages other than English, it is rather clear that theoretical models approaching the issue cannot but face the different manifestations of the phenomenon. English-based accounts have long ignored crucial language-specific features, essential in both the definition of the BP issue and in its description. I will now provide an outline of two major grammatical aspects of the language-specific instantiations of phrasal BPs. The issue of word order is crucial in the definition of BPs as "paradoxes". The purely morphological phenomenon (i.e. BPs attested at the word level) was traditionally conceived as a mismatch between a phrase's semantic scope relations and its structural makeup, since "[...]while the affix seems to be merged with the second constituent following its own usual selection and allomorphy patterns suggesting a [ + [ + ]] analysis, semantically it has scope over the complex base, supporting an opposite [[ + ] + ] analysis" (Bisetto & Melloni 2008:7)

62

A theory involving the assignment of an unambiguous structure is concerned with the choice of the most appropriate bracketing. Since there are two possible bracketings, [ + [ + ]] and [[ + ] + ], the paradox arises because both display some kind of inconsistency, i.e. they both lack a match with either the semantic or the grammatical level. The question to be anwered is why they are both possible bracketings of the word. If we look at English, it is clear from both the word and the phrase level that two alternative structures can be conceived. This paradox is made possible by the linear order of morphemes and the basic syntactic word order of English, which license the formation of structures XYZ in which Y can be attached to both Y and Z. If we take a look at Italian, however, we have a different situation. On the one hand double bracketing is still possible at the word level, but on the other hand most phrases cannot be conceived as structurally paradoxical. Due to Italian adjective-noun basic order, chitarrista elettrico cannot receive two bracketings, because of the constituents' position. Word Eng Ita electric guitarist chitarrista elettrico Bracketing 1 [electric [guitarist]] [[[chitarr(a)]ista] [elettrico]] Bracketing 2 [[electric guitar]ist] ?

English-only studies focussed their insight on the parallel ambiguity of affix semantic and structural scope. In the phrase domain, it is evident that while the external position of the affix allows English to have a double bracketing, Italian cannot provide an alternative structural makeup of the word, making the affix scope mismatch unsolvable by means of linguistic notation. It seems to me that, from a comparative perspective, this finding can itself help to enhance our understanding of the phenomenon. As I have reviewed in the previous section, Spencer's approach to BPs gets rid of the assumption that they must be "repaired" by means of additional levels of description and/ or transformational mechanisms. Turning attention to the paradigmatic level, Spencer both recognizes the crucial role of what is already available in the lexicon and identifies the close relation between phrasal BPs and their alleged base, that is to say a normal phrase. Booij (2009:96) argues that the fourth proportional scheme proposed by Spencer works only up to a certain extent: ln Spencers (1988) analysis, the restricted scope interpretation of transformational grammarian is related to the existence of a lexical

63

unit transformational grammar and the word pair grammar-grammarian, and therefore analysed as an analogical formation. However, restricted scope also occurs in cases where such an explanation cannot be invoked. ln the examples of restricted semantic scope given above such NPs are not available. For instance, [...] there is not a well-formed phrase wetenschappelijke onderzoek scientific research that can be related to wetenschappelijke onderzoeker scientific researcher since the correct phrase is wetenschappelijk onderzoek, without a final schwa on the adjective. The same holds for Italian, where there is no independent *chitarra elettrico on which chitarrista elettrico is based. The gender (and number) agreement discourages the view that a straightforward analogical scheme applies, and highlights the mismatch between the grammatical and semantic properties of the adjective in the final BP outcome. Booij (2009:97) claims that this clearly shows that "word-internal structure must be visible to rules of semantic interpretation" (see figure below).

Italian phrasal BPs are made up of a noun and an adjective agreeing in number and gender. The analogical formation process fails to predict this aspect of the final expression. As a matter of fact, this is caused by Italian inflectional morphology, that unlike English obligatorily marks number and gender in adjectives. Consequently, we have an adjective whose inflection is unpredictable just on the basis of the adjective's occurrence in the base phrase (elettrico is inflected in masculine, whereas the base phrase chitarra elettrica displays a feminine adjective).

64

An analogical scheme thus does not straightforwardly apply to form what we call phrasal BPs. Although strong arguments speak in favour of analogy playing an important role in the emergence of such constructions (e.g. allomorphic complex nouns), a fourth proportional scheme does not capture some facets of the phenomenon. I will argue that this is one of the central issues their linguistic hybrid status. As we have seen from the above examples, the examination of two instances in languages other than English has provided us with a clearer picture of the phenomenon. Italian data, in particular, not only did challenge the analogical approach, unfamiliar with the treatment of complex noun - adjective agreement (in the same way as Dutch), but also offered clear counterevidence to long-standing theoretical models built upon the notion of paradox. Ackema & Neeleman (2004) identify two mapping principles (between morphosyntax and morphophonology) which are relevant for the discussion on phrasal BPs. The Linear Correspondence (Ackema & Neeleman 2004:140) predicts that if X is structurally external to Y, X is phonologically realized as /x/, and Y is phonologically realized as /y/ /x/ is linearly external to /y/.

then

The Linear Correspondence principle thus predicts that affixes are phonologically external to the base they are attached to. Ackema & Neeleman (2004) claim that this mapping principle can be violated because of "conflicting demands". This can be represented by the Input Correspondence, stating that if an AFFIX selects (a category headed by) X,

65

then

the AFFIX is phonologically realized as /affix/, and X is phonologically realized as /x/, /affix/ takes /x/ as its host

According to this principle, the position of the affix is ruled by the base's head and can thus violate standard Linear Correspondence. This is the case of Italian phrasal BPs: a conflict between Input Correspondence and Linear Correspondence is triggered because of the left-headedness of the bases involved. Let us consider the already mentioned example chitarrista elettrico 'electric guitarist'. Deriving from chitarra elettrica 'electric guitar', the output of derivation could conform to either principles: Linear Correspondence: Input Correspondence: chitarraX elettricaY > chitarraX elettricaY > *chitarraX elettricaY +istaZ chitarraX +istaZ elettricaY

Only the input correspondent mapping is grammatical in this case, since the affix istaZ is attached to the head of the noun phrase (i.e. X). The affixation on left-headed constructions must "disobey" one of the two constraints, since the affix cannot contemporarily be external and hosted by X: the result is a [[/x/ /affix/] /y/] construction where standard bracketing is hindered by the "internal" phonological realization of a semantically "external" affix. Some remarks are needed on the number/gender agreement between the derived noun and the adjective. It seems that the formal properties of the phrasal base are maintained throughout the derivational process, so that the agreement is present both in the input and in the output. We can label this as expletive agreement, since the semantic properties related to gender and number are not relevant and the formal agreement has the only function of marking the dependency between noun and adjective.

