Copyright © 1999 Massachusetts Institute of Technology
The lexicon is a list of the morphemes in a language, containing information that characterizes the SYNTAX (hence distribution), MORPHOLOGY (hence PHONOLOGY), and meaning for every morpheme. Thus, for a word such as forget, the lexical entry states its meaning, that it is a verb and takes a nominal or clause complement (I forgot the picnic, I forgot that the picnic was at 3:00), and that its phonological representation is /forgεt/.
However, the study of the lexicon in linguistics and in cognitive science saw a major shift in perspective in the 1980s and early 1990s stemming from the realization that lexicons are much more than this: they are studded with principled phenomena, which any lexical theory must explicate (Wasow 1977 is a classic study). In particular, the lexicon is the domain of a set of linguistic operations or regularities that govern the formation of complex words from lexical components. A much-studied example is passivization (e.g., arrest > arrested as in The thief was arrested by the police). Why does the passive form of arrest, unlike the active, require that its "logical subject" (see GRAMMATICAL RELATIONS) the police appear in a prepositional phrase beginning with by, instead of in object position, or indeed subject position? Why does a verb like suffocate have both a causative and a change-of-state meaning (The pillow suffocated Mary, Mary suffocated), and why is the object of one verb the subject of the other? Work on the lexicon has aimed to uncover the principles underlying regular lexical phenomena and to examine their implications for learning and processing.
Most lexical research has centered on the words of obvious semantic importance in a sentence, primarily substantives such as nouns and verbs, treating words like determiners and complementizers (e.g., that in I think that it is raining) as peripheral to linguistic structure. Recent work argues, however, that these words determine properties of entire phrases, hence their lexical properties are now of considerable interest.
In sum, the theory of the lexicon must encompass two subtheories: one (the primary focus of this article) governing words that are "meaningful" in the obvious if pretheoretic sense, and the other governing words that are functional, or grammatical in character.
The lexicon determines the ability of nouns, verbs and adjectives to combine with arguments. For example give takes three arguments (She gave the box to Peter), eat takes two (She ate the sandwich) and rise takes one (The balloon rose rapidly). The representation that includes this information is often called an "argument structure" (Williams 1981; Grimshaw 1990; Levin and Rappaport Hovav 1995).
The most important source of evidence concerning the representation of argument structure is alternations in argument organization, such as active/passive and causative/change-of-state mentioned above. Others include the "dative alternation": They gave a present to the teachers/They gave the teachers a present; morphological causativization: large > enlarge; and nominalization: legislate > legislator, legislation.
Certain properties of arguments are important in explaining their behavior. Arguments can be usefully classified according to their semantic role as agent, theme, or goal (Jackendoff 1972; see THEMATIC ROLES). Williams (1981) argued that the "external argument" is singled out for special grammatical and lexical treatment. Thus the argument structure of give might be (1), where the first argument is marked as external, the others as internal.
Further work on morphology and syntax suggests that some verbs have no external argument, including unaccusatives and perhaps some psychological predicates, and that this has important morphological and syntactic consequences (Burzio 1986; Belletti and Rizzi 1988; see also RELATIONAL GRAMMAR).
The existence of strong semantic regularities within the alternation system sheds light on the semantic representation of substantives, most prominently verbs. For example, the alternation of argument realization seen in the suffocate examples above, where the object of the transitive verb corresponds to the subject of the intransitive, is a completely regular phenomenon but is found only with causative verbs. The pillow suffocated Mary can be paraphrased roughly as "The pillow caused Mary to suffocate". Verbs that lack this interpretation do not show the alternation, thus The thief robbed the bank does not have a counterpart *The bank robbed.
This constellation of facts, which is crosslinguistically remarkably stable, can be explained by lexical theory, if the verb meanings can be broken down or decomposed into smaller meaning components, one of which (represented here by CAUSE) is part of the meaning of causative verbs. A verb like suffocate has two semantic representations, often called "lexical conceptual structures," one with the CAUSE component and one without.
Because the CAUSE component has its own argument, the causative version has one more argument than the noncausative: this holds mutatis mutandis for all causative verbs in all languages. Because the argument of CAUSE is higher in the argument structure than the other, it occupies the syntactic subject position when it is present; when it is absent, the other argument occupies this position in English, which requires that the subject position be filled in all sentences. Hence we find the alternation in argument realization with these verbs. Verbs like rob do not have an intransitive counterpart because their semantic analysis does not have the required properties.
A related line of reasoning allows us to explain why verbs like apologize occur only intransitively (*He apologized the impolite clerk). Argument structures are subject to a very general constraint: only one argument of a given semantic type is allowed in each one. Analysis shows that apologize and CAUSE have arguments of the same type, with the thematic role agent. Hence the two cannot be combined into a single argument structure, hence the causative cannot exist. Thus generalizations and gaps in the system can be explained, not merely described within this system of representation. Further quite remarkable evidence comes from the discovery by Talmy (1985) that there are crosslinguistic differences in what meaning components can be combined. He showed that French systematically lacks verbs like English float, which encode motion and manner in a single morpheme.
Despite the progress that has been made in understanding lexical alternations, they still pose some powerful challenges. On the one hand, many verbs participate in them and many languages show similar patterns. On the other hand, not all verbs participate in them, and some fail to show the alternation, apparently for no very good reason. For example, donate appears in only the first of the two configurations that give occurs in above, and similarly, rise does not occur in the causative (The balloon rose into the sky, *They rose the balloon into the sky). Yet rise does not have an argument that is plausibly an agent, unlike apologize. Gaps such as these pose an important problem for learning: we can explain the existence of lexical generalizations only if learners generalize heavily, but then how can we explain the existence of exceptions? (See WORD MEANING, ACQUISITION OF; Pinker 1989, Gleitman and Landau 1994.)
