syntax and phonology

7/29/2019 Syntax and Phonology

1/6

'5/hjgltfattltia 0/ Aa5W7i- /phfjf/We tives/its.Syntagmatic LexicalRelations tthere is a paradigmatic class of words that enters into thesyntagmatic relation: all words referring to liquids (coffee.tea. wine, etc.). not to mention metonymically used termsfor containers. such as a glass. a cup, and the like. Becausethe class in question can be characterized semantically ('allwords naming potables or recipients containing them`). thesyntagmatic relationship itself can be characterized moreeasily in conceptual ten'ns than by an enumeration of thelexical items involved; the syntagmatic link between to nodand head. on the other hand. is easy to describe by meansof a lexical enumeration. (SeeCoseriu 1967 for a classifica-tion of syntagmatic relations taking into account thisdistinction.)2. Restrictive or Expectational RelationsKatz and Fodor`s selectional restrictions (l963) wereintended to block (in the algorithmic framework of agenerative grammar) the production of sentences violatingexisting syntagrnatic relationships. Thus, a selection restric-tion to the effect that to hit in the reading to strike with ablow requires a constituent with the feature Physical Objectas its direct object ensures that in theman hit the colorfulball, a reading of ball as spherical object is accepted,whereas the reading festive social assembly for dancingis blocked. However, Weinreich (1966) noted that certainsentences that should be unintelligible according to theKatzand Fodor approach (i.e., that would receive no read-ing) were in fact interpretable, albeit in an unusual way.So, to explain whya sentence such as hewas drinking carrotsdoes receive an interpretation, Weinreich replaced thenotion of selection restriction by that of transfer feature.Instead of saying, for instance, that the verb to drinkrequires a direct objectwith the feature +Liquid. a transferfeature approach would say that the verb imposes such afeature on its direct object. Because such a transferredfeaturewould cancel out features that are incontradictionwith it (such as a feature to the effect that carrots are solid),the sentence could be interpreted as he was drinking carrotjuice.A transfer feature approach does not solve all the prob-lems conceming deviant but interpretable sentences; forinstance, because the interpretative flexibility is not abso-lute, restrictions would have to be formulated with regardto the transfer features capacity to override existingfeatures. That is not, however, the main point to be madehere. The point is this: when syntagmatic semantic relationsare interpreted in the way suggested by Weinreich, theyembody flexible expectations rather than rigid restrictions.(ln this respect, they resemble presuppositions, and someauthors have suggested treating them like that.)3. Motivated or Arbitrary RelationsA Firthian definition of collocation as repeated co-occurrence of words leaves open the possibility that thelexicalpattern in question is not semantically determined inany way. Consider the use of prepositions in the followingsemantically related expressions: you bother about, troubleyourself with, and take care ofsomething. There does notseem to be a readily identifiable distinction between theconcepts expressed by to bother and to trouble that couldexplain why about is the appropriate preposition in one4476

case. and with in the other. Similarly. someone is sate-guardedfrom but protected against something. One con-siders things in connection with but with respect tosomething. Or. to take an example not involving preposi-tions, one can hear something on the radio but never onradio.whereas for some speakers of English at least. it isperfectly normal to have seen something on Tl ' as well ason the TV. In a number of these cases. historical factorslike the bleaching of metaphor probably play a role. and itmight be. of course. that the semantic distinctions involvedare so subtle as to remain unnoticed upon superficialinspection. In general, however. there do seem to be lexicalco-occurrence restrictions that are not synchronicallysemantically motivated. The term collocation is reservedby some authors for such idiosyncratic restrictions on thecombinability ofwords. in contradistinction with the other.conceptually motivated syntagmatic restrictions.See also: Phraseology; Firth and the London School.BibliographyCoseriu E 1967 Lexikalische Solidaritaten. Poetica I: 293-303FirthJ R |957 Papers in Linguistics1934-1951.Oxford UniversityPress, LondonKatzJ J. Fodor J A 1963The structure of a semantic theory. Lg39: |70-210Lipka L 1990 An Outline of English Lext colog_i'. Niemeyer.TiibingenPorzig W 1934Wesenhafte Bedeutungsbeziehungen. Beitriige :u rGeschichte der Deutschen Sprache undLiteratur 58: 70-97Weinreich U 1966 Explorations in semantic theory. In: SebeokT A (ed.) Current Trends in Linguistics. vol. 3. Mouton. TheHague