2.4 Phrases as lexical units


We will now dwell on the role of phrases in the lexicon. The aim of this section will be to highlight how phrasal BPs are rooted in the possible functional equivalence between words and phrases. The role of phrases in word formation has been longly debated, at least since Aronoff (1976:21) "All regular word formation processes are word based. A new word is formed by applying a regular rule to a single already existing word" and Botha (1984:137)'s No Phrase Constraint (henceforth NPC), stating that "lexical rules do not apply to syntactic

66

phrases to form morphologically complex words". Bresnan & Mchombo (1995:181) argue that "words are built out of different structural elements and by different principles of composition than syntactic phrases. Specifically, the morphological constituents of words are lexical and sublexical categories - stems and affixes". These theoretical positions have led many scholars to seek potential counterevidence in the form of words containing phrases as constituents. Lieber (1992:11) (i-iii) and Bauer (1983:164) (iv-v) point out English compounds with a phrasal non-head: i) [over the fence] gossip ii) an [ate too much] headache iii) [God is dead] theology iv) [what do you think?] movement v) [dont tell me what to do] look In such compounds phrases of different types have come to occupy the constituent spot traditionally reserved to words. Bresnan and Mchombo (1995:194) address criticism towards the NPC stating that where syntactic phrases appear to undergo morphological derivation, it is by virtue of their being lexicalized. Although lexicalization can be innovative, the nonsyntactic status of lexicalized phrases embedded in word structure can be detected in properties such as lexical gaps, and can be confirmed by the other lexical integrity tests. Bresnan and Mchombo maintain that violations to the NPC apparently occur because phrases are lexicalized and thus do not pose a real challenge because of their "nonsyntactic" status. Lieber & Scalise (2006) suggest that the insertion/modification test (probably the best known for distinguishing lexical forms from syntactic ones) discloses challenging findings in Italian phrasal compounds, e.g. produzione scarpe 'shoe production', whose constituents reveal a partially syntactic nature being modifiable by adjectives. Let us consider the examples below (from Lieber & Scalise 2006:15): produzione accurata scarpe lit. production accurate shoes 'accurate shoe production' produzione scarpe estive lit. production summer shoes 'summer shoe production' prod. accurata sc. estive lit. prod. accurate shoes summer 'accur. sum. shoe prod.'

67

The authors conclude that an interplay between morphology and syntax must not be excluded a priori. With respect to phrasal BPs, it is important to stress that Englishonly accounts have equated expressions containing phrasal constituents with phrasebased expressions, i.e. old maidish with transformational grammarian, both represented as suffixed phrases: [[old maid]Nish]A [[transformational grammar]Nian]A As we have already pointed out, left-head languages offer clear counterevidence to this generalisation, because if an adjective modifying a noun is on the right side, the affix scoping on the whole phrase will not always find itself on the right side. Italian displays two different word formation patterns involving phrases. Let us consider the following examples: (i) [[pan dolc(e)]Naio]N (ii) [chitarrista elettrico]N Whereas the former can be represented as a lexicalized phrase + an affix, the latter is conceivably a phrase which underwent derivation (chitarra elettrica 'electric guitar' > chitarrista elettrico 'electric guitarist'), so that no more than one bracketing structure can be stipulated. Can syntactic phrases be the base of word formation processes? In his considerations on the relationship between syntax and morphology, Botha (1984:137) specifies that syntactic rules are excluded not only from the class of word-formation rules, but also from the class of rules that form the bases to which word-formation rules may apply. Bresnan and Mchombo (1995:190) defend the position that most "phrasal" constructions are often based on quotes or foreign expressions: a mea culpa look the ich bin ein Berliner speech a certain je ne sais quoi quality

68

We will not argue that every possible phrase can enter paradoxical constructions like (ii). Data, however, confirm that a wide range of phrasal bases can enter word formation, albeit not quotative and/or foreign. A certain degree of "fixedness" is required, but Italian phrasal nouns display a structure with clear syntactic properties, both in derivation input and output (N/A number and gender agreement). Sprenger (2003:4, cited in Booij 2009) delves into the notion of "fixed expression": Fixed expressions (FEs) refer to specific combinations of two or more words that are typically used to express a specific concept. [...] The defining feature of a FE is that it is a word combination, stored in the Mental Lexicon of native speakers, that as a whole refers to a (linguistic) concept. As highlighted by Booij (2009:219) the naming function is central with respect to this issue. Apart from complex and simplex word, in fact, phrases too can establish a direct link between an expression and a concept. The phrasal naming strategy is one among the many mechanisms for the formation of new names: De Caluwe (1990) enlists phrase constructions, borrowings, brand naming, acronyms, clippings and semantic extension (in addition to standard word formation, i.e. derivation and compounding). So phrases can perform the same function of words and give rise to lexical units. Let us consider some examples: Italian Spanish Greek English mulino a vento 'windmill' guerra fredda 'cold war' luna nueva 'new moon' (Rainer & Varela 1992) tritos kosmos Third World (Ralli & Stavrou 1998) french fries

Italian phrasal BPs are rooted in the functional equivalence between phrases and words. Both can be bases for word formation processes. In the literature pertaining to both the NPC and bracketing paradoxes, the boundary between words and lexical units has been ignored. As we have illustrated, although a certain degree of fixedness is a central feature of phrasal naming units, this does not imply that phrasal units involved in word formation should be equated with words (lexemes, syntactic atoms). Booij (2009:221) concludes that "lexical units can be constructed by means of syntactic rules".

69

The constructionist approach to word formation assumes as its starting point the hypothesis that there is no evidence of a strict division between what scholars have labelled "lexicon" and "syntax" (see Chomsky, 1965). According to Goldberg (1995) both lexical and syntactic constructions form essentially the same type of "[...] data structure: both pair form with meaning". On this view, the boundary is not clear-cut, but blurred, due to basic commonalities between the two constructions. This assumption allows the theoretical description to combine structures pertaining to both traditionally separated levels and let them cohabit in order to provide an in-depth characterization of linguistic phenomena. For the same reason, rules are not accepted, and it is rather put forward an all-encompassing notion of generalization as a primary source of linguistic processes. Within this framework, Masini (2009) focuses phrasal lexemes, "a kind of intermediate unit between syntax and the lexicon", analysing the properties of Italian Noun + Adjective and Noun + Preposition + Noun constructions in Italian. Masini (2009:259) shows that some tests may be applied which demonstrate phrasal lexemes' paradigmatic and syntagmatic cohesion (and thus lexical nature), e.g. paradigmatic variation is blocked, since words cannot be replaced by near-synonyms, which pose no problems for standard phrases: casa di cura lit. home of treatment nursing home *abitazione di cura lit. dwelling of treatment Furthermore, the constituents cannot be "individually modified". Neither number inflection on non-head nor adjectivation with narrow scope can be realised: *casa di cure lit. home of treatments *una casa accogliente di cura lit. a home cosy of cure a nursing cosy home Phrasal lexemes display a certain degree of lexicalization: some remarks should be done on this. The notion of lexicalization has received many definitions and the boundaries between the conceptual and the formal side of it have not always been very clear, i.e. scholars have often referred to lexicalization addressing some possible peculiar formal counterparts, e.g. univerbation (in grammaticalization processes, a complex expression may be reanalysed as a single unit and be assigned a new function/meaning, see e.g. Bybee 1985:18). Bauer (1983) operates a distinction between syntactic, semantic, morphological and phonological lexicalization. The notion of semantic lexicalization is however far from that required in the present discussion. Bauer

70

(1983:55) presents instances of compounds and derivatives which lost and/or lack semantic compositionality, so that the notion has often been considered synonymous of idiomatization (Lipka 2002:113). As we have seen, phrasal lexemes and the potential "bases" for phrasal BPs are not inherently idiomatized. It is thus maybe more fruitful to refer to the notion of fixedness rather than lexicalization and/or idiomatization. Sprenger (2003:4, cited in Booij 2009) states that typical examples of fixed expressions (FEs) [...]often have an opaque meaning or a deficient syntactic structure, for example, by and large or kick the bucket. However, these properties are not essential. [FEs are] non-compositional in the sense that the combination and structure of their elements need not be computed afresh, but can be retrieved from the Mental Lexicon. However, the degree of lexical and syntactic fixedness can vary. Booij (2009:221) highlights that "a fixed expression may be completely compositional, but nevertheless stored because it is a conventional name for a particular concept". In sum, the "bases" of BPs are fixed expressions with independent constituents (no univerbation); their meaning can have different degrees of idiomatization and this is not a prerequisite. Masini (2009) clearly illustrates the semantic affinity between morphological and phrasal lexemes (see figure below, from Masini 2009:268).

