《Entity and Event Coercion in a Symbolic Theory of Syntax实体和事件的语法符号理论胁迫.doc》由会员分享，可在线阅读，更多相关《Entity and Event Coercion in a Symbolic Theory of Syntax实体和事件的语法符号理论胁迫.doc（51页珍藏版）》请在三一办公上搜索。
1、Entity and Event Coercion in a Symbolic Theory of SyntaxLaura A. MichaelisUniversity of Colorado at Boulder1. IntroductionWhere does sentence meaning come from? Leaving aside the inference strategies targeted by the Gricean paradigm, formal models of the syntax-semantics interface have supplied a si
2、ngle answer to this question: the words of the sentence and the frames which those words project. Since word combination in formal theory is inextricably tied to phrase building, the drivers of sentence semantics are not simply words but, more particularly, the heads of syntactic projections. The as
3、sumption that the licensing of sisterhood relations is the unique privilege of lexical heads is woven into the formal conventions of formal theory, e.g., phrase-structure rules like those in (1), in which the optionality of complements, specifiers and adjuncts is defined over a set of lexical classe
4、s distinguished by their projection behaviors:(1)a.VP V (NP) (PP)b.NP (determiner) NModels of sentence meaning based on lexical projection provide a straightforward picture of the syntax-semantics interface: while words determine what a sentence means, rules of morphosyntactic combination determine
5、how a sentence means. While rules of syntactic combination assemble heads and their dependent elements into phrases, they play no role in either the licensing or construal of arguments. It is apparent, however, that the syntax-semantics mapping is less tidy than the foregoing statement would imply.
6、In particular, the identification of licensors with syntactic heads cannot always be maintained. This is shown by the following examples, in which the projection properties of the boldfaced items are distorted in various ways. These distortions involve, respectively, nominal morphosyntax (2), verbal
7、 thematic structure (3), and those aspects of verbal morphosyntax which are determined by the aspectual class of the verbal projection or situation radical: (2)Nominal Morphosyntaxa.Give me some pillow.b.They sampled some wines.c.She had a beer.(3)Semantic Framea.Down at the harbor there is teal-gre
8、en clubhouse for socializing and parties. Beside it sparkles the community pool. (Vanity Fair 8/01). b.When a visitor passes through the village, young lamas stop picking up trash to mug for the camera. A gruff police monk barks them back to work. (Newsweek 10/13/97)(4)Aspectual Morphosyntaxa.She li
9、ked him in a minute.b.Im feeding him a line and hes believing every word.c.She washes the car. In (2a), a word which denotes a bounded entity, pillow, is embedded in the morphosyntactic frame ordinarily projected by a mass noun, while in (2b-c) the inverse is the case. In (3), two monovalent verbs,
10、sparkle and bark are embedded, respectively, in bivalent frame comprising a location and a theme and a trivalent frame, comprising an agent, a theme and a goal. In (4a-b) stative situation radicals are combined with aspectual operators which logically require tenseless propositions denoting events.
11、In (4a), the state radical She like- him combines with a frame adverbial (in a minute), which is logically compatible only with those predications which do not entail downward to subintervals, i.e., telic events (Herweg 1991). In (4b), the state radical He believe- every word combines with Progressi
12、ve morphology. This combination is unpredicted by verbal aspect. Since the Progressive maps events to medial states, it appears to apply vacuously in this context (see Vlach 1981, Langacker 1987, Herweg 1991, De Swart 1998). In (4c), an event radical, She wash- the car, combines with Present inflect
13、ion. While this combination is widely attested it too involves a distortion of verbal aspect: (4c) does not denote a unique event, as would its simple Past counterpart. As a momentaneous sampling device, the Present cannot accommodate the positive temporal profile of an event. Instead, the Present a
14、ppears to index the class of stative situations, e.g., a state of the world in which car-washing takes places at regular intervals. A model of the syntax-semantic interface based solely upon lexical-head licensing would, of course, fail to account for the fact that all of the examples in (1-3) have
15、coherent, consistent interpretations. For example, the verb bark in (3b) is uniformly construed as denoting (metaphorical) caused motion, while the situation radical I like- him in (4a) receives an inchoative interpretation. Although these interpretive effects might be dismissed as the products of m
16、anner- or relevance-based implicatures, the relevant implications do not obviously qualify as generalized implicata: because they are based on the presence of specific lexical items, these implications, like conventional implicatures, are neither detachable nor defeasible. The foregoing examples the
17、refore suggest that there is not in fact a single source of sentence meaning: conceptual content comes not only from words but also from an inferential procedure which bridges semantic gaps in morphosyntax. I will refer to this procedure as implicit type-shifting, reserving the more widely used term
18、s coercion and coercion effect to refer to the enriched representations produced by the reconciliation mechanism in question. Our exploration of implicit type-shifting thus far enables us to draw the following three generalizations: First, semantic operators can apply even in the absence of an appro
19、priate situation-type argument, since the argument can adapt to the requirements of the functor. This fact is difficult to model in a noncircular way, since a given operator must not only operate on the output of an inference rule, but also trigger the very inference rule which enables it to apply.
