The term 'mismatch' has been used to describe a number of linguistic phenomena involving mappings between (apparently) incongruent elements or structures, where incongruity is defined relative to some typical or default condition. There might, for example, be a mismatch in the number of elements involved at different levels of representation. Jackendoff (1987) argues that a single syntactic NP may have multiple thematic roles, even though normally the relationship between thematic roles and NPs is one to one. To take a somewhat more complex example, Farrell Ackerman (this volume) discusses cases from several languages in which a single morphosyntactic feature set is realized by multiple elements in the syntax, even though normally such a set would be realized by a single syntactic unit. Another common type of mismatch is a category mismatch, in which there is a violation of the default correspondence between categories at different levels of representation. For example, Francis (1999) argues that predicate nominals in English (e.g., a cat in Max is a cat) may be described as involving a mismatch between syntax and semantics. Although typically a syntactic noun phrase corresponds to a semantic argument, in the case of predicate nominals the post-verbal noun phrase corresponds to a semantic predicate.
The notion of mismatch has been especially important for parallel-architecture constraint-based theories of grammar such as LFG, HPSG, Construction Grammar, Autolexical Syntax, RRG, and Representational Modularity. In particular, mismatch phenomena help to justify the distinctions between different levels of grammatical representation. If there is an apparent mismatch between elements or categories at two different levels of grammar, this may be interpreted as evidence that the two levels are distinct. In some cases, mismatch is also taken as evidence against a Chomskyan derivational representation and in favor of a parallel representation. For example, Alsina (this volume) discusses V-NP idioms in Chichewa as examples in which the NP clearly bears its own grammatical function (as shown by its ability to passivize), but unlike ordinary NPs it is not an independent semantic argument and does not bear its own thematic role. Alsina’s data point toward a model of grammar in which grammatical function and thematic structure are represented independently of each other and in parallel. The data would be a problem, however, for a derivational analysis in which both thematic roles and grammatical functions are conflated with constituent structure, for there would be no straightforward way of representing the distinct properties of each type of grammatical information. Similarly, Sadock (this volume) presents sentences involving ‘raising’ verbs such as seem as examples of mismatch between syntax and semantics, where the usual association between the syntactic subject and the semantic argument of the main verb is violated. Sadock takes this kind of mismatch as evidence that semantics and syntax should be represented independently and in parallel. Unlike a derivational ‘raising’ analysis, a parallel representation of syntax-semantics mismatch requires no extra levels of syntactic structure (d-structure), no empty categories, and no movement.
At the same time that mismatch phenomena help to justify parallel-architecture models of grammar in opposition to derivational models, they also pose a number of difficult questions. First of all, what exactly constitutes an incongruity or mismatch? Given that many phenomena identified as mismatch are a common and pervasive part of language, how can we decide whether or in what sense a particular mapping is actually incongruent? This question is especially difficult when dealing with cross-linguistic differences in mapping relationships, since default correspondences between elements may differ from language to language. Working within the theory of LFG, Bresnan (1995) presents evidence from English and Chichewa that constructions which are semantically and functionally similar across languages may have distinct syntactic (c-structure) categorizations. For example, she compares that-complements in English with kuti-clauses in Chichewa (1995: 34-40). Although the semantic and functional properties of the two kinds of complements are virtually identical, Bresnan shows that English that-complements are categorized as CP while Chichewa kuti-clauses are categorized as NP. In such a case, which syntax-semantics mappings, if any, should be considered mismatch? Can a mapping that is perfectly normal for a particular language still be deemed a mismatch in some universal sense?
Even assuming that we can clearly identify a mismatch when we see one, the implications of mismatch phenomena remain controversial. As discussed above, evidence of mismatch is often used to justify different levels of grammatical representation. But practitioners of different grammatical theories and even practitioners of the same theory often disagree as to which levels are necessary. The levels of syntax, for example, differ from theory to theory and even within theories. GB theory (Chomsky 1981) distinguishes d-structure, s-structure, and LF (logical form), but has no distinct level of syntax for representing grammatical functions such as subject and object. In contrast, LFG distinguishes between c-structure (phrase structure) and f-structure (functional structure), but has no (syntactic) level of LF and no d-structure (Horrocks 1987). Autolexical Syntax (Sadock 1991) has only one level of syntax, which corresponds to c-structure in LFG. And in HPSG, there is theory-internal disagreement as to whether valence lists are a part of argument structure, or whether argument structure and valence are separate levels of syntax (Manning and Sag 1998).
In theoretical discussions of the differences among these various levels
of syntax, mismatch phenomena are usually involved either explicitly or
implicitly. However, the mismatch phenomena themselves do not always
tell us conclusively whether the levels in question are actually syntactic.
For example, in GB theory (Chomsky 1981), the special argument structure
properties of verbs like seem (as in He seems to like it)
help justify a distinction between the two levels of phrase structure called
d-structure and s-structure. The basic argument structure is represented
in d-structure, while the actual word order is represented in s-structure.