Unlike regular phrases and morphological lexemes, characterized by both either compositional or non compositional form and meaning, phrasal lexemes manifest a dissociation between form and meaning, i.e. a syntactic form and a "morphological" meaning (naming a unitary concept, replaceable with an either complex or simplex word). It is reasonable to assume that this mismatch is at the bottom of the bracketing paradox phenomenon, since phrasal lexemes are not syntactic atoms albeit displaying a semantics licensing unitary derivation, and cannot thus undergo a derivational process as words do. Derived phrases maintain their syntactic internal dependency, e.g. the

71

number/gender agreement in the case of Noun + Adjective constructions: all four Italian declensional possibilities (masculine or feminine and singular or plural) can be realised: chitarrista elettrico 'electric guitarist'
m.sing. f.sing. m.plur. f.plur.

un eccellente chitarrista elettrico e [http://www.hwupgrade.it/forum/] promettente chitarrista elettrica, avevo [http://italiano.peterschaefer.info/] Molti chitarristi elettrici preferiscono [http://gobilumusic.wordpress.com/] violiniste che chitarriste elettriche, ci non [http://forum.musicoff.com/]

Summarising, phrasal BPs are rooted in the functional equivalence between words and a certain class of phrases which, albeit maintaining their structural and formal properties, have been institutionalized and brought to express a unitary concept. The semantic equivalence of the latter with morphological lexemes itself licenses derivation; their formal properties prevent derivation from being formally realised on the phrase as a whole (i.e. externally). The output of the derivational process is thus a phrasal lexeme in which a derivational morpheme can be found which has scope on the whole expression but is attached to only one constituent.

2.5 Semantic issues


In order to provide a clearer picture of phrasal BPs, some remarks on their semantics are necessary. The main point to address is to test their being "special" under a theoretical viewpoint. This section will outline some of the issues related to the reasons which set phrasal BPs apart from other type of constructions, addressing both Noun + Adjective and Noun + Prepositional Phrase types. At least since Beard (1995), some claims were advanced against the existence of Noun + Adjective bracketing paradoxes. Beard (1995)'s claim is semantic in nature: he considers bracketing paradoxes from a semantic scope perspective (i.e. the word internal scope of the adjective in a construction like criminal lawyer), and argues that similar issues concern some cases of underived head nouns. In the examples below, no ambiguity between a base vs complex word adjectival reading can arise, since their structure is different from attribute phrases like first violinist, i.e. involving a complex word (violin+ist) and thus cannot be related to different word formation levels: good athlete usually refers to someone who is good as an athlete, rather than "an athlete who is good (as a person)". According to Beard, this would support the claim that BPs are not special in their semantic composition, since instances of compositional ambiguity can be

72

found also among phrases with unequivocal morphosyntactic structure, namely made up of simplex words. good athlete old friend 's.o. good as an athlete' vs. 'an athlete who is good as a person' 'actor in an old friendship' vs. 'old actor in a friendship'

In order to provide a semantic notation of the adjectival modification process for both structurally ambiguous and unambiguous phrases, Beard analyses the emergence of multiple meanings in terms of featural composition among constituents, i.e. leading the adjective to compose with one of the features of the head noun's semantic representation. Be it a simplex (e.g. friend) or a complex (e.g. lawyer) word, the adjective will modify one particular feature of the head, licensing two different semantic readings (see figure below, from Beard 1995:34).

Although Beard puts forward interesting insights on the similarities between phrasal BPs and ambiguous phrases, some distinctions should be laid out. It should be clear from the previous sections that according to the position defended here phrasal BPs are not analysable in terms of noun-adjective combination. Number and gender agreement of the adjective in Romance languages always follows the complex noun, so it would speak in favour of a misleading semantic composition (a chitarrista elettrico is not a chitarrista who is elettrico). This additional mismatch should contribute to defining phrasal BPs as a special class of expressions. Phrasal BPs represent an instance of phrase-based word formation: this means that the adjective has no semantic relationship with the derived noun, since the latter "comprises" the former, i.e. has scope over the whole original phrasal lexeme (e.g. chitarra elettrica) and has no proper modifying function. In most cases, the adjective could not be analysed as a modifier because of semantic selectional restrictions, e.g. [-HUMAN] adjectives. The adjectives involved cannot be analysed as intersective (the semantic composition cannot be characterized in terms of the intersection of noun and adjective extensions): let us consider chitarrista elettrico vs. chitarrista famoso 'famous guitarist'. We can assume

73

that whereas the latter can be defined as the intersection of chitarrista and famoso, the former cannot: chitarrista () famoso () [[ ]] = [[]] [[]] chitarrista () elettrico () [[ ]] [[]] [[]] 'famous guitarist' 'electric guitarist'

The same holds for the following examples: sociologo () svizzero () [[ ]] = [[]] [[]] sociologo () criminale () [[ ]] [[]] [[]] sciatore () professionista () [[ ]] = [[]] [[]] sciatore () alpino () [[ ]] [[]] [[]] flautista () anziano () [[ ]] = [[]] [[]] flautista () dolce () [[ ]] [[]] [[]] 'Swiss sociologist' 'criminal sociologist'

'professional skier' 'alpine skier'

'elderly flutist' 'sweet flutist'

The representation of phrasal BPs' semantic composition must take into account the properties of the base (a phrasal lexeme), the unit on which the modifying or purely classifying function of the adjective takes place. As highlighted by McNally & Boleda (2004:180), the fact that John is a male architect entails that John is male and that he is an architect, whereas if John is a technical architect we cannot infer that John is technical and that he is an architect. In formal semantics a more adequate account can be provided thanks to Siegel (1976)'s definition of predicate modifiers, i.e. properties of properties, rather than properties of individuals. If we conceive the adjectives as properties of another predicated property, we can have a better approximation of the actual semantic composition of phrasal BPs. Let us consider the above mentioned examples:

74

Mario un sociologo criminale Mario uno sciatore alpino Mario un flautista dolce

T(sociol. crim.) = x[((sociol.)criminale)(x)] T(sciatore alpino) = x[((sciatore)alpino)(x)] T(flautista dolce) = x[((flautista)dolce)(x)]

The predicate modifier representation sheds light on the subordination of the adjective and on the properties of its base, namely a phrasal lexeme in which the adjective is combined with a noun. With respect to N + PP phrasal BPs, standard syntax strongly deviates from the actual lexical reading of the construction. The already mentioned pcheur sous la glace-like class investigated by Kerleroux (2007), for instance, is not based on a locative relation between PPs and complex nouns, since the former are part of the lexical unit which functioned as base. The nouns embedded in PPs have no referential force having become part of a fixed expression, i.e. a phrasal lexeme, as clearly stated in Masini (2009:264): The loss of referential force of incorporated elements indicates that we do not have to do with a true phrase. Therefore, incorporated elements do not introduce new referents, but have the function of creating a new reference in combination with other elements. Let us consider for instance locative PPs: in pattinatore sul ghiaccio lit. skater on.the ice 'ice skater' the PP sul ghiaccio does not have any autonomous locative meaning in the phrasal BP since it is part of the lexical meaning of the phrasal lexeme on which the derivation is based. As we will see later, this is reflected in the combinatorial properties of these units in context. The semantics of N+PP nominals is the object of Johnston & Busa (1996). The meaning composition of such complex constructions has been longly debated and no absolute generalisation can be proposed regarding the individual contribution of prepositions and/or embedded nouns, i.e. the way in which meaning is built combining two nouns connected by a preposition. Adopting Pustejovskys (1995) Generative Lexicon (GL) model, however, they sketch a qualia-based account of how head and modifier nouns can compose in different Italian phrase schemata. Johnston & Busa (1996:3) highlight, in fact, that In English, the form of the semantic relation that holds between the modifying noun and the head noun is unspecified and left implicit, while in the

