VI. Einheiten der morphologischen StrukturUnits of morphological structure

39. Lexical, morphological and syntactic symbolization

I. Introduction2. Symbolization types3. Reduction in phonetic content4. Fusion5. Grammatical factors6. Semantic factors7. Meanings possible for each type of

expression8. References

1. Introduction

Languages offer a variety of means of expres-sion, especially for grammatical notions.Morphology, which is concerned with thecombination of meaningful units into words,can only be fully understood in the contextof the complete array of expression and com-bination types. In this article, similarities anddifferences among the major expression typesor symbolization types - syntactic, free gram-matical, inflectional, derivational, composi-tional and lexical - are described in terms ofa series of continua with both diachronic andcognitive motivation (cf. 2). These continuainvolve phonetic, morphophonemic, gram-matical and semantic criteria (cf. 3-6). Therelation between the formal and semantic cri-teria is non-arbitrary, and thus part of thediscussion will be directed toward explainingwhich meanings are expressed by which sym-bolization type (cf. 7).

2. Symbolization types

Two or more meaningful units may be com-bined in different ways. The form of combi-nation with the least fusion between the twounits and fewest restrictions on possible co-occurrence is syntactic expression, in whichtwo meanings are expressed in two indepen-dent words. For instance, the causative no-tion and an intransitive verbal notion are ex-pressed separately in the phrases cause to dieor make [something] fall. In morphological

expression, the two meaningful units are ex-pressed in distinct morphemes, but they arecombined into a single word. In English, thecausative suffix -en on an adjective yields themeaning 'cause to become .. .', e.g. blacken,straighten. In Turkish, the causative suffixcan be added to dynamic intransitive verbs,as in diisilrmek '[to] make fall, drop' vs,diismek '[to] fall'. Lexical expression involvesthe combination of various semantic featuresinto a single indivisible morpheme which haslexical status, that is, functions as a noun,verb, adjective or other lexical category andhas material content. For instance, walk andwade describe a similar type of movement,but wade includes the additional semanticspecification that the movement be throughshallow water. Lexical distinctions sometimescorrespond to distinctions made grammati-cally elsewhere, such as the English distinc-tion between die and kill ('cause to die') andfall and drop ('cause to fall, allow to fall').Within morphological symbolization it is

customary to distinguish between inflectionaland derivational morphology (cf. Art. 38). In-flectional morphemes are those bound gram-matical morphemes that belong to obligatorycategories, that is, categories from which on.emember must be chosen given the grammati-cal context. English inflections include thecategories number (singular, plural) fornouns and tense (present, past) for verbs. Be-sides affixation, inflectional expression in-cludes other formal processes, such aschanges in the base, as in English ring vs.rang (cf. Ch. VIII). Derivational morphemesare also bound, that is, either affixal or ex-pressed through stem changes, but they donot belong to obligatory categories. Theyalso may change the lexical category of thestem, as when the adjective happy becomesthe abstract noun happiness, or the nounreason becomes the adjective reasonable ~cf.Art. 86, 89). Composition or compoundmgmay also be considered a type of morpholog-

'cal, morphological and syntactic symbolization

bolization. It involves the juxtaposi-in a single word of two lexical units -orphemes that have lexical status andcontent. Thus headache and type-

are compounds (cf. Art. 37, 86-88).tactic expression may occur in morene type. The words involved in syntac-pression may be lexical or grammatical.latter case we speak of free grammati-rpbemes. Examples include non-boundatical morphemes such as articles,sitions or verbal auxiliaries. Free gram-morphemes are usually associated

a particular construction, which mayinvolve certain affixes, and which ex-grammatical meaning. English be go-

t for example, uses the erstwhile lexicalgo, the grammatical morphemes -ing10, and the copula be, whose status ishat intermediate between lexical andatical. This construction qualifies as

grammatical because although the com-t parts are fused (as in gonna) and thection has a fixed position, it is not af-

to a lexical unit. The term periphrasis istimes appropriate for constructions in-. g non-bound morphemes in their ex-'on (cf. Art. 68, 78). Clitics are also a sub-of free grammatical expression. Cliticsammatical morphemes that behave asthey are non-bound in terms of their

ibution - that is, they may change their'on under different grammatical condi-(e.g. Spanish clitic object pronouns pre-the finite verb, as in me 10 dio '(s)he gave it" but follow the imperative and non finite,as in damelo 'give it to me!'), or they arenstruction with a phrase or clause rathera single lexical unit (e.g. the English pos-e, as in the man next door s car, where shes to the end of the noun phrase). De-the distributional independence, how-, clitics are unstressed and behave as

they are phonologically part of theing or following word (cf. Art. 41). Thusrepresent an intermediate type betweentic and morphological symbolization.