D.Geeraerts

Syntax and PhonologySyntax, phonology, and the lexicon are the central andindispensable parts of themediation between meaning andsound in language. The lexicon provides (at least) theirreduciblebasisof sound-meaning associations. in lexemesand idioms. Syntax and phonology together concern therule-govemed portion of this association, syntax serving(together with morphology) as the gateway to semantics,phonology as the gateway to phonetics. Both are complexsystems involving units of various sizes (in syntax, theseinclude words, phrases, clauses, and sentences; in phonol-ogy, these include features, segments, syllables, phono-logicalwords, and phonological phrases) and also involving(language-particular) generalizations about how units com-bine to make larger units. How syntax articulates withphonology is then a central question in the description ofany particular language, as well as a major issue in theframingof a general theory of grammar.1. Phenomena and ComponentsThe labels syntax and phonology` are used both forclasses of phenomena and for components of a grammar.In the first sense, syntax concerns the meaningful freeunits (ofword size or larger) in a language and compriseseverything that has to do with the combination of these


2/6

S_tnta.\' and Phonologyinto meaningful composites. and consequently with the dis-tribution of such units with respect to one another.(Morphology` in this sense concems the expression ofmeaning by form within words.) Also in the first sense,phonology concems those properties of sound that are rel-evant in a language and comprises everything that has todo with the combination of these into pronounceable unitsof various sizes.The divisions thus made between syntax and phonologyare not necessarily those appropriate for a theory of gram-mar,which divides components according to differences inthe types of generalizations (also known as rules') appli-cable within them. The distribution of meaningful units ina language can be contingent on their stylistic values (as inthe anomalousconditional clause wereI tobuzzof ifI left,with itsconflict between the formal style of the inversion init and the casual style of the idiom buzz ojj), or on theirsemantics (as in the anomalousclause I am containingDNA,with its conflict between the ongoing-event semantics of theprogressive and the state semanticsoftheVPcontainDNA),or on their morphological properties (aswhen clauses withmodals, like I can sing, fail to occur in a wide variety ofconstructions-*I want to cansing, versus I want to be ableto sing--because the modals lack nonfinite inflectionalforms), or their phonological properties (as in the anomal-ous clause to would make me happy-cf. to do so wouldmake me happy and not to wouldmake me happy-with itsaccentually stranded to).But that does not necessarily mean that rules in the syn-tactic portion of a grammar attend to the stylistic values,semantic properties, morphological properties, and phono-logical propertiesof syntactic units. In these particular casesa sufiicient account is available in other parts of grammar,based on the assignment of pragmatic, semantic,morpho-logical, or phonological properties to the elementsof syntaxand lexicon.It is also true that the distribution of phonological prop-erties in a language can be contingent (a) on themorpho-logical function of the material involved (as in the contrastbetween the phonological effects of the /i/ suffixes inpiracy,based on pirate, and carroty, based on carrot), (b) on itssyntactic function (as when the auxiliary is can be un-accented and contracted to /z/ in I know theparty`s tonightbut not in *I know where the partys tonight), (c) on itssemantics (as when the auxiliarymust be contracted in theidiomatic Wha!s with him?, the uncontracted What is withhim? having some quite different literalmeaning), or (d) onits pragmatic values (as when a falling intonation on whatmakes Myname iswhat?a quizquestion,while a high risingintonation makes it an incredulity or reclamatory question).In examples like (d), it is plausible to suggest that prag-matic values are assigned to phonological elements, but in(a-c), something rather different seems to be going on. Itappears that another component of grammar (morphologyor syntax) or the lexicon is imposing conditions on themakeup of phonological units.2. The Modularity of GrammarThe picture just sketched presumes a modular view of lin-guistic organization, iriwhich (at least) phonology, phonet-ics,morphology, syntax, semantics,pragmatics, and lexiconare treated as to some degree autonomous domains. Most