75

corresponding Italian complex nominals, it is partially specified by the preposition which introduces the modifying noun. In the GL, qualia are a representational tool for expressing the componential aspect of word meaning, each expressing a different facet:
FORMAL:

The formal quale distinguishes an individual within a larger set CONSTITUTIVE: The constitutive quale specifies the internal constitution of an entity TELIC: The telic quale consists of the typical function of the entity AGENTIVE: The agentive quale expresses the origin of the entity Johnston & Busa (1996) argue for instance that we can account for the meaning of 'N da N' Italian constructions on the basis of an operation of "qualia modification'. According to the authors this construction's preferred interpretation is that in which the modifier noun relates to the purpose of the head noun, encoded in the telic quale, representing the inherent typical purpose of that lexical item, e.g. to cut things for a knife. Let us consider a small sample of Italian 'N da N' phrasal lexemes: bicchiere da vino 'wine glass' polvere da sparo 'gun powder' fucile da caccia 'hunting rifle' In these examples the meaning of the complex expression can be derived by postulating that the modifier noun specifies an individual argument of the head noun's telic quale, e.g. vino 'wine' specifies what the bicchiere da vino typically contains. Analogous qualia operation are postulated for 'N di N' and 'N a N' constructions, i.e. an 'Agentive Modification' and a 'Constitutive Modification', that is to say operations through which arguments of either the agentive ('N di N', e.g. succo di limone 'lemon juice') or the constitutive ('N a N', e.g. porta a vetri 'glass door') qualia are specified by the modifier noun. Johnston & Busa (1996) thus provide a semantic characterization of three productive phrasal lexeme patterns, which seem to capture the most common semantic properties of these N Prep N expressions. Although some inconsistencies may be found (it seems to me that a productive class of 'N a N' constructions has been ignored, i.e. that containing mulino a vento 'windmill', treno a vapore, lampada ad olio 'oil lamp', in which I argue that the modifier specifies the agentive subquale of the head's telic quale),

76

Johnston & Busa (1996)'s approach seems to be succesful in handling the semantics of a wide range of constructions. See below (taken from Johnston & Busa 1996:7) an instantiation of the constitutive qualia operation (the modifier glass specifies the constitutive quale of the head noun door):

If we rely on this semantic representation, it is apparent how a wide range of phrasal BPs cannot be interpreted according to head-modifier relation usually introduced by the corresponding preposition. Let us consider some examples from the constitutive qualia schema: il padre calciatore a cinque affermato con un palmares
[http://www.calcioa5anteprima.com/lazio/]

David Drudis, strumentista a corde di Barcellona.


[http://www.musicclub.it/musicclub/]

il famoso corridore a ostacoli di origine Cinese


[http://www.motoclub-tingavert.it/t353872s60s.html]

Such expressions cannot be understood in terms of relations between modifier and head nouns, i.e. there is no adequate qualia operation which would allow an appropriate interpretation of the complex nominals on the basis of the conceptual interplay between them, and the standard constitutive qualia operation is excluded: in no way a calciatore a cinque lit. soccer.player to five 'futsal player' can be analysed as a player 'formed by five (players)'; the interpretation of strumentista a corde 'string instrumentist' and corridore a ostacoli 'obstacle runner' poses analogous challenges. Saying that no argumental relation could be available between the head and the modifier noun would be a very strong claim; rather, it is essential to highlight how the phrasal lexemes on which such phrasal BPs were derived perfectly fit into Johnston &

77

Busa (1996)'s generalisations on 'N a N' constructions. Unlike English, which displays totally underspecified NN nominals, Italian allows to trace phrasal BPs back to their phrasal lexeme bases. Once again, the formal and semantic properties of the output provide clues about the derivational path followed by the construction.

2.6 Differences between NA and NPP constructions


Although having a common origin, NA and NPP phrasal BPs display differences in their syntactic behaviour. It seems that, even if both are rooted in a phrasal lexeme, the NPP pattern leads to a higher level of internal cohesion. If we apply standard syntactic tests, we can see that in both constructions constituents are not individually modifiable: *chitarrista esperto elettrico lit. guitarist expert electric *hockeista giovane su ghiaccio lit. hockeist strong on.the ice Such constructions can be modified as a whole, i.e. externally (with an adjective preceding or following the expression, e.g. chitarrista elettrico esperto 'expert electric guitarist', giovane hockeista su ghiaccio 'young ice hockeist'). In order to investigate whether exceptions to this preference existed, specific Google searches were performed to find instances of NA or NPP BPs internally "interrupted" by adjectives. The search yielded significant results for the NA pattern, but none for NPP. Let us consider the examples below: Davig Gogo, chitarrista blues elettrico canadese [http://www.ird.it/oldnews.htm] dei massimi chitarristi italiani acustici: l'abruzzese [http://www.tricesimohomepage.it/] It seems that NA constructions can be arranged more freely. Apart from occasional internally modified ones, a significant number of occurrences were found in which constituents were separated. In particular, factorized expressions were found in which the noun head was omitted in order to avoid its repetition: un chitarrista classico, acustico ed elettrico, frequenta [http://www.kisskiss.it/] hanno aderito filologi classici, romanzi, moderni [http://prin2008.miur.it/] degli scienziati naturali, sociali, economici e politici. [http://www.oilcrash.com/] The difference between the two constructions probably lies in the nature of the internal modifier. Unlike adjectives inside phrasal lexemes, which can be reanalysed

78

as relational modifiers of the derived forms, PPs cannot have but a locative/temporal function outside of the phrasal lexeme in which they were originated. In order to better understand such difference, an informal survey was submitted to 20 native speakers of Italian (mean age: 26, ranging from secondary school to university education), asking them to give judgments (Yes = acceptable sentence, No = unacceptable sentence) about sentences containing NA and NPP agent nouns with separated constituents (either A/PP modifiers were employed as copular (dislocated) predicates or agent nouns were factorized). Two set of sentences were built, each one containing either the NPP or the NA version of each sentence, so that none of the participants could judge two versions of the same sentence. The table below summarizes the results:
Sentence 1a. Quel chitarrista elettrico. 1b. Quell'hockeista su ghiaccio. 2a. nautico quello sciatore. 2b. di fondo quello sciatore. 3a. Il flautista che ho visto traverso. 3b. Il flautista che ho visto di pan. 4a. C'erano nuotatori sincronizzati e pinnati. 4b. C'erano sociologi della comunicazione e della letteratura. 5a. Mario un filologo sia romanzo che germanico. 5b. Mario un hockeista sia su prato che su ghiaccio. 6a. Non cognitivo quello scienziato. 6b. Non della materia quel fisico. 7a. un linguista bravissimo computazionale. 7b. un giurista bravissimo del lavoro. Total (NA) 5 2 3 2 2 0 4 3 3 1 2 0 2 0 21 (30%) 8 Yes 3 4 4 5 7 9 4 5 3 7 8 9 8 9 37 No 2 4 3 3 1 1 2 2 4 2 0 1 0 1 12 ?

(52.8%) (17.2%) 48 14 (20%)

Total (NPP)

(11.5%) (68.5%)

79

These acceptability judgments confirm the assumption that although NA and NPP forms are both uninterruptible (both are rejected in more than 50% of the sentences), adjectival modifiers are not perceived as bound to the nominal constituent as PPs in NPP forms (30% accepted vs. 11.5%).