(1) Syntax Free grammatical expression

reduction

371

The boundaries between these differentsymbolization types are not discrete, butrather each type is a focal point on a conti-nuum. Therefore the different phonetic, mor-phophonemic, grammatical and semanticproperties of the types will be presented interms of several continuous scales (cr. 3-7).

3. Reduction in phonetic content

As a general rule, lexical morphemes tend tobe longer than grammatical morphemes interms of the number of constituent phono-logical segments. In addition, affixes, espe-cially inflections, tend to use only a limitednumber of the phonological distinctionsavailable in a language. For instance, Englishuses only coronal consonants and the re-duced vowel in its inflectional suffixes. Thiscorrespondence creates a kind of iconicity(cf. Art. 30) in which the more specific, lesspredictable and perhaps contextually moreimportant semantic content of lexical mor-phemes is expressed by the longer, less pre-dictable phonological material (cf. Givan1990: 969). The actual mechanism that cre-ates this iconicity is, however, diachronic. Aslexical material erodes into grammatical ma-terial in the process of grammaticalization,semantic content and phonological contentare eroded in parallel, producing a situationin which the units that have the least seman-tic content are also the phonologically mostreduced (cf. Bybee et al. 1994; Art. 145).Since free grammatical expression evolves

out of lexical units in syntactic combination,we can observe a difference in phoneticcontent between freely formed syntactic con-structions and grammaticized periphrasticones. Thus going to has its full phonetic real-ization in I am going to the library, while inits grammatical use, as in I think I'm gonnacry, both the vowels and the consonants aresubstantially reduced. Periphrastic construc-tions are in turn usually longer than inflec-tions. This relationship is summarized in (1),where the arrow points in the direction ofincreasing reduction.

Inflection

similar relationship can be seen betweenpounds and derivational morphemes.elements of the compound are lexicalenter into the compound as unreduced,t that the stress may be reduced on one

element. If one element is frequently used incompounds it may begin to reduce in size.Thus the -ly suffix that is used on nouns tomake adjectives (e.g. friendly) and on adjec-tives to make adverbs (e.g. quickly) derives

372

from the Old English -lice 'having the ap-pearance or form of' which in turn derivesfrom a noun meaning 'body'. As this one ele-ment of the compound became frequentlyused in many different compounds, it lost itsfinal consonant, unlike the independent formlike which has a long vowel and final conso-nant. (2) shows this diachronic relation:

(2) Composition Derivation

reduction

4. Fusion

The degree of fusion of two elements can bedetermined by both grammatical and phono-logical criteria. The grammatical criteria in-clude the extent to which the relative positionof one is fixed with regard to the other, and

VI. Einheiten der morphologischen Struktur

the extent to which other elements, particu_larly open class elements, are allowed to in-tervene between the two. Phonological cri-teria are applicable in some languages: twounits are more fused if the combined unit istreated as a word by stress and by otherrules, such as vowel harmony rules, Whosedomain is the word. Morphophonemic fusionis manifest in allomorphic variation that isconditioned in the combination. Such varia-tion includes effects that the stem may haveon an affix, either phonologically motivatedor more arbitrary, as in the regular alternatesof the English past tense (as in walkedplayed, and waited) or the more arbitrary al:ternates of the past participle (-ed vs. -en),and effects that the affix has on the stem, asin English past tense forms kept and left (cf.keep and leave). The continuum (3) showshow all of the expression types are rankedwith regard to fusion:

less fused

(3) Syntax Free grammatical expression Composition Inflection Derivation

more fused

The freest type of combination and the onewith the least fusion is the syntactic combina-tion of lexical units, although there are somesyntactic combinations, such as idioms, thatare relatively fixed. Free grammatical expres-sion is by definition not affixal, but some de-gree of fusion exists to the extent that relativepositions are fixed and the construction as awhole has meaning that is not just derivablefrom the sum of the parts. Compounds alsoexhibit fusion since no lexical elements cancome between the parts of a compound, andthere is usually some prosodic feature ofcompounds that treats them as a single unit.Inflection and derivation (which derive