20、Second, the patterns which trigger coercion effects do not have a uniform syntactic characterization. The coercion trigger may be a syntactic head, as in the case of the Progressive, where the auxiliary head be selects for a participial complement of the appropriate aspectual class, forcing a dynami
21、c reading in the mismatch condition (4b). The coercion trigger may be a specifier like some in (2b), which selects for a noun whose denotatum is a mass. Finally, it may be an open schema, as in (3), where the relevant scene-construal properties follow from the presence of specific grammatical functi
22、ons, rather than being attributable to a given verb or argument. Third, coercion effects are produced by both type-shifting schemas and type-sensitive schemas. An example of the former is the Progressive construction. An example of the latter is the Frame Adverbial construction (3a). Coercion effect
23、s appear to indicate a modular grammatical architecture, in which the process of semantic composition may add meanings absent in the syntax in order to ensure that various functors, e.g., the indefinite article, receive suitable arguments. One such model, proposed independently by both Jackendoff (1
24、990, 1997) and De Swart (1998), involves the interpolation of coercion operators in semantic structure. In the case of (2c), for example, a specific coercion operator would be used to derive a count type from a mass type, making beer a suitable argument for the indefinite article. The interpolated-f
25、unctor model successfully extricates two widely conflated head propertiesthat of being a syntactic head (determining the distribution of the phrasal projection), and that of being a semantic headcalling for an argument of a particular type (Zwicky 1985, Croft 1996). However, this model also has thre
26、e significant failings: First, it requires a powerful indexing mechanism to constrain coercion operations. Jackendoff (1997:50) notes this issue, pointing out that such operations might “insert arbitrary material into arbitrary arrangements”. De Swart (1998:361) seeks to avoid such overgeneration by
27、 assuming that a coercion operator is introduced only when there is a trigger for it. For example, a unitizing coercion operator might be indexed to the class of linguistic expressions requiring count-noun sisters, e.g., the indefinite article. However, by enabling a given linguistically expressed o
28、perator to invoke a given coercion operator on an as needed basis we do not thereby ensure that that this coercion operator will appear only where needed. For example, there is no obvious means by which to prevent the unitizing operator from intervening between the determiner the and a mass-noun sis
29、ter (e.g., beer) in the expression the beeran unwelcome result since this expression need not denote a portion or variety of beer. Coercion operations may be morphosyntactically invisible, but if their representation owes nothing to morphosyntax it is not obvious how they can be constrained. Second,
30、 it misses the following generalization: both the match conditions upon which lexical projection is based and the mismatch conditions which trigger implicit type-shifts are created by morphosyntax. On the modular account, there no obvious relationship between strict (projection-based) composition an
31、d enriched (coercion-based) composition. The enriched representations do not appear to owe anything to the syntactic configurations in which the particular functor appears. In fact, Jackendoff (1997:50) admits that enriched composition considerably complicates the syntax-semantics interface. Third,
32、it cannot account for cases of template-based coercion, as in (3). As noted above, the coercion effects in question cannot be traced to the presence of a specific functor, be it a verb or an argument. Instead, the modulation of meaning is the result of the verbs conformity to a linking pattern whose
33、 valence set properly includes that projected by the verb. Fourth, it provides no rationale for the existence of type-sensitive operators. What use does an interpretive module have for a set of identity functions? Since functions in construal-based semantic theories are intended to represent cross-d
34、omain mappings, type-sensitive operators, whose input and output types are identical, appear to serve no explanatory role.As an alternative to modular models, I will propose an account of implicit type-shifting based upon the grammatical construction. This account will draw upon the mechanisms and a