theorists adopt such a separation of levels.` but there aresome (including many of the generative semanticists' ofthe |9705) who have seen linguistic organization as a seam-less web, in which stipulations and generalizations aboutthe association of sound with meaning can make referenceto any properties of expressions. even properties of verydilierent sorts, and will treat these properties as all on apar with one another.Autonomous levels of analysis are posited partly out ofa desire to restrict the expressive power of grammars. soas to constrain the range of languages consistent with theanalytic framework, and partly as a way of embodying theobservation that interactions between two domains of lin-guistic organization often tum out to be quite limited. bothin that the interacting properties are only a subset of thoserelevant within the domains and in that there is a logicaldirectionality to the relationship, rather than a mutualconditioning.3. Morphology-free and Phonology-free SyntaxIn particular, it is generally assumed that syntax has onlylimited aooess to morphological organization. Many wouldsay syntax is morphology-free (the lexicalist hypothesis`).in the sense that syntactic rulesdo not distribute specificallymorphological properties of expressions. On this hypoth-esis, syntax can be sensitive to abstract properties realizedin morphology, but not to specific inflectional marks forthese properties (to dative case, say, but not to a particulardative case marking, or to a declension class for nouns);and itcanbe sensitive to syntacticsubcategories of lexemes,but not to specificderivational marks for these subcategor-ies (to abstract Ns, say, but not to just those abstract Nswith the derivational suffix -ness).I t is also generally assumed that syntax has only limitedaccess to phonological properties. Indeed, it has beenclaimed (Pullum and Zwicky |988) that syntax is phonol-ogy~free, in the sense that syntactic rules do not distributespecificallyphonological properties of expressions. On thishypothesis, syntax can be sensitive to abstract propertiesrealized in the distribution of phonological features, butnot to the specific phonological features. Though the condi-tions in a syntactic rule can have certain sorts of indirector ultimate phonological consequences (like the temporalordering of the parts of an expression), these conditionsnever seem to distribute phonological properties directly;no language has a syntactic rule stipulating that some con-stituent begin with an obstruent,or have no more than twosyllables, or contain only unrounded vowels, or have stresson its penultimate syllable.The literature on the connection between syntax andphonology contains many apparent counter-examples tothe hypothesis that syntax is phonology-free. But Pullumand Zwicky (1988) maintain that all of these dissolve oncloser examination. Some are merely incorrect statementsof the generalization; it is sometimes said that the invertedmaterial in English subject-auxiliary inversionmust consti-tute a phonological word (Wouldn't I sing? versus *Wouldnot I sing?), but it can be argued that the correct conditionrequires a syntactic word, that is, a constituent of word(rather than phrase) rank in the syntax. Others involvepreferences or tendencies in language use that should notbe seen as rulesof grammar, in particular as rules of syntax;

4477


3/6

Syntax and Phonologyan instance in point is the preference for having long. com-plex, and prosodically heavy constituents ordered at theend of their constructs (favoring the altemative orderingWesaw withpleasure sixteen beautifulbirds that none ofuscould_find in our guide books over Wesaw sixteen beautifulbirds that none of us could find in our guide books withpleasure).Still others involve genuine regularities. but ones thatbelong to some extragrammatical domain; here belongs therequirement in certain fortns of verse that each part of anutterance must have a fixed number of syllables in it, arequirement that is not part of the grammar of the languagein question, but rather is a matter of a set of conventionsfor language use that build upon, or are overlaid on, therules of grammar. A fourth collection of cases comprisesgenuine regularities in the grammar, regularities thatbelong, however, not to syntax but rather to morphology/lexicon; the presence (in geschlagen hit) or absence (intrompetet `trumpeted)of a prefix ge- in German past parti-ciples tums on phonological properties of the verb (it ispresent only for a verbstem with stress on its first syllable),but the rule in question is oneof (infiectional) morphology,not of syntax.A final group ofcases is exemplified by the acceptabilityof coordinations like They have never and will never cometo ourparries, where the existence of syntactically distinctverb forms (bare infinitive and past participle) that happento be phonemically identical (come) permits one word toserve simultaneously in two different, and normally incom-patible, syntactic functions. There is evidently an inter-action here between phonological properties of syntacticwords and their ability to occur in syntactic constructions,an interaction that Pullum and Zwicky (1988) suggest isnot stipulated in the syntactic rules of individual languages,but rather ismade available by a general condition on theapplicability of such rules.4. SubmodularityThe large domains of linguistic organization might them-selves tum out to comprise several autonomous sub-domains. Pragmatics certainly embraces principles of quitedifierent sorts, and the same has been argued for syntax,phonology, and morphology/lexicon.4.1 Modularily in SyntaxIn the case of syntax, Bloomfield (1933) distinguished rulesdescribing sentence types (the declarative type of Penguinscannot jiy. the yes-no question type of Can penguins jiy?,the fragment type of more penguins), those describing thedistribution of anaphoric elements (the reflexive pronounin We congratulated ourselves, the anaphoric gap in Iordered sushi. and Robin sashimi), and those describing thecombination of immediate constituents into constructs (forinstance, the combination of the head V donated with itsdirect and indirect object arguments huge sums and to goodcauses to yield the VP donated huge sums to good causes).Modern theoretical frameworks would further separaterules describing syntactic valency (for instance, the fact thatthere is a class of verb lexemes, donate among them,eligibleto occur with three syntactic arguments-a (nominative)subject, an (accusative)direct object, and an indirect objectmarked by the preposition to) from those describing how4478