diachronically from free grammatical expres-sion and compounding respectively) are bydefinition affixal and thus more fused thanany other expression types. Again by defini-tion affixes allow no open class items to in-tervene between them and the stem. As to thedegree of fusion, this can be further dif-ferentiated by the extent of allomorphy instem and suffix, since the more the two unitsare phonetically compressed, the greater theireffect on one other will be. The most extreme

degree of fusion in derivation and inflectionis pure stem change with no affix. Inflectionand derivation cover approximately the sameformal range of expression, both allowing af-fixes and stem change, but there are probablymore derivational categories that can have aneffect on the stem than there are inflectionalones, since agreement, tense and mood arevery infrequently expressed by stem changeor even accompanied by stem change (cf. By-bee 1985: 36f.).Greater fusion in both compounding and

derivation leads to lexical expression, al-though in both cases the process usually in-volves meaning as well as form. For instance,the fusion of breakfast involved the reductionof both vowels, and the loss of secondarystress on the second syllable. Derived wordscan also undergo similar fusion as when theproductively formed highness, e.g. in thephrase Your Highness, loses secondary stresson the suffix and comes to rhyme with minus.The fusion criterion gives us two dia-

chronic continua, one leading to inflectionalexpression (4) and one leading to lexical ex-pression (5):

(4) Syntax Free grammatical expression Inflection

fusion

(5) Syntax Composition Derivation Lexical expression

fusion

Lexical, morphological and syntactic symbolization

Grammatical factorse major criterion that distinguishes lexicald grammatical morphemes is membershipopen vs. closed classes, where class refers

o a set of mutually exclusive forms and opens. closed refers to the ease with which thelass accepts new members. Even this crite-'on is scalar, since both parameters are con-inuous. First is the definition of classhood,hich relies upon shared distribution, as thisistribution may be more or less heavily gov-rned by grammatical rules. Derivation andompounding are word-formation devices ine sense that they create new lexical wordsat then enter into grammatical structure ac-ording to the rules for their category, butheir internal make-up is not influenced byroperties of the clause. On the other hand,oth bound and free grammatical mor-hemes are heavily restricted in their distri-bution.The second parameter is the size of the

class: open classes are always large, butclosed classes may be small or large. In inflec-tion, we often find categories with only oneovert member contrasting with a zero, whilein a system that fuses the expression of per-son, number and gender of subject and ob-ject, one class may have dozens of members(e. g. Maung, of North Australia; cf. Capell &Hinch 1970: 66-85). Periphrastic classes canbe larger also: while the class of Englishmodal auxiliaries is heavily bound by gram-matical rules, it still contains nine members- somewhat on the large side for a closedclass in the languages of the world. Cate-gories that are usually thought of as lexicalmay also be represented by closed classes.Adjectives and adverbs in some languagesconstitute closed classes (cf. Dixon 1977; Art.74). Nouns that may be incorporated oftenconstitute a fixed set, which nonetheless mayhave as many as a hundred members (e.g. inTiwi; cf. Osborne 1974: 47-50; Art. 88).Another important grammatical criterion

is the one that best distinguishes inflectionfrom derivation: the criterion of obligatorin-ess (cf. Greenberg 1954; Matthews 21991:43-54; Anderson 1982: 585-591). Obliga-tory categories are associated with the partic-ular structure of the clause or phrase to suchan extent that one member of the categorymust appear. Thus English noun phrases withcount nouns require marking for number.Every Spanish finite verb must have an indi-cation of tense or aspect (either present, fu-ture, preterite or imperfective), mood, person

373

and number. While inflectional categories areby definition obligatory, periphrastic cate-gories mayor may not be. An example of anon-bound obligatory category is the Englishdeterminer system, since a singular nounmust have some type of determiner and thelack of a determiner is meaningful, signallinga non-specific noun (e. g. rabbits multiplyquickly). On the other hand, derivation, com-pounding and lexical expression are neverobligatory in this sense. A consequence ofobligatoriness is that every stem must occurwith some member of the obligatory cate-gory. Thus gaps in inflectional paradigms arerelatively rare, and occur only where they aresemantically motivated (e.g. the first personsingular of the verb 'to rain' may be missing;cf. Art. 67). On the other hand, idiosyncraticgaps in the application of derivational mor-phology are common. Thus we have prevent,prevention but not *preventment and resent,resentment but not *resention.