35、rchitecture of Construction Grammar (Fillmore et al. to appear, Kay & Fillmore 1999, Zwicky 1994, Goldberg 1995, Michaelis 1994, Michaelis & Lambrecht 1996, Koenig 1999). In this model, the grammar is a network of symbolic rules of morphosyntactic combination. As in Bybees (1995) conception of morph
36、ological storage and processing, rules traditionally conceived in processual terms are replaced with schemas which differ from one another with regard to the level of specificity (e.g., whether or not particular words or affixes are invoked) and productivity, as determined both by the restrictivenes
37、s of the schema and its type frequency (see Bybee 1995:432). In addition, constructions represent diverse formal objects. Grammatical constructions determine: constituent-structure relations, dependency relations, role-function linkages, linear orderings, and combinations thereof (Zwicky 1994). Gram
38、matical constructions are combined with one another, and with lexical items, via superimposition, a mechanism whose technical implementation is unification (Fillmore et al. forthcoming, Kay & Fillmore 1999). Grammatical constructions refer in the same way that words do: they denote typesamong them c
39、lasses of entities and events. Accordingly, coercion is not merely the resolution of semantic conflict, but is instead the resolution of conflict between constructional and lexical denotata.Endnotes The idea that constructional requirements may override lexical requirements in the case of NPs like a
40、 beer is not part of the conception of Construction Grammar put forth in Kay & Fillmore 1999. In that version of the model, conflict of this type would represent a unification failure, since the bounded- feature of the noun beer would conflict with the bounded+ requirement that the Indefinite Determ
41、ination construction imposes upon its nominal daughter. Therefore, the licensing of tokens like a beer requires the intercession of type-shifting constructions. A type-shifting construction has an external semantic value which is distinct from that of it sole daughter node. The MassCount constructio
42、n, for example, unifies with a mass noun like beer. Its external semantics is that of a count noun, which can thereby unify with Indefinite Determination. Type-shifting constructions are essentially lexical rules, and as such fail to capture an important generalization, since type-shifted nominals a
43、re freely generated but not indexed to the morphosyntactic contexts which trigger the relevant type shifts. Further, use of the box-within-a-box constructions for type-shifting violates the spirit of a model which, in the interest of concreteness, eschews nonbranching domination in phrase structure.
44、 That is, in CG, no phrase consists simply of a noun. If a given lexical noun is of the appropriate semantic class, it will simply unify directly with any grammatical-function position in a construction. In accordance with Goldberg (1995), I therefore employ a version of the CG architecture which al
45、lows for unification with overrides, as per the Override Principle to be described in Section 2. This interaction is subject to a principle which I will refer to below as the Override Principle. The construction-based model of coercion has the following explanatory features: First, it uses a single
46、combinatory mechanism, the construction, to account for both coerced and syntactically transparent interpretations. Rather than representing a special form of composition, coercion effects are predictable by-products of construction-word combination: they mediate conflicts between the meaning of a c
47、onstruction and the meaning of a superimposed lexical item. This means that the constraint which requires semantic concord between the syntactic sisters in the string a bottle also triggers the coerced interpretation found in a beer. Since this concord constraint is stated for a rule of morphosyntac
48、tic combination, the same construction underlies both strict and enriched composition. Second, it captures head-driven and non-headed (exocentric) coercion effects by means of a single combinatory mechanism. Since combination in unification-based syntax has nothing per se to do with phrase building,
49、 licensing is not the unique domain of syntactic heads. Further, since its combinatory mechanisms are based upon schemas rather than sisterhood relations, Construction Grammar provides a straightforward model of functor-free coercion, as exemplified in (3). Third, it predicts the existence of two sources of coercion effects: type-selecting constructions (e.g., Indefinite Determination) and type-shiftin