A very appealing answer holds that there are no entirely arbitrary exceptions once the generalizations themselves are properly understood. This is probably a fair way to characterize the general direction of the field, but it is not an easy commitment to make good on. One idea is that the meaning of the verb in one configuration is systematically different from the meaning of the verb in others, as the result of a lexical rule changing the verb's semantic representation (see Levin and Rappaport Hovav 1995). Each such lexical rule carries out a specified meaning change on any verb with the appropriate semantics, with syntactic changes emerging as a consequence.
However, consider a marginal but interpretable example like He sang his mother out of the house, where sang, under this line of analysis, would mean "to cause Y to move by singing." Why couldn't there be a verb which meant only this and did not also mean what sang usually does? The best answer is that this is not a possible word meaning, but then it cannot be the meaning of sang as used here. Some current work suggests that effects such as this extended use of sang are due to an interaction between the verb meaning and the syntactic and semantic context the verb appears in, which makes it possible to construe the verb in a certain way. From this perspective, related to that developed by Pustejovsky (1995) and in Construction Grammar (Goldberg 1994), verbs that do not alternate have some special property prohibiting alternation; massive alternation is the norm rather than the exception.
Although the argument structure of the predicate of a clause determines the number and kind of arguments that can appear in it, the grammatical structure of the entire clause is determined by properties of the function words, previously labeled misleadingly as "minor categories." A noun along with its satellite expressions forms a noun phrase of which it is the "head," the element that determines the properties of the phrase as a whole. Remarkably, it turns out that the same is true for the "minor" syntactic categories, like complementizer and determiner: they also head phrases. Thus the head of the entire complement clause in I think that it is raining is that, and not, as previously thought, raining or a phrase of some kind. It follows then that the primary item determining the grammatical properties of the clause is the complementizer itself (see HEAD MOVEMENT and X-BAR THEORY). Since this discovery, the field has faced a completely new question: what properties does the lexicon of functional morphemes have? Clearly, notions like thematic role are entirely irrelevant for words like that and the. These morphemes code properties like "type" (e.g., interrogative versus propositional) or definiteness. How much crosslinguistic variation can be shown to follow from differences in the functional lexicon? Related work on learning attempts to establish whether these morphemes and associated phrases are present in early child language (Déprez and Pierce 1993; Clahsen 1990/91; Radford 1990; see also PARAMETER-SETTING APPROACHES TO ACQUISITION, CREOLIZATION, AND DIACHRONY and MINIMALISM).
The content-function word distinction has a long history in cognitive science (see Halle, Bresnan, and Miller 1978). The lexical representation of both types of morpheme is now an issue of central importance within linguistic theory.
Belletti, A., and L. Rizzi. (1988). Psych verbs and theta theory. Natural Language and Linguistic Theory 6:291-352.
Burzio, L. (1986). Italian Syntax: A Government-Binding Approach. Dordrecht: Reidel.
Clahsen, H. (1990/91). Constraints on parameter setting: a grammatical analysis of some acquisition stages in German child language. Language Acquisition 1:361-391.
Déprez, V., and A. Pierce. (1993). Negation and functional projections in early grammar. Linguistic Inquiry 24:25-67.
Gleitman, L., and B. Landau, Eds. (1994). Lexical acquisition. Lingua 92: 1.
Goldberg, A. E. (1994). Constructions: A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press.
Grimshaw, J. (1990). Argument Structure. Linguistic Inquiry Monograph 18. Cambridge, MA: MIT Press.
Halle, M., J. Bresnan, and G. Miller, Eds. (1978). Linguistic Theory and Psychological Reality. Cambridge, MA: MIT Press.
Jackendoff, R. (1972). Semantic Interpretation in Generative Grammar. Cambridge, MA: MIT Press.
Levin, B., and M. Rappaport Hovav (1995). Unaccusativity: at the Syntax-Semantics Interface. Cambridge, MA: MIT Press.
Pinker, S. (1989). Learnability and Cognition: The Acquisition of Argument Structure. Cambridge, MA: MIT Press.
Pustejovsky, J. (1995). The Generative Lexicon. Cambridge, MA: MIT Press.
Radford, A. (1990). Syntactic Theory and the Acquisition of English Syntax. Oxford: Blackwell.
Talmy, L. (1985). Lexicalization patterns: semantic structure in lexical forms. In T. Shopen, Ed., Language Typology and Syntactic Description 3: Grammatical Categories and the Lexicon. Cambridge: Cambridge University Press, pp. 57-149.
Wasow, T. (1977). Transformations and the Lexicon, in Formal Syntax. In P. Culicover, T. Wasow, and A. Akmajian, Eds. New York: Academic Press, pp. 327-360.
Williams, E. (1981). Argument structure and morphology. Linguistic Review 1:81-114.
Baker, M. (1988). Incorporation: A Theory of Grammatical Function Changing. Chicago: University of Chicago Press.
Dowty, D. (1991). Thematic proto-roles and argument selection. Language 67:547-619.
Hale, K., and J. Keyser. (1993). On argument structure and the lexical expression of syntactic relations. In K. Hale and J. Keyser, Eds., The View from Building 20. Cambridge, MA: MIT Press.
Grimshaw, J. (1979). Complement selection and the lexicon. Linguistic Inquiry 10:279-326.
Jackendoff, R. (1990). Semantic Structures. Cambridge, MA: MIT Press.
Copyright © 1999 Massachusetts Institute of Technology