2.7 Survey of phrasal nouns


We will now have a look at a wide range of phrasal nouns involving a bracketing paradox. The data collected come from Google searches and La Repubblica corpus (Baroni et al. 2004). The collection relied both on pattern queries and individual focused searches. During the collection it was clear that some major categories presented a huge number of phrasal BPs, all in the domain of "personal nouns" (Spencer 1988), i.e. denoting actors of specific (often professional) activities. The next sections will illustrate instances belonging to the classes of academic, sport and musical activities. This does not mean that phrasal BPs can be found in these domains only: the following outline has only an illustrative function and it should not be forgotten that they come from disparate domains (see the following examples): azionista ordinario 'owner of common stocks' banconista alimentare 'owner of food desk' cronista rosa 'gossip columnist' nudo proprietario 'holder of a residual life estate' parlamentare europeo 'member of the European Parliament' giocatore di ruolo 'RPG player' 2.7.1 Academic Many instances of phrasal BPs were found in the field of jobs and activities related to the academic world. The complex constructions involved often denote specialists in academic subdisciplines named by a phrasal lexeme. Many academic subdisciplines are represented in Italian by a phrasal lexeme, comprehending a noun (the academic discipline, e.g. economia, chimica, linguistica, sociologia) followed by either an adjective (e.g. politica, cognitiva, organica) or a PP (e.g. della letteratura, della comunicazione). The nouns to name professionals in these academic sectors are mostly formed by the agentive form derived from the noun (e.g. economista, chimico, linguista, sociologo)

80

followed by the same modifier of the base phrasal lexeme (in case of adjectives, agreeing in number and gender), e.g. chimico industriale 'industrial chemist' chimico molecolare 'molecular chemist' chimico nucleare 'nuclear chemist' chimico organico 'organic chemist' economista culturale 'cultural economist' economista finanziario 'financial economist' economista politico 'political economist' filologo classico 'classical philologist' filologo germanico 'germanic philologist' filologo moderno 'modern philologist' filologo romanzo 'romance philologist' filosofo del diritto 'philosopher of law' filosofo del linguaggio 'philosopher of language' filosofo della mente 'philosopher of mind' filosofo etico 'ethic philosopher' filosofo morale 'moral philosopher' filosofo politico 'political philosopher' filosofo teoretico 'theoretical philosopher' fisico della materia 'matter physicist' fisico molecolare 'molecular physicist' fisico nucleare 'nuclear physicist' fisico tecnico 'technical physicist' fisico teorico 'theoretical physicist' giurista civile 'civil lawyer' giurista del lavoro 'work lawyer' giurista europeo 'european lawyer' giurista penale 'criminal lawyer' linguista applicato 'applied linguist' linguista cognitivo 'cognitive linguist' linguista computazionale 'computational linguist' linguista dei corpora 'corpus linguist' matematico applicato 'applied mathematician' matematico teorico 'theoretical mathematician' psicologo cognitivo 'cognitive psychologist'

81

psicologo del lavoro 'work psychologist' psicologo sociale 'social psychologist' sociologo clinico 'clinical sociologist' sociologo del diritto 'sociologist of law' sociologo della comunicazione 'sociologist of communication' sociologo della letteratura 'sociologist of literature' sociologo politico 'political sociologist' scienziato applicato 'applied scientist' scienziato cognitivo 'cognitive scientist' scienziato giuridico 'law scientist' scienziato informatico 'computer scientist' scienziato politico 'political scientist' scienziato sociale 'social scientist' 2.7.2 Sport Many sport names in Italian are phrasal lexemes, e.g. hockey su ghiaccio 'ice hockey', lotta svizzera 'swiss wrestling', calcio a cinque 'futsal'. The professionals involved in such sports are named drawing on the agentive form derived from the head noun, e.g. hockeista, lottatore, calciatore combined with the base modifiers. Here some forms encountered: arrampicatore sportivo 'sport climber' artista marziale 'martial artist' calciatore a cinque 'futsal player' calciatore gaelico 'gaelic football player' cestista in carrozzina 'wheelchair basket player' ciclista su pista 'track cyclist' ciclista su strada 'street cyclist' corridore ad ostacoli 'obstacle runner' corridore campestre 'cross-country runner' corridore di resistenza 'resistance runner' footballista americano 'american football player' ginnasta acrobatico 'acrobatic gymnast' ginnasta aerobico 'aerobic gymnast' ginnasta artistico 'artistic gymnast' ginnasta attrezzista 'trampolining, tumbling gymnast'

82

ginnasta ritmico 'rythmic gymnast' golfista su pista 'minigolf player' hockeista in-line 'in-line hockey player' hockeista su ghiaccio 'ice hockey player' hockeista su pista 'minihockey player' hockeista su prato 'field hockey player (grass)' lottatore greco-romano 'greco-roman wrestler' lottatore svizzero 'swiss wrestler' nuotatore pinnato 'finswimmer' nuotatore sincronizzato 'synchronized swimmer' pattinatore artistico 'artistic skater' pattinatore corsa 'speed skater' pattinatore di figura 'figure skater' pattinatore sincronizzato 'synchronized skater' pattinatore su ghiaccio 'ice skater' pattinatore su rotelle 'roller skater' pescatore a mosca 'fly fisherman' pescatore in apnea 'apnea fisherman' pescatore sportivo 'sport fisherman' saltatore a ostacoli 'obstacle jumper' saltatore in alto 'high jumper' saltatore in lungo 'long jumper' sciatore alpino 'alpine skier' sciatore di fondo 'cross-country skier' sciatore nautico 'nautic skier' sciatore nordico 'nordic skier' slittinista su pista artificiale 'sledder on artificial track' tiratore alla fune 'tug-of-war competitor' tiratore al piattello 'trap shooter' tiratore a volo 'clay shooter' tiratore con l'arco 'archer' 2.7.3 Music

Musical instrument players represent a productive domain with respect to phrasal BPs. In Italian, as well as in many other languages, a number of musical instruments

83

(expecially subtypes) are named through NA or NPP constructions, resulting on the paradoxical forms made up of the agentive noun followed by the base modifiers. bassista acustico 'acoustic bass player' bassista elettrico 'electric bass player' batterista acustico 'acoustic batterist' batterista elettrico 'electric batterist' batterista elettronico 'electronic batterist' chitarrista acustico 'acoustic guitarist' chitarrista a dodici corde 'twelve-string guitarist' chitarrista a sei corde 'six-string guitarist' chitarrista classico 'classic guitarist' chitarrista elettrico 'electric guitarist' contrabbassista elettrico 'electric double bass player' cornista inglese 'English horn player' flautista basso 'bass flutist' flautista di pan 'pan flutist' flautista diritto 'recorder player' flautista dolce 'recorder player' flautista traverso 'flutist' oboista d'amore 'oboist d'amour' organista elettrico 'electric organist' pianista elettrico 'electric pianist' pianista elettronico 'electronic pianist' sassofonista soprano 'soprano saxophonist' strumentista a corde 'string instrument player' strumentista a fiato 'wind instrument player' tastierista elettronico 'electronic keyboard player' violinista elettrico 'electric violinist' violoncellista barocco 'baroque cello player' violoncellista elettrico 'electric cello player' 2.7.4 Locative/collective nouns Apart from the class of BPs which Spencer (1988) defines "personal nouns", there seems to exist a productive class of nouns with locative/collective meanings connected with specific referents. In Italian the suffix -eria, deriving nouns from nouns, seems to

84

be the most productive with respect to phrasal BPs. Web searches yielded examples like the following: birreria alla spina 'draft brewery' camiceria da donna 'woman shirt shop' chioderia da falegname 'carpenter nail shop' gioielleria da viaggio 'travel jewelry' occhialeria da vista 'glasses collection' orologeria da polso 'wristwatch collection' pescheria d'acqua dolce 'freshwater fish market' pizzeria al taglio 'sliced pizzeria' pizzeria da asporto 'takeaway pizzeria' segheria elettrica 'electric saw shop' utensileria da cucina 'cooking tools collection' valigeria da uomo 'man suitcase shop' verniceria a polvere 'powder paint shop'

2.8 Adjectival phrasal BPs


Are phrasal bracketing paradoxes limited to the domain of derived nouns? The issue of bracketing paradoxes in phrasal adjectives has to my knowledge never been raised. Web data, however, attest many instances of relational adjectives derived from phrasal lexemes, see e.g. dallaggressione ondosa marina fin dagli [http://www.autoritabacino.marche.it/] costituzione di un tribunale ecclesiastico romano [http://www.vatican.va/news_services/] Direttiva parlamentare europea [http://issuu.com/adpware/docs/mc108] frequentare gli istituti scolastici dell'obbligo ubicati [http://www.comune.lagnasco.cn.it/] Similarly to nominal counterparts, phrase-based adjectives pose bracketing challenges because the adjectival suffix semantically scopes over the whole phrase although being attached to the noun only, e.g. atmosferico terrestre lit. atmospheric terrestrial 'related to Earth's atmosphere', chirurgico estetico lit. surgical aesthetic 'related to aesthetic surgery', ciclistico su pista lit. cyclistic on track 'related to track cyclism'. An analysis will be provided of Italian phrase-based adjectives with respect to their morphosyntactic and semantic properties, as well as their usage and coinage properties.