6. Semantic factorsThe phonetic, morphophonemic, and gram-matical parameters governing the expressiontypes correlate to some extent with certain se-mantic parameters. Some of them are con-cerned with the type of semantic contentfound in the individual unit (cf. 6.1), while asecond set is concerned with the semantic re-sults of the fusion of two units (cf. 6.2).

6.1. Semantic contentLexical content may be characterized as basicor concrete, describing as it does objects, ac-tions, and qualities, and grammatical contentmay be called relational: it constructs thefundamental form of the proposition bytying concrete concepts together (cf. Sapir1921: 93). Relational concepts may be moreor less concrete; the more concrete ones areexpressed by derivation, and the less concreteones by inflection and word order (cf. Art.27). Lexical content is also more specific,with each unit being specified by multiplefeatures, while relational content is highlygeneralized and sometimes fully described bya single feature, such as 'before the momentof speech' (for past tense), or 'more than one'(for plural). Of course, lexical items may bemore or less specific, with a noun like cupbeing more specific than thing, and the verbcalculate being more specific than do. Gram-matical meaning may also be more or lessspecific; a dual is more specific than a plural,a completive, meaning 'to do something thor-

374

oughly and completely, totally affecting thepatient' as in Tucano bapeoami 'he ate all ofit', more specific than a simple past or perfec-

VI. Einheiten der morphologischen Struktur

tive 'he ate' (cf. Sorenson 1969: 174). In gene-ral, however, the following scale holds forgenerality of meaning:

(6) Inflection Free grammatical expression Derivation Lexical expression

generalrelational

specificconcrete

The degree of specificity or generality ofthe semantic content of a morpheme is di-rectly related to the size of the class that mor-pheme belongs to. Inflectional morphemessuch as those signalling tense tend to belongto small sets and to have very general mean-ing. Animate nouns tend to belong to verylarge sets and to have more specific meaning.A number of recent studies suggest that

the range of meanings expressed inflection-ally is characterizable as a small universal set(cf. Bybee 1985; Dahl 1985; Ch. XIII - XIV).The same can be said for some (perhaps all)areas of derivational morphology (cf. Comrie1985; Comrie & Thompson 1985; Wood-worth 1991), and for free grammatical ex-pression (cf. Dahl 1985; Bybee et al. 1994).Lexical meaning may be somewhat more lan-guage-specific, especially the lexical contentof verbs, which are more abstract and rela-tional than concrete nouns (cf. Talmy 1985;Gentner 1981).

6.2. Semantic combinationIn addition to these criteria, which distin-guish types of meaning expressed by indivi-dual morphemes, differences may be found inthe way that meanings are combined in thevarious expression types. The first differenceconcerns the predictability versus idiosyn-cracy of the meaning resulting from the com-bination of two semantic units (cf. Art. 82).Meanings expressed by word order, such asgrammatical relations of subject or object ofverb tend to remain stable over differentnoun and verb categories. Periphrastic con-structions and inflections also tend to havepredictable meaning across a range of combi-nations, although there are some cases of in-flections that have different interpretationswith different stem classes, such as anteriors

or perfectives whose meaning is 'presentstate' with stative predicates. For instance, inKanuri, the suffix -na is used for 'anterior'(or 'perfect') with dynamic predicates, suchas lsana 'they have arrived' but for 'presentstate' with stative predicates such as nongna'I know' or raakna 'I want to, I like to' (cf.Hutchison 1981: 12lf.).Derivational combinations have predicta-

ble meaning when they are formed and oftenretain such predictable meanings, as in English-able, in washable, or curable, but it is alsocommon for derived combinations to take onunpredictable meaning over time. Thus awfulno longer means 'full of awe' but rather can beused about the object or situation which in-spired awe, and is further restricted as an ad-jective to describing the object or situationas bad, which was not part of the originalmeaning. As an adverb with -ly the negativesense is lost and it is just an intensifier: Thisis awfully good. Lexical elements forming acompound have predictable meaning only inthe particular speech or cultural context be-cause a variety of relations may hold betweenthe two members of the compound. Thus inair condition the noun is the object, but in airflow it is the subject; in air brush, air is themeans or instrument and in aircraft, it is themedium for travel. That compounds are notpredictable in meaning becomes evident whenone compares them to new or rare deriva-tional combinations, such as ethnicness orringable, whose parts have a predictable rela-tionship regardless of the context of their use.Lexical expression represents the extremepole of idiosyncracy, since there is no way topredict the meaning of the sum of the partsas there are no formal parts to be summed.Continuum (7) summarizes the degree of pre-dictability of each symbolization type:

predictable

(7) Syntax Free grammatical expression Inflection Derivation Composition

idiosyncratic

The second parameter in semantic combi-nation, relevance, concerns the extent towhich the meaning of one unit directly affects

or modifies the meaning of the other (cf. By-bee 1985: 13). Degree of relevance predictsthe likelihood that two meaningful elements