The two levels are claimed to be related by a movement operation of 'raising',
whereby the subject of the embedded clause in d-structure moves up into
the subject position of the matrix clause in s-structure. In
LFG, the same argument-structure properties of seem are represented in terms
of functional (f-structure) relations which are independent of phrase structure
(Horrocks 1987). In this case, no movement is involved, and the
two levels of syntax are more independent of each other than in
GB theory. Although GB and LFG are rather different in their architecture,
the unusual relationship between the argument structure of seem and the word
order of a sentence like He seems to like it may be similarly used in each
theory to distinguish between two different levels of syntax (s-structure
and d-structure, or c-structure and f-structure). But there is yet
another way of viewing the same phenomenon. As Sadock (this volume)
argues, verbs like seem may just as easily justify the distinction
between syntactic (phrase) structure and semantic structure, but say nothing
about any additional levels of syntax. This is because for Sadock,
the relevant facts regarding the argument structure of raising verbs are
captured in semantics. Thus, the relationship between mismatch phenomena
and the distinctions among levels of grammatical representation is far
from straightforward, and continues to be debated among theorists.
The first two papers by Robert Malouf and Jerrold Sadock deal with issues about the architecture of grammar from a big-picture perspective. Malouf (this volume) presents a new model of constraint interactions designed to specify the conditions under which mismatch is possible. Incorporating insights of OT proposals, Malouf conceives of mismatch in terms of conflicts between constraints. In contrast to OT proposals, however, he takes a constructional view of constraints. According to Malouf, the inventory of constructions is a set of mutually cooperating constraints, and conflicts between constraints may be resolved most parsimoniously at the level of the construction, rather than at the level of the individual linguistic form. He argues that this view of constraint interactions avoids the psychologically implausible aspects of OT, such as the evaluation of an infinite set of candidate forms, while still capturing the important insights of the theory.
Sadock (this volume) also offers an original approach to mismatch within his multi-modular theory of grammar, commonly known as Autolexical Syntax. According to Sadock, syntax, semantics, and morphology are represented independently and in parallel, and mismatches are represented in a simple, straightforward manner through imperfect correspondences across modules. He argues that important principles of grammar, such as the Theta Criterion and the Case Filter, need not be stipulated in a multi-modular theory, but rather automatically follow from a modular architecture. Other principles, such as conditions on traces and chain formation, turn out to be completely unnecessary in a multi-modular framework. Finally, new principles of grammar designed to capture general constraints on mismatch, such as the Generalized Interface Constraint, take the place of a number of less general principles developed in derivational theories.
The second two papers by Farrell Ackerman and Alex Alsina deal more specifically with the issue of mismatch in complex predicates. Ackerman discusses cases of morphology-syntax mismatch in which a single morphosyntactic feature set is realized on more than one syntactic element of the predicate, for example the case of the Latin periphrastic perfective passive. Such cases may be contrasted with the default case, in which such a morphosyntactic feature set is realized on a single syntactic head. He focuses in particular on preverb-verb constructions in Uralic and Caucasian languages which display both a morphology-syntax mismatch and a morphology-semantics mismatch whereby inflections on the verb (or otherwise within the predicate) may have semantic scope over the entire preverb-verb complex. He argues that the same kinds of morphosyntactic processes apply in both the default cases and in cases of morphology-syntax mismatch, concluding that complex-predicate formation is actually a kind of lexeme formation that applies independently of syntactic constituent structure.
Alsina (this volume) analyzes three distinct kinds of complex predicates by identifying different levels at which the complexity is located. By definition, all complex predicates involve some kind of mismatch in which a single predicate maps onto two or more elements at some level. The first kind of complex predicate that Alsina discusses, the m-complex predicate, involves a single predicate which maps onto two or more morphemes of the same word, as illustrated by reciprocal verb forms in Chichewa. Similarly, the second kind of complex predicate involves a single predicate which maps onto two or more distinct words. Alsina illustrates this kind of predicate with the example of reflexives in Chi-Mwi:ni, which have an argument structure similar to reciprocals in Chichewa, but in this case the reflexive is a (non-function-bearing) NP rather than a bound morpheme. Finally, Alsina discusses f-complex predicates, which involve a single predicate which maps onto two different f-structures (and two different words). He illustrates this type of complex predicate with the example of verb-NP idioms in Chichewa, in which the NP bears its own grammatical function but is still clearly part of the verbal predicate.
Finally, de Swart and Michaelis discuss the mismatch phenomenon of aspectual coercion. Importantly, aspectual coercion appears to differ from the kinds of mismatch discussed so far. Rather than involving a mismatch in the number or constituency of elements at different levels of grammar, aspectual coercion involves a mismatch within the semantics. Specifically, the semantic properties of an aspectual operator (such as progressive -ing in English) conflict with the semantic properties of the sentence in which the operator is used, but the sentence is interpretable anyway. For example, even though the verb flash denotes a punctual event, we can assign an acceptable meaning to the progressive sentence The light is flashing by interpreting the flashing as iterative.
