The Selective Properties of Verbs in Reflexive Constructions

This paper investigates the relationship between verbs and reflexive markers within reflexive constructions, setting forth the hypothesis that the verb plays a determining role in anaphoric binding. Building upon Dalrymple's (1993) argument that binding constraints are lexically specified by anaphoric elements I demonstrate reflexive requirements can be lexically specified for distinct groups of verbs, an approach which offers another level of descriptive clarity to theories of anaphoric binding and introduces a means of predicting reflexive selection in domains where syntactic constraints do not readily apply. This is shown to be particularly pertinent in languages with more than one reflexive type that have overlapping syntactic binding domains.

Much of the previous work on the grammar of reflexive constructions has generally been concerned with nominal types. Regardless of whether the approach has been one of invariant universal constraints, language-by-language parameterization, or lexically specified antecedent requirements, the antecedent and the anaphor have been repeatedly recognized as the most important elements within a reflexive construction. However, instances in which reflexive constructions demonstrate identical syntactic environments but result in different requirements on the reflexive have been observed in many of the world's languages. This is illustrated in Dutch below:

The verbs in (1) are all used transitively and, thus, do not result in differing syntactic environments. Equally, all three sentences share the f-structure illustrated in (2):

[ PRED 'verb<SUBJ, OBJ>'
  SUBJ [ PRED 'R-EXPRESSION' ]
  OBJ  [ PRED 'PRO'
         PRONTYPE REFL ] ]

The lexical forms (Bresnan 1980) of the Dutch verbs presented in (1) are also identical:

Since the verbs in these sentences are all transitive, they require both a subject and an object; however, the direct objects that the Dutch sentences permit in (1) are realized differently for each of the verbs illustrated. Dutch is a language with two reflexive types and it is on the selection of the acceptable reflexive type, in particular, that these sentences differ. Unlike the verb wast, 'to wash', which can be paired with either of the two Dutch reflexive types, the verbs haat, 'to hate', and gedraagt, 'to behave', can only be paired with one of the two available reflexive types. Moreover, gedraagt and haat require opposite reflexive types: the verb haat requires the complex reflexive whilst the verb gedraagt can only be used with the simple reflexive zich.

Taking such observations into account and incorporating data from five typologically distinct languages (English, Dutch, French, Russian, and Fijian), I put forward the following three arguments:

1. The verb plays a determining role in reflexive selection:
   • a. In languages with multiple reflexive strategies the verb can select for a reflexive strategy according to the referential range of the reflexive type and the semantic transitivity of the verb.
   • b. In languages with a single reflexive type the verb can influence the referential range of a reflexive with respect to its antecedent.
2. Reflexive constructions are determined over four different components of linguistic representation: c-structure, f-structure, lexical structure, and semantic structure.
3. Languages with multiple reflexive strategies can be described as two overlapping but different types: those that are primarily syntactically motivated and those that are primarily semantically motivated.

Relating to, and proceeding from these arguments, the following inferences are reached:

1. Binding constraints are lexically specified.
2. Reflexive requirements are lexically specified by the verb.
3. Anaphoric binding conditions are both syntactically and semantically determined.

Following Sells et al. (1987), I suggest that reflexive constructions are definable over four different components of linguistic representation and adopt a quadripartite linguistic analysis that incorporates c- structure, f-structure, lexical structure, and semantic structure within a Lexical Functional Grammar theoretical framework. The level of semantic structure is found to be particularly interesting since the realization of a reflexive construction is shown to be influenced by differing semantic requirements between verbs and reflexives. On the basis of several semantic tests, I find that verbs in reflexive constructions have two different predicate structure types, 'transitive' and 'intransitive', and reflexive markers have three different possible internal semantic structures, 'strict identity' (x,x), 'sensory-perceptual near identity' (x,f(x)), and 'representational near identity' (x,y).

The syntactic, semantic, and lexical characteristics of the reflexives and verbs analyzed over the data set presented in this work result in the identification of eight different reflexive/verb types and the establishment of two implicational relationships:

1. Reflexive markers in lexically intransitive reflexive constructions have no semantic content.
2. Verbs that take a reflexive argument with a strict (x,x) or close (x,f(x)) internal structure must be intransitive at the semantic component of linguistic structure.

These results contribute to our understanding of anaphoric binding theory, directed verb categories, the syntax-semantics interface, and the licensing of multiple reflexive types within a given language.

References:

• Dalrymple, Mary. 1993. The Syntax of Anaphoric Binding. Stanford, CA: CSLI.
• Sells, Peter; Annie Zaenen, and Draga Zec. 1987. Reflexivization Variation: Relations between Syntax, Semantics, and Lexical Structure. In M. Iida, S. Wechsler, and D. Zec, eds., Working Papers in Grammatical Theory and Discourse Strucutre, 169-238. Stanford, CA: CSLI.