heads. their arguments. and their modifiers are assembledinto constructs (for instance. the fact that a head V and i tsnon-subject arguments can be assembled into a VP),4.2 Modularity in PhonologyMost phonological theories have distinguished at least twosubcomponents. either one concerned with morpho-phonemic altemants (like the word~final z - s alterna-tion in dogs-cats) versus one concerned with allophonicvariants (like the partial devoicing word-finally in tag asagainst the full voicing medially in tagging). or one con-oemed with nonautomatic-morphologically. lexically. orsyntactically targeted and/or triggered-phenomena (likethe morphologically triggered /'t/-/sf' alternation inpirate-piracy and the lexically targeted /G-/n, altemation in a. an)versus automatic, entirely phonologically targeted andtriggered, phenomena (as in dogs-cats and tag-tagging).4.3 Modularity in Morphology/lexiconThe intemal organization of a phonological component isclearly not independent of the organization of a morpho-logicalcomponent and of the lexicon. Morphological theor-ies building on the Greco-Roman tradition distinguish atleast two subcomponents that incorporate phonologicalgeneralizations about the stock of lexemes: derivationalmorphology, relating the phonological properties of dilier-ent lexemes stems (sane-sanity, anaphoric-anaphoricity);and infiectional morphology, relating the phonologicalproperties of inflectional forms of a lexeme to the propertiesof one of its stems (dream-dreamt. sleep-slept). To thesesubcomponents can be added one dealing with phono-logical relationships between the different stems of a lexeme(Latin present versus perfect stems ama versus amav love,`curr- versus cucurr- run) and one dealing with lexeme-specific facts about the phonological make-up of forms-their shapes-in external sandhi (an versus a). Each ofthese subcomponents concems itself. at least in part.withphonological relationships, all of them nonautomatic andmorphophonemic.A rather different division of morphology (that of lexicalmorphology and phonology`; see Kaisse and Shaw 1985)posits several layers of morphology, along with its (again,nonautomatic and morphophonemic) phonological con-comitants, arranged out from the stem, without anyassumption that these levels, or strata,necessarily corre-spond to derivational versus inflectional morphology.In addition to these subcomponentsofmorphology, thereare in every language several types of generalizationsaboutthe properties of lexemes: rules that predict some suchproperties from others. Since they describe redundancies inthe lexicon, they are sometimes called lexical redundancyrules. There are, for instance, rules relating some phono-logical properties of fonns to others (though these are usu-ally classed as a species of phonological rule rather thanas lexical redundancy rules). For instance, in German thedefault rule for stress placement is that if a syllable is thefirst one in a fonn (of some lexeme), then it is stressed.And there are rules relating phonological properties of lex-emes to their morphological properties. For instance, inEnglish the default rule is for adjectives and adverbswithstems of more than two syllables to lack an inliectionalcomparative and superlative (shyer, worldlier, atherlier).

