Valencies versus argument structure constructions: quantitative corpus studies on the distribution of argument structure patterns of psych-verbs
Institut für Deutsche Sprache, Mannheim
Thursday 21 March 2013, 14.15-16.00, PAM 10
Psych-verbs exhibit a fairly large variation of argument structure patterns, alternating between stimuli and experiencers in subject position, between nominal and sentential realizations of arguments, and between explicit and implicit arguments. Therefore, they lend themselves to an investigation of the quantitative distribution of argument structure patterns that is expected to contribute to the current debate about the adequacy of valency-based versus construction-based approaches to argument structure.
The data presented are based on two types of corpus studies. (i) “Verbprofile Analyses” are based on corpus samples of sentences for each of the psych-verbs investigated; the samples are processed by assigning each sentence to the argument realization pattern it represents. The number of sentences for each argument realization pattern of a verb are counted and subjected to statistical analyses (cluster analysis, association measures); three studies of this types have been carried out so far: German psych-verbs in newspaper texts, German psych-verbs in corpora representing different registers, and German psych-verbs and their Romanian counterparts in newspaper texts. (ii) In “Exhaustive Clause Pattern Analyses” argument structure patterns are embedded into partial clauses in which the argument structure pattern is known to occur frequently. The clause patterns are searched for in the entire corpus and the frequency data statistically analyzed.
The data show that three types of factors determine the quantitative distribution of argument structures: (i) verb-independent linguistic factors, among them stylistic properties of text genre, requirements from information structure, functional dependencies between different argument structure patterns, and cross-linguistic conceptual-semantic properties of verbs; (ii) general properties of the cognitive system, in particular of the memory system, which account for the strong entrenchment of a handful of verbs in each argument structure pattern; (iii) idiosyncratic preferences of verbs or verb classes for particular argument structure patterns.
While the last factor points to the existence of valency-based argument structure representations, there is also evidence for construction-driven behavior. In particular, argument structures of small or medium frequency all show similar frequency distribution patterns with respect to the respective verbs they occur with; this reveals how constructions are entrenched with particular verbs and slowly spread over other parts of the verb lexicon. Thus, the data give rise to a discussion of the interplay of valency and argument structure constructions.