In her paper, de Swart (this volume) explores cross-linguistic variation in the division of labor between the semantic contribution of aspectual operators and that of coercion. As specified in the Discourse Representation framework developed in earlier work (de Swart 1998, de Swart and Molendijk 1999), any mismatch between the semantics of the proposition and the semantics of the aspectual operator triggers the insertion of a coercion operator, which repairs the mismatch and allows the sentence to be assigned an interpretation. In this paper, de Swart discusses examples of coercion in English, French, and Dutch which show cross-linguistic variation in the kinds of semantic information encoded in aspectual operators, and in the kinds of aspectual coercion that are possible in each language. She argues that the inventory of aspectual meanings is like a toolbox which different languages may use in different ways. In any given language, some meanings may be grammaticalized as aspect markers, some meanings may be allowable through coercion, and some meanings may be entirely disallowed. De Swart explores the very interesting consequences of this cross-linguistic variation for the representation of sentence-level semantics and discourse.
Michaelis (this volume) argues in favor of a construction-based approach to coercion. In contrast to the approach of de Swart (this volume, 1998), Jackendoff (1997), and others, Michaelis conceives of coercion not just as a case of mismatch between lexical semantic and compositional semantic properties. Rather, aspectual coercion is viewed as a mismatch between the semantic specifications of two conventional symbolic constructions, and therefore necessarily involves syntax as well as semantics. Thus, aspectual coercion on this view may not be so different from other kinds of mismatch involving multiple levels of grammar. Unlike other approaches to coercion, this approach requires no special mechanisms such as coercion operators to deal with coercion effects. In much the way that Ackerman argues that a single kind of morphosyntactic process underlies both ordinary lexeme formation and complex-predicate formation, Michaelis argues that a single kind of morphosyntactic combinatory process underlies both coercion and ordinary instantiation. Thus, Michaelis presents a viable alternative to purely semantic approaches to coercion.
These six papers show clearly that mismatch phenomena bring up very
interesting and challenging problems for grammatical theory. In particular,
mismatch phenomena often defy principles of grammar designed to account
for the ordinary default cases, and as a consequence they challenge us
to develop new and better theories of grammar. Mismatch phenomena
have been crucial, for example, in the development of constraint-based
parallel architecture theories. They provide important evidence for
distinctions among different levels of grammar, for the parallel representation
of these distinctions, and for the existence of conflicting constraints.
However, as these workshop papers have shown, the specific theoretical
implications of mismatch phenomena are not yet fully known, and still a
matter of lively and productive debate. This workshop has, we hope,
furthered the debate and provided interesting insights and useful connections
among different approaches to the diverse range of linguistic phenomena
called mismatch.
Ackerman, Farrell and Gert Webelhuth. 1998. A theory of predicates. Stanford: CSLI Publications.
Alsina, Alex. 1993. Predicate composition: a theory of syntactic function alternations. PhD thesis, Stanford University.
Baker, Mark. 1996. Unaccusativity and the adjective/verb distinction: English and Mohawk Evidence. Proceedings of the Eastern States Conference on Linguistics 13, 12-35.
Bresnan, Joan. 1995. Category mismatches. In Akinbiyi Akinlabi (ed.) Theoretical approaches to African linguistics, 19-45. Trenton, NJ: Africa World Press.
Bresnan, Joan. 1997. Mixed categories as head sharing constructions. In Miriam Butt and Tracy Holloway King (eds.) Proceedings of the LFG ’97 conference. Stanford: CSLI Publications.
Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris.
Creider, Chet and Richard Hudson. 1999. Inflectional morphology in Word Grammar. Lingua 107/3-4, 163-187.
Culicover, Peter and Ray Jackendoff. 1997. Semantic subordination despite syntactic coordination. Linguistic Inquiry 28, 195-217.
Francis, Elaine J. 1999. Variation within lexical categories. PhD thesis, University of Chicago.
Horrocks, Geoffrey. 1987. Generative grammar. London: Longman.
Malouf, Robert. 1998. Mixed categories in the hierarchical lexicon. PhD thesis, Stanford University.
Manning, Christopher and Ivan Sag. 1998. Dissociations between argument structure and grammatical relations. In Gert Webelhuth, Jean-Pierre Koenig, and Andreas Kathol (eds.) Lexical and constructional aspects of linguistic explanation, 63-78. Stanford: CSLI Publications.
Pustejovsky, James. 1996. Aspectual coercion and logical polysemy. In Boguraev Branimir (ed.) Lexical Semantics: the problem of polysemy, 133-162. Oxford: Clarendon Press.
Sadock, Jerrold M. 1991. Autolexical Syntax. Chicago: University of Chicago Press.
de Swart, Henriette. 1998. Aspect shift and coercion. Natural Language and Linguistic Theory, 16/2, 347-85.
de Swart, Henriette and Arie Molendijk. 1999. Negation and the temporal
structure of narrative discourse. Journal of Semantics, 16/1, 1-42.