85

2.8.1 Extension The existence of adjectival phrasal BPs (of the A+A type) could be rejected on the assumption that the sequence of adjectives should be treated as coordinate, so that the two following NPs should receive the same semantic account: i) dell'inquinamento atmosferico italiano proviene La politica energetica italiana non pu contro l'inquinamento atmosferico terrestre La crisi energetica solare fu risolta nel 1939 [http://www.rinnovabili.it/] [La Repubblica corpus] [La Repubblica corpus] [La Repubblica corpus]

ii)

Similar arguments to those valid for phrasal nouns can be put forward with respect to semantic composition. The NPs in (i) can be defined through the intersection of the extensions of the two adjective and the noun, so that the three constituents are combined in a [[ [ ]]] = [[[]] [[] [[]]] way, and the following formalisms can be proposed: T(atmosferico italiano) = x [atmosferico(x) italiano(x)] T(energetica italiana) = x [energetico(x) italiano(x)] On the contrary, the semantics of (ii) constructions cannot be based on the extensional intersection of its constituents: T(atmosferico terrestre) x [atmosferico(x) terrestre (x)] T(energetica solare) x [energetico(x) solare(x)] In this case too, Siegel (1976)'s predicate modifiers better suits the needs of our semantic analysis. Adjectival phrasal BPs are rooted in phrasal lexemes (the same can derive both nouns and adjectives, e.g. chitarra elettrica > chitarrista elettrico, chitarristico elettrico). The second adjective specifies information related to the other (noun-based) adjective. T(energetico solare) = x[((energetico)solare)(x)] T(atmosferico terrestre) = x[((atmosferico)terrestre)(x)]

86

With phrase-based adjectives too, however, it is difficult to provide a semantic account based on individual constituents. Rather, the meaning of adjectival phrasal BPs is rooted in that of the phrasal lexeme which functioned as base for the derivation. 2.8.2 Hyphenation During the perusal of web data, a number of instances of phrase-based adjectives were found which were hyphenated, i.e. an hyphen was introduced to mark that the two adjectives are actually forming a complex unit with a unitary meaning. Al via l'Itinerario turistico-religioso a cavallo Pietrelcina [http://www.sanniotradizioni.it/] proposta di Piano faunistico-venatorio regionale [http://www.regione.piemonte.it/] di stampo giornalistico-scandalistico, le prime [http://www.altremappe.org/] questione chitarristica-elettrica negli arrangiamenti [http://www.soundsblog.it/] Il percorso terapeutico-riabilitativo sempre [http://www.ulss7.it/] in piena tempesta bellica-civile.[http://www.linamangionesavatteri.it/] It seems reasonable to assume that language users occasionally mark graphically the link between the adjectival constituents because unlike other adjectival sequences, a unitary meaning is perceived in these constructions. This, however, holds also for NA phrasal BPs, but no instance of hyphenated forms was found in that domain. Hyphenated forms probably reflect a tendence towards univerbation: unlike NA phrasal BPs they display a formal structure in which both constituents share the same syntactic category. This may encourage univerbation, since syntactic marking could be perceived as redundant: another related piece of evidence comes from another phenomenon encountered during the analysis of web data, i.e. factorization. 2.8.3 Factorization Two adjectives can be factorized in Italian if they are adiacent and share an inflectional mark, e.g. un'analisi storico-filosofica degli approcci [http://www.lawrence.altervista.org/] Storico-filosofica 'historico-philosophical' displays only one inflectional mark (feminine singular on filosofica) standing for both adjectival forms. The uninflected word is in its unmarked form. Among phrase-based adjectives a large number of encountered forms were factorized. See for instance

87

la cultura pokeristico-sportiva e poi [http://forum.assopoker.com/] l'analisi energetico-solare diventa alla portata [http://spa.casaccia.enea.it/] Cresce l'offerta turistico-religiosa nella Capitale [http://www.newsfood.com/] La comunit terapeutico-riabilitativa di Olgiasca [http://www.gabbianoonlus.it/] pochi gruppi chitarristico elettrici che riesco [http://www.ilmucchio.net/] delle indagini archeologico-medievali territoriali su [http://prin2007.miur.it/] Such forms display a different structure from the already mentioned storicofilosofica 'historico-philosophical', because the latter has a coordinative function, i.e. storico-filosofica means 'historical and philosophical'. We can conclude that both hyphenation and factorization contribute to defining some properties of phrase-based adjectives: first, they are likely to be tendentially perceived as regular as other sequences of (coordinated) adjectives, as hyphenation shows. In some cases, however, hyphenation could be regarded as a strategy for marking in the written form the clustering of units in a sequence of more than two adjectives, see e.g. the already reported proposta di Piano faunistico-venatorio regionale [http://www.regione.piemonte.it/] delle indagini archeologico-medievali territoriali su [http://prin2007.miur.it/] Factorization, on the other hand, may display a certain degree of lexicalization of adjectival expressions, favoured by bases (phrasal lexemes) provided with a fixed, unitary meaning. 2.8.4 APP costructions Adjectival phrasal BPs are not limited to AA constructions. As already mentioned, NPP phrasal lexemes can in fact derive adjectival expressions too. A significant source of these expressions seems the field of anthroponimic names, see e.g. La coreana del nord Kye Sun Hui [http://www.fijlkam.toscana.it/] La bandiera cipriota del nord [http://www.agenziadogane.it/] la maggioranza del popolo osseto del sud [http://it.euronews.net/] As already reported, nouns embedded in these PPs generally lose their referential force. This probably happens at various levels: let us consider the alternation between prepositions with/without embedded determiner, e.g. sul/su 'on.the:M:Sing/on'. Hockey

88

su/sul ghiaccio lit. hockey on.the:M:Sing/on ice 'ice hockey', for instance, should in principle derive both a phrasal noun and a phrasal adjective with alternation, namely hockeista su/sul ghiaccio 'ice hockeist' and hockeistico su/sul ghiaccio 'related to ice hockey'. As a matter of fact, however, web searches for this case and a small sample of similar constructions yielded limited occurrences of embedded determiners: although the form hockey sul ghiaccio has a significant presence in on-line texts, namely 18,300 hits (846,000 for hockey su ghiaccio), determiners were present in less than 0,01% of the corresponding agent nouns and in none among phrasal adjectives. hockeista sul ghiaccio 19 hockeista su ghiaccio 270,000 hockeistico/i/a/che sul ghiaccio 0 hockeistico/i/a/che su ghiaccio 3 It is reasonable to assume that a progressive loss of referential force is at play, ranging from phrasal lexemes, where "N+PREP+DET+N constructions are [...] less systematic and frequent than the N+PREP+N" (Masini 2009:261), to phrasal nouns and phrasal adjectives. It could be hypothesized that in the adjectival forms the loss of referential force of the embedded constituents is even stronger because the link between the adjective and the constituent (e.g. hockeistico with ghiaccio) is looser than that of the complex noun (e.g. hockeista with ghiaccio), in turn looser than that of the simplex noun in the original phrasal lexeme (e.g. hockey with ghiaccio).