J9. Lexical, morphological and syntactic symbolization

will occur together and the degree to whichthey will have fused expression. On the syn-tactic level, relevance predicts that the ele-lI1ents internal to the noun phrase and verblPht"ase will occur adja~ent. t~ o~e anotherand in general that the linguistic distance cor-responds to the conceptual distance (cf. Hai-an 1983: 814). With free grammatical andectional expression, the relevance prin-

eiple predicts the ordering of grammatical el-iments with respect to the lexical stem ac-rding to the meaning of the grammatical

,lements and how relevant the meaning is toe stem. Thus since valence and aspect aree categories that are most exclusively rele-t to the verb, they tend to be expressedelements that are closest to the verb stemeither by periphrases or affixes that areose in, or by derivation or lexical expressionf. Bybee 1985; Art. 77). For nouns, gendereets the meaning of the stem more thanber or case, and thus tends to be closerthe stem, or to have derivational or lexicalpression (cf. Art. 73). Verbal categoriesch as tense and mood are relevant to theb, but have the whole proposition in theirpe and .thus are more commonly ex-

375

pressed by more external affixes and periph-rasis, or even clitics.Derivational categories often have a high

degree of relevance, evidenced by the factthat the combination of the affix with thestem produces a new word - that is, there iscohesion in the resulting concept. For in-stance, causatives may have periphrastic, der-ivational or lexical expression, but the degreeof semantic cohesion of the concepts differsaccording to expression type. The Japaneseperiphrastic causatives sin + sase 'die +cause', and tamar + sase 'stop + cause' donot mean the same thing as the more fusedcausatives such as tome 'stop' or lexical caus-atives such as koros 'kill'. The synthetic formsconnote simultaneity of cause and result,with physical contact between causer andcausee, while the periphrastic forms allow theinference that the cause was less direct andtwo events may have been involved (cf. Hai-man 1983: 784). This example shows alsothat in lexical expression the units of meaningare so relevant to one another that they pro-duce a coherent and discrete concept of theirown. The scale of degree of relevance of twosemantic features to each other is shown in(8):

less relevant

Syntax Free grammatical expression Inflection Derivation Lexical expression

more relevant

Meanings possible for each type ofexpression

Jkcause of the formal and semantic charac-teristics of each expression type, certainineanings have a tendency to be expressed insome of these ways but not others. For com-pounding the mere juxtaposition of lexicalitems is meaningful. This meaning does notderive from linguistic units with inherent se-mantic content as in the other expressiontypes, but rather has its source in the infer-ences that can be made from the linguisticand non-linguistic context. The relationshipbetween the two elements of a compound canbe any of the relationships that may hold be-tween the two elements in the categories rep-resented, that is, two nouns, noun-verb, etc.(cf. 6.2). Which relation is intended dependsUpon the meanings of the two lexical itemsand the broader context in which the com-pound is used.Inflection, derivation and free grammati-

cal symbolization express many similar se-mantic notions. Inflectional expression is

used for the smallest, most constrained set ofnotions, because it must be semanticallygeneral enough to be applicable to all (or al-most all) lexical stems (cf. Bybee 1985: 16-19). Derivation tolerates much more lexicalidiosyncracy and specificity of meaning. Freegrammatical expression is also often specific.Thus in the domain of tense and aspect themore specific and lexically restricted mean-ings - completives, inchoatives and iteratives- may be expressed either periphrastically orderivationally. Progressives are usually peri-phrastic, and futures are sometimes peri-phrastic and sometimes inflectional (cf. Dahl1985: 184-189). The more generalized tenses,such as present and past are usually inflec-tional (although they may be periphrastic),but they are never derivational, as theircontent is deictic and has the whole proposi-tion in its scope. The more generalized as-pects, perfective and imperfective, are almostalways inflectional. In the less usual circum-stance that perfective and imperfective areexpressed derivation ally, as they are in Slaviclanguages, the sense is more specific, in that