AA L '


4/6

S_\n!a.\' and P/ionologyLexical redundancy rules do not necessarily involvephonological properties. There are. for instance, such rulesrelating semantic properties of lexemes to their paradigmclasses (in English. nouns referring to creatures hunted forsport-quail. moose-belong, at least as a default, to thezero plural' declension class) or their grammatical categor-ies (as when the default is for nouns referring to malecreatures to belong to the gender labeled `masculine).The lexical redundancy rules of special interest here arethose expressing some association between phonologicalpropertiesof lexemes, inparticular, phonological propertiesof their stems, and their syntactic properties, in particular,membership in a major syntactic category or in a syntacticsubcategory. There are two types of such associations:those in which syntactic (sub)category predicts aspects ofstem phonology, and those inwhich aspectsof stem phonol-ogy predict syntactic(sub)category.Associations of thefirsttype-e.g., the default in English is that if a lexeme is aninterrogative proform, then its stem begins with /hw/-have been reported with modest frequency in the literature,though it is not always clear whether they represent gen-

eralimtions that should be captured by rules of grammar,preferences or tendencies in language use, oraccidental co-occurrences of the properties in question (having no placein a description of regularities in linguistic organization).It appears that associations of the second type are evenless secure than those of the tirst type. it is unclear whetherthere areany firm examples of reallinguistic generalizationsthat predict syntactic (sub)category of a lexeme fromphonological properties of its stem. For instance, it hasbeen claimed that the default in English is that if a verblexeme with semantics involving agent, patient, and recip-ient roles has a monosyllabic stem, then it can occur ashead in the double-NP-object construction of Illgive Chrisflowers and We told Robin stories; indeed, it has beenclaimed as well that membership in this syntactic sub-category also predicts (at least as a default) themonosylla-bicity of stems. The first claim says that give and tell shouldbelong to the subcategory, the second that donate anddivulge should not. But when semantic regularities are fac-tored out of the data-for example, means-of-communica-tion verbs like cable and telegraph generally belong to thesubcategory (however many syllables they happen to have),and manner-of-speaking verbs like lispandscream generallydo not (again. regardless of their number of syllables)-there seem to be no significant generalizations left. beyondthe overall tendency in English for lexemes, or at least fre-quent lexemes, to have monosyllabic stems.5. Phonological and Syntaetic RulesThe discussion to follow does not further explore the rela-tionship between syntax and the phonological side of mor-phology. lnstead it focuses on automatic phonology-simply phonology in this discussion-and its relationshipto syntax.Rules of the syntactic and phonological components ofa grammar distribute target properties of very dilierentsorts and operate within domains of different types.5.1 Phonological RulesPhonological rulesdescribe the content of phonological (orprosodic) domains of various sizes, from segments and

syllables through phonological words. phonologicalphrases, intonational phrases. and phonological utterances.(The inventory of domains is a matter of controversy: inparticular. several domain types lying between the phono-logical word and the phonological phrase. and between thephonological phrase and the intonational phrase. have beenproposed.) The ultimate units are purely phonologicalproperties like syllabicity_ voicing. nasality.obstruency. andhigh tone. For each type of domain. rules describe howphonologicalconstituents of some smaller domain are dis-tributed with respect to one another-how they are organ-ized in time, by simultaneous or successive occurrence-soas to form instances of the larger domain.Phonological rules can stipulate that some property isdistributed within a domain (e.g.. nasality in the coda ofan English syllable spreads to the nucleus of that syllable.but not to an adjacent syllable, as in the second syllable ofiron, Haldane, and balloon). at the edge of a domain (e.g..syllable-final obstnients inGerman are devoiced, as in Bundband, covenant' and Bundbrucli 'treaty violation`). oracross domain boundaries (e.g.. in fast speech. syllable-final /n/ in English assimilates in point of articulation toa following syllable-initial obstruent. so that infamous canhave a labio-dental nasal in its tirst syllable and incongruousa velar one).It canoften beunclear as to whether the domains withinwhich,at the edge of which, or between which phonologicalproperties are distributed are phonological or morpho-syntactic. Given that there are generalizations about thephonology of stems of Iexemes (the default Latin perfectstem is the present stem plus -tr/-u), fomis of lexemes (thedefault English past tense verb form is the stem plus /d_/).and shapes these forms take in syntactic combinations (thereduced shape of a form of an English auxiliary verb isits tinal consonant: /z/ for is and /ias. /d," for would andhad, etc.), it might not be clear whether a particular phe-nomenon involves a generalization about forms of lexemesversus phonological words. or about shapes of forms versusphonological words/phrases.5.2 Syntactic RulesTo a large extent syntactic rules distribute purely syntacticproperties of various sorts: major syntactic category (e.g.,V), syntactic subcategory (e.g.. auxiliary V eligible to invertwith a subject), rank (word, phrase, or clause), grammaticalrelations (e.g., subject-of), and constituency (e.g., divisi-bility into a subject constituent and a VP constituent).But there are also properties that are cashed out ininliectional morphology (e.g., case in government; person,gender, and number in agreement), in the selection of par-ticular grammatical marker` lexemes (like inlinitival to inEnglish), and in phonology-in prosodic properties likeintonation contours and boundary tones (e.g., the risingterminal intonation of English yes-no questions), in phono-logical altemations (in eliect, phrasal inflections) affectingthe edge segments of syntactic constituents (like the Welshconsonant mutations affecting the first word of a phraseand triggered by, among other things, certain specific pre-positions, as inwedipob mth before every cat, ar bob caihon top of every cat, a'phob /fob/ carh with every cat; orthe English possessive suliix /z/, attached to the last wordof a phrase, as in one person I knots-s opinions), in the