2.9 Pragmatic constraints


Some remarks are needed with respect to the potential constraints on the formation of phrasal BPs. It was apparent from the data collected that no proper semantic constraint could be found: there does not seem to be any kind of restriction either on the base or on the output regarding meaning. Some forms generating potential ambiguity (e.g. phrasal lexemes containing an adjective combinable with [+human] bases) are nonetheless attested, e.g. lottatore svizzero 'swiss wrestler', cornista inglese 'English horn player' etc. Although not every phrasal lexeme can derive phrasal BPs, this seems rather to be ascribed to pragmatic factors rather than semantic. The coinage of phrasal BPs is subject to pragmatic word formation constraints in the sense that there must be a need (a "pragmatic pressure" or "real-world need", Lieber 2004:96) for a lexical item that

89

would otherwise be redundant for language use. According to Plag (1999:39) this need is "a reflection of their capacity of labeling a new concept or referent". Lipka (1977) introduces the concept of Hypostasierung ('hypostatization'), stating that the existence of a certain word suggests the existence of the denoted entity: "the object of the label needs to exist, so only existing things can be named" (Lipka 1977:161). No attestations of phrasal BP was encountered with phrasal lexemes denoting referents with limited extension: birra al malto 'ale beer' denotes a kind of beer as birra alla spina 'draft beer' does; although the noun birra 'beer' present in both phrasal lexemes has a paradigmatic relation with the complex noun birreria 'beerhouse', birreria alla spina 'draft beer-house' is attested whereas birreria al malto 'ale beer-house' is not. This is motivated by the non-existence of a commercial activity exclusively devoted to ale beer and its scarce pragmatic plausibility. As (Plag 1999:40) puts it, "one can only label something that does exist". If there does not exist a significant number of entities to refer to no coinage occurs. This is why, on the same line, a range of potential constructions can (at the present state of affairs) be discarded, see e.g. ?pizzeria ai funghi 'mushroom pizzeria' ?morfologo compositivo 'compounding morphologist' ?analista delle urine 'urine analyst' ?calciatore a otto 'eight-soccer player' All the meanings expressed above would probably be realised through the just madeup forms if significant reasons existed for their coinage (and maybe will: let us imagine that calcio a otto 'eight-soccer' catches on and a larger and larger set of individuals playing that sport needs to be referred to).

2.10 A constructionist account


The kind of paradoxes illustrated above can be accounted for by assuming a Construction Grammar approach to word formation. Goldberg (1995:4) asserts that phrasal patterns are considered constructions if something about their form or meaning is not strictly predictable from the properties of their component parts or from other constructions. This definition captures the main property of phrasal BPs: their meaning cannot be directly derived from their component parts. If we consider the NA pattern only,

90

the standard construction combining a noun and an adjective cannot itself adequately represent the meaning of the output:

[[N]N [A]A]cN
<N with a relation S to A>

The pattern above can be employed to represent combinations like chitarrista esperto 'expert guitarist', but fails to predict the meaning of chitarrista elettrico. Bisetto & Moschin (forthcoming)'s costructionist account of phrasal nouns assume that the suffix is attached to the nominal constituent of the complex NA base construction because it selects nouns, and that this selection is justified by the consistency of the semantics of both the N and the suffix. Let us consider their proposal for the example flautista barocco 'baroque flutist': [[[flauto]N [barocco]A]cN [-ista]]dcN
Formal constraint: the suffix attaches to the first constituent Semantic constraint: the suffix has semantic scope over the cN 0

Can such forms be accounted in terms of pure suffixation? Some remarks are needed on this issue. The paradox itself is caused by the position of the suffix, mismatching with its semantic scope. Assuming that suffixation can occur on phrasal lexemes, generating phrasal BPs, can however lead to overgeneralize that this derivational process can occur regardless of any paradigmatic constraint. It is apparent from a number of phrasal BP forms pertaining to both nominal and adjectival domains that no phrasal BP is generated in the absence of an already existing derived complex word. Let us consider the following examples: (i) sci nautico 'nautic sky' pizza al taglio 'sliced pizza' pianta grassa 'succulent plant' carne alla brace 'grilled meat'

(ii)

The heads of the forms in (i) have already established paradigmatic relations with a derived noun, e.g. sciatore and pizzeria, whereas those in (ii) do not. The phrasal BPs built on the latter sound unacceptable.

91

sciatore nautico 'nautic skier' pizzeria al taglio 'sliced pizzeria' *piantista grasso *carnista alla brace Bisetto & Moschin (forthcoming)'s representation makes no claim about the paradigmatic dimension of the process. It seems useful to recall Spencer (1988)'s remarks on personal nouns. He points out that the availability of the complex lexical unit is crucial for the proper functioning of the fourth-proportional schema which licenses phrasal BPs. As we have pointed out earlier, the major weakness of the analogical pattern lies in the morphological properties of the derivational output (namely number and gender agreement in case of NA constructions), not considered even by Bisetto & Moschin (forthcoming). The construction addressed here displays features belonging to two different constructions: [[N]N[A]A]NP : phrasal BPs have their constituents agreeing in gender and number if the second constituent is an adjective; [[N]N suff]X : the derivation operates on a unitary concept; In order to deal with these aspects, I will illustrate a proposal of constructionist schema aiming at representing both nominal and adjectical instances of phrasal BPs, assuming as a prototype those with an NA base. The pattern specifies some conditions on the derivation of phrasal BPs: the derived form in the output (i.e. [[N]suff]X e.g. chitarristaN, energeticoA must be already present in the lexicon) and the complex form derived from the noun must agree with the adjective. The construction is partially underspecified with respect to lexical category (X), since according to the availability of forms we can have from the same NA base nouns only, adjectives only, or both: N,*A pattinaggio artistico *N, A energia solare N, A chitarra elettrica > pattinatore artistico, *pattinaggistico artistico > *energista solare, energetico solare > chitarrista elettrico, chitarristico elettrico

Here the construction for XA BPs (e.g. chitarrista elettricoN, energetico solareA):

92

[[N{AGR}]N [A{AGR}]A]N
<naming a concept >

>

[[N]suff{AGR}]X [A{AGR}]A]X
<+suff>

| [[N]suff]X A less complicated construction can be assumed for XPP bracketing paradoxes (e.g. calciatore a cinqueN, scolastico dell'obbligoA): [[N]N [PP]]N
<naming a concept >

>

[[N]suff]X [PP]]X
<+suff>

| [[N]suff]X

2.11 Processing Issues


The formal properties of phrasal BPs raise interesting questions concerning both language production and comprehension. The mismatch between formal and semantic structure is not only a matter of linguistic representation, and arguably reflects itself on the strategies employed by language users in order to correctly interpret the input. A range of theoretical assumptions concerning language processing can be called into question if we focus on the way phrasal BPs are interpreted in everyday language. It should be clear from the above discussion that they represent a productive class and thus cannot be conceived as special status items: rather, their interpretation could in no way differ from standard lexical and/or phrasal units. The first issue to consider is related to their hybrid status between lexemes (from a semantic viewpoint) and phrases (from a formal viewpoint). Speakers must be able to access a unitary meaning instantiated in a phrasal structure. With respect to this, phrasal BPs conceivably pose a number of challenges to many models of language comprehension: first, the mentioned dissociation between unitarity in meaning and separation in form; second, the anomalies this can cause at the level of part-of-speech sequences in utterances containing phrasal BPs: head nouns in NA phrasal BPs, followed by two adjectives, must be matched with the actual modifier; phrasal adjectives, on the other hand, can produce sequences of three or more adjectives (e.g. un futuro energeticoA1 solareA2 brillanteA3 'a brilliant solar energetic future'), normally ruled out. According to a leading school of thought in psycholinguistics, parsing (i.e. the analysis of syntactic structures) precedes lexical access. Frazier & Clifton (1996) assume that the syntactic structure of an utterance is computed on the basis of part-of-speech