376

the perfective expresses the attainment of alimit, and there is considerable lexical idio-syncracy (cf. Dahl 1985: 84-89; Art. 109 and110). The moods and modalities that can beinflectional are again the most generalizedones, those which have a whole propositionin their scope - subordinating moods, im-peratives and epistemic moods. Modalitiesexpressing desire, obligation, or ability mostoften take free grammatical expression (cf.Bybee 1985: 166; Art. Ill). Derivational orperiphrastic expression is also possible forvalence-changing constructions, such as caus-ative, transitivizers and intransitivizers, andperiphrastic expression is possible for voicefunctions, such as passive, reciprocal and re-flexive (cf. Art. 107 and 108). Morphemes in-dexing person, number and gender agree-ment on verbs are almost always inflectional,but some languages have a derivational cate-gory of "plurality of action" that partiallyoverlaps with number agreement (cf. Art.100).Morphemes in construction with nouns

such as numeral classifiers are usually notbound to the noun, but may have some in-ternal fusion of their own (cf. Art. 75). Whilecase markers are usually bound to the nounif they are postposed, preposed markers ofgrammatical and locative relations are usu-ally not bound (cf. Art. 102). Definitenessmarkers may be bound or not bound, butneither case nor definiteness have any deriva-tional counterparts. Plural, on the otherhand, which is often inflectionally expressedon nouns, has a derivational counterpart inthe collective, and plural in Indo-Europeanlanguages in some ways resembles derivation(cf. Art. 100 and 101). Gender and similarnoun classifications are lexically determinedand thus more closely resemble derivationalcategories, even where the classification in-teracts thoroughly with case and number (cf.Art. 97). Diminutive and augmentative mustalso be considered derivational, even wheretheir use is very productive and interacts withinflection (cf. Art. 99). Adjective comparisonmay be either derivational or periphrastic (cf.Art. 114). Agreement for gender and numberin adjectives is inflectional, even though innouns gender is derivational or lexical (cf.Kurylowicz 1964). In adjectives gender andnumber are not inherent to the adjective, butrather an index of the same categories innouns. Other common derivational processesare those which nominalize verbs and verbal-ize nouns. Occasionally these same processes

VI. Einheiten der morphologischen Struktur

are accomplished periphrastically (cf.Woodworth 1991).Languages utilize the different symboliza-

tion types to varying degrees. The well-known distinction between analytic and syn-thetic languages is based on the observationthat some languages (e. g. Mandarin Chi-nese), termed analytic, make very little use ofinflectional marking and rely instead on syn-tactic expression and composition, sometimeswith derivational morphology as well. Typi-cal synthetic languages (such as Latin), onthe other hand, make use of inflectional mor-phology to a greater degree, while also usingthe other symbolization types. Since the dis-tinctions between expression types involveboth formal and semantic criteria, languagesthat use certain formal types of expressionwill express certain types of meaning, andlanguages that lack certain types of expres-sion will lack certain types of meaning (cr.Sapir 1921). Thus languages without inflec-tion will lack person and number categoriesand often also lack tense distinctions such aspast and present, and the most generalizedaspectual distinctions, such as perfective andimperfective (cf. Dahl 1985). On the otherhand, if these languages have periphrasis,which they usually do, they may have com-pletive, iterative and progressive aspects (cr.Dahl 1985; Bybee et al. 1994). Some lan-guages have very rich derivational possibilit-ies while others have none, again affectingthe types of meaning that are overtly ex-pressed (cf. Art. 115 and 116).

8. References

Anderson, Stephen R. (1982), "Where's Morphol-ogy?". Linguistic Inquiry 13, 571-612Bybee, Joan L. (1985), Morphology: A Study of theRelation between Meaning and Form. Amsterdam,Philadelphia: Benjamins (Typological Studies inLanguage 9)Bybee, Joan & Perkins, Revere & Pagliuca, Wil-liam (1994), The Evolution of Grammar: Tense, As-pect and Modality in the Languages of the World.Chicago, London: Chicago Univ. PressCapell, A[rthur] & Hinch, H. E. (1970), MaungGrammar, Texts and Vocabulary. The Hague, Pans:Mouton (Janua Linguarum Series Practica 98)Comrie, Bernard (1985), "Causative Verb Forma-tion and Other Verb-Deriving Morphology". In:Shopen (ed.), 309- 348Comrie, Bernard & Thompson, Sandra A. (1985),"Lexical Nominalization". In: Shopen (ed.), 349-398