4479


5/6

Syntax and Phonologypresence or absence of some constituent with phonologicalcontent (e.g.. the gap in infonnation questions like Whatdid you make? or the missing subject in imperatives likeBehave _vourself'). and indeed in the temporal ordering ofwords.6. Stacking of DomainsA striking difference between syntactic and phonologicalorganization is that in syntax it is common for units ofsome type to properly contain units of the same type (asin the stacked VPs in might have been being attacked), evento the point of recursivity (aswhen finite clauses containfinite clauses: That your evidence demonstrates the paintdemonstrates that my argument demonstrates nothing).Stacking is rare for the units of phonology-many wouldmaintain that it does not occur at all (the strict layerhypothesis of Nespor and Vogel 1986)-and recursivity isunknown; there is no such thing as a syllable containingother syllables, or a phonological phrase containingphono-logical phrases of similar type. This difference presumablyfollows from the different functions of syntax and phonol-ogy, syntax providing units (syntacticconstituents) that aresemantically interpreted. phonology providing units(phonological domains) that are phonetically interpreted.7. The Morphosyntax-phonology InterfaceGiven the very different functions served by syntactic con-stituents and phonological domains, it is no surprise that asingle expression can be subject to very different hierarch-ical organizations in the two components.Thispoint is familiar frommorphology. Though phonol-ogy for themost part pays little attention to morphologicalstructure, there are circumstances in which morphologicalstructure conditions phonological organization. Forinstance, in English the strong (secondarily stressed andsemantically transparent) prefixesmis-and dis-, as inmistab-ulate and distaste, can be maintained in separate syllables,whereas /s/ is otherwise obligatorily syllabified with a fol-lowing voiceless stop, as inpastiche.Phonology always pays some attention to syntactic struc-ture. The organization of an expression into syntacticwords, phrases, and clausesserves as the basis for itsorgani-zation into phonological words and phrases. For instance,there is normally a phonological phrase break in Englishbetween the subject of a clause and its VP ; [Allofus] [lovespt nach], [People from Italy] [all love spinach]But phonological domains can cut across syntactic con-stituents, as frequently happens with syntactic words thatare phonologicallydependent on adjacentmaterial: Englishcomplementizer that, phonologically dependent on immedi-ately following material. as in the phonological phrasing [Iknow] [that soon] [I`lI win]; and infinitival to, phono-logically dependent on immediately precedingmaterial if itcannot lean on following material, as in the phonologicalphrasing [1ll soon] [persuade them to]. Consequently it isnot enough for a grammar to specify for each expressionwhat constituents of morphology/syntax it has within itand what domains ofphonology it has within it. Thegram-mar must also include generalizations about how (morpho)syntactic spans are associatedwith phonological domains:rules of prosodic domain forrnation.4480