93

recognition, maintaining all the structural alternatives in parallel, before the meanings of the individual lexical items are accessed. Future research could evaluate (for instance through eye-tracking studies on self-paced reading) whether phrasal BPs cause processing delays. Their natural occurrence in everyday spoken and written language, however, suggests that they are processed in a normal fashion and encourages those theoretical approaches claiming that parsing is lexically-driven (see e.g. MacDonald 1993), i.e. it is based on the access to lexical information, which progressively constrains the syntactic analysis which the same sequence items can produce according to their parts-of-speech. With respect to the comprehension level, i.e. the level at which speakers integrate the meaning of words/utterances into the broader discourse context, two major theoretical trends can be evaluated. The modular approach to language comprehension (see e.g. Fodor 1983) assumes that it is the outcome of the functioning of different cognitive modules, each devoted to a specific aspect of comprehension. On the contrary, MarslenWilson & Tyler (1987:51) defend the view that our processing system is highly flexible, even opportunistic, in its use of different types of processing information to achieve the perceptual goal of relating an utterance to its discourse context. [...] The flexibility in using different sources of constraints, as and when they are available, means that the process of discourse linkage is not dependent on information being made available to it in a fixed order or in a fixed format. If we look at the comprehension of phrasal BPs, challenges are posed apart from syntactic parsing, i.e. speakers must be able to on-line discern between a standard compositional N+A/PP combination and a phrasal BP. In order to achieve this, speakers have to access the meaning of the phrasal lexemes functioning as bases: unlike standard derived words in which a stepwise affixation can be stipulated, phrasal BPs do not display the morphological cues necessary to infer the hierarchical relationships to derive the final meaning. This supports the hypothesis that language comprehension draws on a wide number of sources in order to be able to establish links between linguistic items and discourse referents.

2.12 Novel phrasal BPs


Do phrasal BPs belong to a productive pattern in Italian? Does comprehension rely on already established phrasal lexemes? Are speakers able to map novel phrasal BP

94

expressions onto their correct semantic representation? In order to ascertain this, a test was performed aiming at evaluating whether and to what extent Italian speakers are able to understand novel phrasal BPs (nouns). A set of 5 novel phrasal BPs was created to be tested for comprehension among 20 native speakers (mean age: 31, ranging from secondary school to postgraduate education). The expressions created had to meet the following requirements: (i) be unattested in the web; (ii) conform to either the NA or the NPP "paradoxical" pattern; (iii) be plausible, both from a pragmatic and a semantic viewpoint. Adjectival expressions were put into sentences. Participants were asked to paraphrase the meaning of the phrasal BP. One would predict that, if the template can be retrieved by speakers, both phrasal BPs built on attested phrasal lexemes (terapeuta aerobico 'aerobic therapist') and on unattested ones (hockeista sull'acqua 'water hockeist') should be comprehended. Below the results:
Novel phrasal BPs terapeuta aerobico medico agopunturale hockeista sull'acqua orologiaio a cuc calciatore a tre flautista a tasti sociolinguista cognitivo astrologo lineare tuffatore sincronizzato strumentista ad acqua Paraphrase "che pratica/ esperto in/ terapia aerobica" (75%) "specializzato in/dottore in/ medicina agopunturale" (60%) "giocatore di/che gioca a/pratica hockey sull'acqua" (85%) "che vende/venditore di/ esperto in/ orologi(o) a cuc" (85%) "giocatore di calcio a tre" (80%) "suonatore di/ esperto in/ flauto a tasti" (75%) "esperto di/ professore di/ studioso di/ sociolinguistica cognitiva" (60%) "esperto in/praticante dell'/ astrologia lineare" (70%) "che pratica (come sport) i(l) tuffo/i sincronizzato/i" (80%) "suonatore di (uno) strumento/i ad acqua" (70%)

2.13 Conclusions
The previous sections have outlined an analysis of the phenomenon of bracketing paradoxes in the domain of phrasal constructions. The focus on Italian, at my knowledge unprecedented, has allowed to disentangle the complex interplay of formal and semantic properties resulting in phrasal BPs, unlike the long-standing English-based generativetransformational tradition which has both underestimated the complexity of the phenomenon ignoring its cross-linguistic manifestations, and exclusively focussed on

95

the attempt to "repair" a malfunctioning morphosyntactic representation without insights into its possible explanations from a linguistic viewpoint. The present chapter has collocated phrasal BPs at the heart of the boundary between syntax and the lexicon: the functional equivalence between phrasal lexemes and "standard" words leads to a situation in which a formally syntactic object is derived due to its word-like meaning: both for XA and XPP constructions, the result in languages with NA basic order is impossible to be brought back to a proper bracketing. A semantic analysis of phrasal BPs revealed their divergence with formally parallel syntactic phrases, and it was apparent from different perspectives that the central meaning to be considered is that provided by the derivational root of the construction, namely the phrasal lexeme functioning as base of the complex BP. No space to my knowledge ever had ever been devoted to adjectival bracketing paradoxes. The chapter documented the attestation of adjectival phrasal BPs and their functional similarity with nominal constructions, so that it was possible to describe a constructionist pattern on which both are based. Their morphosyntactic properties have shown how such forms trigger phenomena like hyphenation and factorization, which reflects the fact speakers tend to mark graphically their being unitary from a semantic viewpoint and start a process of univerbation. Phrasal BPs do not seem to pose particular challenges related to productivity: ambiguity in adjective-noun matching does not prevent many expressions from being coined, and the only criterion valid to discard potential constructions is pragmatic plausibility. Comprehension data elicited from native speakers allowed to establish that novel instantiations of the described constructionist template were easily integrated. In addition to this, phrasal BPs encourage the view that speakers normally follow flexible cognitive paths in order to be able to map such unconventional morphosyntactic structures in which no stepwise derivation and standard bracketing can be assumed, and relate them to the right discourse referent. Bracketing paradoxes in the domain of phrase-like constructions can thus be conceived as a natural state of affairs in Italian. They represent a productive word formation pattern, rooted in the functional equivalence between morphological and phrasal lexemes.

96

Final remarks

The present work has contributed to the analysis of the bracketing paradox phenomenon. It has provided evidence on their motivation rather than trying to solve a theoretical puzzle. BPs are paradoxical only if a rigid theoretical and notational approach is applied: they represent a class (variegated inside) of word formation patterns in which a standard bracketing cannot be assigned for a range of functional reasons. We can conclude that bracketing paradoxes are not paradoxes (in the sense of something exhibiting inexplicable or contradictory aspects) at all. They rather represent instances of incompatibility between theoretical descriptive assumptions and language form. In addition to this, they speak in favour of a meaning-centered language comprehension model in which speakers are able to integrate expressions although their imperfect match between meaning and structure. In order to ascertain this empirically, future research in psycho-/neurolinguistics could evaluate whether different indexes of cognitive processes are elicited -for instance- by phrasal BPs vs standard phrases in comprehension tasks. A quote attributed to the physicist Niels Bohr says "How wonderful that we have met with a paradox. Now we have some hope of making progress" (quoted in Moore 1967). The problems posed by bracketing paradoxes were thus useful to highlight general and language-specific properties and be more aware about the complex interplay of traditionally separated theoretical domains which produced this result.

97

S-ar putea să vă placă și