Verb representations are closely associated with syntactic constructions in sentence production: Evidence from aphasic patients with short-term memory deficits

• ¹ Rice University, Department of Psychology, United States
• ² University of Maryland, Department of Psychology, United States

Introduction One central issue concerning the mental representation of syntax is the relationship between structural and lexical knowledge. The syntactic priming paradigm offers insights into these issues. Speakers have a tendency to repeat syntactic constructions (the abstract syntactic priming effect; Bock, 1986), and this tendency is even stronger when adjacent utterances share the same main verb (the lexical boost effect; Pickering & Branigan, 1998). A lexicalist account of syntactic encoding argues that processing a prime sentence leaves residual activation of representations at both the structural and verb lemma level (e.g., Pickering & Branigan, 1998). By comparison, an implicit learning account assumes no close connection between structural and lexical representations. This account proposes that speakers implicitly learn the mapping between the message representation and the syntactic structure that expresses the message. The lexical boost effect is an incidental by-product of explicit memory of the sentence (e.g., Chang, Dell & Bock, 2006). Consistent with this account, speakers show persistent syntactic priming and a short-lived lexical boost (Hartsuiker et al., 2008). We hypothesize that if the lexical boost depends on explicit short-term memory (STM), it would not be observed among patients with STM deficits while the syntactic priming effect may be obtained; however, if it reflects automatic activation of structural and lexical representations, aphasic patients would show both syntactic priming and lexical boost effects. Methods and results Six aphasic patients and six age- and education-matched healthy control speakers participated in the study. According to the Quantitative Production Analysis (Saffran, Berndt, & Schwartz, 1989), patients had generally preserved morphological aspects of production but varied considerably on structural aspects (from no deficits to deficits on all measures). Patients also had varying degrees of deficits on semantic and phonological STM measures. In an experimental trial, participants heard and read aloud a transitive prime sentence before describing a transitive picture. We found that patients, with one exception, showed syntactic priming and lexical boost effects within the range of control speakers (Figure 1a-1b), i.e., both patients and control speakers were more likely to produce active sentences after active primes than after passive primes, and this tendency was stronger when the prime and the target sentences shared the same verb. Individual patient syntactic priming effects were not correlated with any of the structural measures from the QPA (ps > .12). More importantly, patients’ lexical boost effects were not correlated with semantic/phonological STM capacities (ps > .82) (Figure 1c-1d). Conclusion Our findings suggest that the lexical boost effect does not rely on temporary maintenance of the wording in the prime sentence. Instead, processing a prime sentence activates both the verb lemma and the sentence structure, as well as the connections between the two levels of representations, though different decay rates may be associated with activations at different levels. The residual activation of the verb-structure association automatically facilitates the subsequent reuse of the same sentence structure with the same verb. The evidence supports a lexicalized system of syntactic representations in language production in which verb lemmas are closely related to structures.