7.1 Prosodic Domain FormationThe investigation of prosodic domain formation rules is inits infancy. Some such rules are specific to individuallexemes or small classes of them; this is the case for thefacts just cited about that and to. and it also applies tothe distribution of unaccented object pronouns in English.which must form phonological phrases with their immedi-ately preceding heads: Igave thefight up. I gave it up (wheregave is it`s head). I gave up thefight. 'I gave up it (whereup is not it`s head).Other prosodic domain formation rulesare more general. gathering a variety of different syntact-ically related items into phonological domains.A number of proposals have been pu t forward for suchgeneral mappings between syntactic constituency andphonological domain organization; see Kaisse and Zwicky(1987) for a variety of these. It has been suggested thatphonological domain fonnation is sensitive. among otherthings, to the difference between maximal and nonmaximalphrases in syntax (in They might have been beingattacked.might have been being attacked is a maximal VP, and allthe other VPs are nonmaximal). to the location of the edgesof syntactic phrases. to the location of the edges of clauses.to the difference between syntactic constituents that are(nonbranching) syntactic words and those that are multi-word phrases (unmodified nouns versus modified nouns.orintransitive verbs versus transitive verbs). and to syntacticrelations between adjacent constituents. inparticular to thedifference between modifying (or adjunct) dependents of ahead and argument (or complement) dependents.A typical proposal is that ofChen (in Kaisse and Zwicky1987) for tone groups in Xiamen Chinese. These domains,within which tone sandhi rules apply. are picked out (inpart) by requiring that every end of a maximal phrase insyntax is also the end of a tone group in phonology.7.2 Direct versus Indirect Syntactic ConditioningProsodic domain fomiation rules provide ameans for indi-rect syntactic conditioning of phonological rules. Syntaxspecifies one set of properties of expressions; prosodicdomain formation rules relate these to a different set ofproperties. phonological incharacter. which include organi-zation into phonological domains; and phonological rulesapply within the domains of the latter sort.The discussion in Sect. 4.2 and 4.3 above also allowedfor direct syntactic conditioning of rules distributingphonological properties, though such rules would be classi-fied as lexical/morphological rather than properly phono-logical. These are nonautomatic phonological mles thatmust be viewed as applying within, at the edges of, orbetween syntactic constituents-rules governing shapes` inthe terminology used above, Pl rules in the terminologyof Kaisse (|985), `precompiled lexical phonology in theterminology used by Hayes (in lnkelas and Zec 1990).There is some controversy as to whether there are any suchrules; the altemative is to maintain that any syntactic condi-tioning is indirect, to claim that prosodic domain formationrules, themselves sensitive to (morpho)syntactic organiza-tion, establish the appropriate spans, within which entirelyautomatic phonological rules apply.The altemative has a theoretical point in its favor, sinceit entirely eliminates one class of rules (syntactically con-strained nonautomatic rules) in favor of types of rules (pro-sodic domain formation rules sensitive to syntactic


6/6

S_\'nta.\' and Pragmatiesorganization; automatic phonological rules) whose exist-ence is not in question. ln practice, deciding between directand indirect syntactic conditioning is not an easy matter.turning largely on the question of whether generalizationsare better captured in nonautomatic phonological rules orin prosodic domain formation rules;The literature coversa wide range of phenomena, both clitics (see Clitics), suchas the English reduced auxiliaries, and extemal sandhi. in avariety of languages including italian (syntactic doubling),French (liaison), Mandarin Chinese (tone sandhi), andMatumbi (vowel shortening; see also Sandhi).7.3 Surface versus Remote Syntactic ConditioningSome syntactic theories posit more than one level at whichthe syntactic properties of an expression are described-one or more levels of remote representation, in additionto the surface level. In such theories thequestion arises ofwhether factors available only at a remote level can condi-tion phonological rules (either directly or indirectly).Though there issome tradition for permitting such a relaxa-tion in theoretical assumptions, the weight of opinion is infavor ofthe simpler theory, inwhich all syntacticcondition-ing is superficial.Several English phenomena have been the focus of dis-cussion on this point: the reduction of auxiliaries (I knowwhere its been versus *I know where its), contraction of towith a preceding verb (What do you wanna make? versus*What doyou wanna vant'sh?), and the assignment of stressand intonation contours to sentences(the infinitival relativeinterpretation favored for I have glans to leave versus thenoun-complement interpretation favored for I have plansto Lage).BibliographyBloomfield L 1933Language. I-Iolt, New YorkInkelas S, Zec D (eds.) |990 The Phonology-syntax Connection.University of Chicago Press, Chicago, ILKaisse E M |985 Connected Speech: The Interaction of Syntaxand Phonology. Academic Press, NewYorkKaisse E M. Shaw P A (eds.) 1985 Phonology Yearbook. Vol. 2:Phonology and the Lexicon. Cambridge University Press,CambridgeKaisse E M, Zwicky AM (eds.) 1987 Phonology Yearbook. Vol.4: Syntactic Conditions onPhonological Rules. CambridgeUni-versity Press, CambridgeNespor M, Vogel I 1986 Prosodic Phonology. Foris, DordrechtPullumG K, Zwicky AM 1988The syntax-phonology interface.In: Newmeyer F J (ed.) Linguistics." The Cambridge Survey,vol. I. Cambridge University Press, Cambridge