40. Word boundaries

Dahl, Osten (1985), Tense and Aspect Systems. Ox-ford, New York: Blackwell

Dixon, R[obert] M. W. (1977), "Where Have Allthe Adjectives Gone?". Studies in Language I, 19-80 [reprinted in: Dixon, R[obert] M. W. (1982),Where Have All the Adjectives Gone? And Other Es-says in Semantics and Syntax. Berlin etc.: Mouton(Janua Linguarum Series Maior 107), 1-62]

Gentner, Dedre (1981), "Some Interesting Differ-ences between Verbs and Nouns". Cognition andBrain Theory 4, 161-178

Giv6n, Talmy (1990), Syntax: A Functional-typo-logical Introduction, Vol. II. Amsterdam, Philadel-phia: Benjamins

Greenberg, Joseph H. (1954), "A Quantitative Ap-proach to the Morphological Typology of Lan-suage". In: Spencer, Robert F. (ed.), Method and'erspective in Anthropology. Papers in Honor ofWilson D. Wallis. Minneapolis: Univ. of MinnesotaPress, 192-220 [reprinted 1960 in Internationaloumal of American Linguistics 26, 178- 194; alsoin: Greenberg, Joseph H. (1990), On Language: Se-lected Writings, ed. by Keith Denning & SuzanneKemmer. Stanford/CA: Stanford Univ. Press,-25]

aiman, John (1983), "Iconic and Economic Moti-tion". Language 59, 781-819

IntroductionSemantic criteriaDistributional criteriaPhonological criteriaWord boundaries in phonological rulesConcluding remarksReferences

Introduction

e main problem to be discussed in this arti-can be illustrated by asking the question:w many words does its very title, viz. theuence word boundaries, consist of? The an-r obviously depends on what we mean by(cf. 1.1 and Art. 26). And in order to

ify the notion of word boundaries (cf. 1.3),relations to other linguistic boundariesuld be considered (cf. 1.2).

• Wordst ordinary people who have goneugh an elementary school education un-btedly identify the notion word with or-

377

Hutchison, John P. (1981), A Reference Grammarof the Kanuri Language. Madison: African StudiesProgram, Univ. of WisconsinKurylowicz, Jerzy (1964), The Inflectional Cate-gories of Indo-European. Heidelberg: WinterMatthews, P[eter] H. eI991), Morphology. Cam-bridge etc.: Cambridge Univ. Press [11974]Osborne, C[harles] R. (1974), The Tiwi Language.Canberra: Australian Institute of AboriginalStudies (Australian Aboriginal Studies 55; Linguis-tic Series 21)Sapir, Edward (1921), Language: An Introductionto the Study of Speech. New York: Harcourt, BraceShopen, Timothy (1985, ed.), Language Typologyand Syntactic Description, Vo!. III: GrammaticalCategories and the Lexicon. Cambridge etc.: Cam-bridge Univ. PressSorenson, Arthur P. (1969), The Morphology of Tu-kano. Ph.D. dissertation, Columbia Univ.Talmy, Leonard (1985), "Lexicalization Patterns:Semantic Structure in Lexical Forms". In: Shopen(ed.), 57 -149Woodworth, Nancy L. (1991), From Noun to Verband Verb to Noun: A Cross-linguistic Study of Class-changing Morphology. Ph.D. dissertation, StateUniv. of New York at Buffalo

Joan Bybee, Albuquerque (U S.A.)

tho graphical word, which can be informallydefined as an unbroken sequence of letters,viz. a sequence which contains neither blanks,or spaces, nor commas, parentheses or thelike. In that sense, the sequence word bound-aries consists of two words. This is quitestraightforward, but what if we add a hy-phen: word-boundaries? Most people wouldprobably agree that now there is only oneword, although nothing else seems to havechanged; this appears unproblematic, how-ever, since the definition can easily be ad-justed to the effect that a hyphen is allowedwithin the letter sequence of a single word.But there can be meaningful (as well as

meaningless) discussions on this aspect of or-thography: To take just one example, shouldboth in so far and insofar be allowed? In sucha discussion the argumentation transcendsorthography, involving e. g. semantics (is themeaning composed of the meanings of in, soand far - as is the case with normal wordsforming syntactic phrases - or not? cf. 2),