AmoldM. Zwicky

Syntax and PragmatimA Rottweiler bit me last week. ( l a )I was bitten by a Rottweiler last week. (lb)It was a Rottweiler that bit me last week. (lc)Last week I was bitten by a Rottweiler. (ld)What bit me last week was a Rottweiler. (le)

There are few lexicaldifferences among these five sentencesbut each of them has a distinct syntactic structure. Thesedifferences in the type and order of phrases seem to make

no difference to the objective information content of thesentences. They are all true at a given time if and only if amember of the class of Rottweilers is in the relation ofhaving bitten the speaker aweek prior to the time of utter-ance; otherwise they are all false. So a speaker who wantsto convey that infomiation has at least these tive means ather disposal. The questions of interest here are what gov-ems the speaker`s choice of one rather than any other ofthese options-is it random?-and what. if any. differencedoes the option chosenmake to the hearer s understandingof the utterance? Before responses to these questions areconsidered some preliminaries are necessary.I. Syntax, Semantics, and PragmaticsThe subject here is the relation between syntax and prag-matics but coming to grips with this is difficult withoutbringingin a third participant. semantics. In order to appre-ciate thepeculiarly pragmatic effects that particular syntac-tic structures may have some understanding of thedifference between semantics and pragmatics is needed.Traditionally, syntax is taken to be the study of the combi-natorial properties of words, semantics to be the study ofmeaning and pragrnatics to be the study of language usage.Meaning is a very vague tenn and a distinction betweensemantics and pragrnatics along these lines leaves open awide rangeof quite different ways of construing the subjectmatter of the two fields. A sharper distinction can bemadebetween two aspects of utterance meaning: (a) those ele-ments of meaning that can be directly decoded from thelinguistic expressions used, that is.meaningwhich they haveacross all contexts of use, and (b) those which depend onextralinguistic contextual information and the interpreter`sinferential abilities. So, for example, B s response to A in(2) communicates informationabout B s impression oftheperson referred to by A in the previous utterance asMary.A would most likely understand B as making a remarkabout Marys personality rather than her temperature andmight well infer that B did not very much like Mary, onthe assumption that coldness of personality is not a likablefeature:

d you likeMary? (2)found her very cold.Now the words actually uttered fall far short of encapsu-lating this infomiation in and of themselves: her is a wordthat can be used to refer to any female creature, cold isambiguous between the temperature and the personalityunderstandings, and the sentence contains no words enco-ding coneepts to do with liking. These processes of assigninga referent to the pronoun, of disambiguation and of deriv-ing certain implications from an utterance are pragmaticprocesses. What is meant by this is that they depend onthe assumption that speakers observe certain standards ofrational cooperative behaviorwhen they communicate andthat hearers interpret utterances with these standards inmind.Without this assumption A (and the reader)wouldhave no grounds for thinking that B is talking about Maryand is making a remark about her personality. Just whatthese standards guiding communication are will be consid-ered in the next section.So utterance understanding is a two-phase process. Thefirst phase is the automatic decoding of linguistic content

4481

syntax and phonology

Documents