Corpus-based Transitivity Biases in Individuals with Aphasia

• ¹ University of Arizona, United States

Introduction Spontaneous speech samples in individuals with aphasia (IWA) have been analyzed to examine many different psycholinguistic features. The present study focused on how IWA use verbs in spontaneous speech. Some verbs can occur in more than one argument structure, but are biased to occur more frequently in one frame than another. For example, "watch" appears in transitive and intransitive structures, but is usually used transitively. This is known as a transitivity bias. It is unknown whether IWA show the same transitivity biases in production as those reported in previous corpus studies with unimpaired individuals. Studies of sentence comprehension have shown that IWA are sensitive to verb biases (e.g., DeDe, 2013). In addition, IWA have shown an overall preference for transitive structures, which are the most frequent structures in English (Roland, Dick, & Elman, 2007). The present study investigated whether IWA show the same pattern of transitive and intransitive biases in spontaneous speech as unimpaired individuals. Method Participants: 278 interviews with IWA were taken from AphasiaBank. The IWA represented a range of aphasia types. Participants were omitted if they spoke English as a second language. Materials: 54 verbs were coded. We chose verbs with the goal of representing different bias types (e.g., transitive, intransitive, sentential complement). Of these, data from 11 transitively biased and 11 intransitively biased verbs (matched for frequency of use and number of syllables) are presented here. Coding: All productions of the 54 verbs were coded. The coding protocol was based on Gahl, Jurafsky, and Roland (2004). We implemented an additional level of coding to indicate erroneous verb productions, such as ungrammatical structures and verb agreement errors. Results The (in)transitivity biases for IWA were compared to biases from a previously published corpus study (Gahl et al., 2004). The IWA used transitively biased verbs in transitive structures 43.5% of the time and in intransitive structures 14.5% of the time [controls: 50.2% transitive and 9.3% intransitive productions]. The IWA produced intransitively biased verbs in intransitive structures 54.7% of the time and in transitive structures 3.8% of the time [controls: 75.2% intransitive and 4.9% transitive uses]. Next, consider the relative frequency of transitive and intransitive constructions. The IWA used intransitive verbs in 62.4% of their overall productions and transitive verbs in 37.7% of overall productions. In Gahl et al. (2004), control populations produced intransitive verbs in 44.5% of their overall productions and transitive verbs in 55.5% of their productions. Discussion Our results showed that IWA retain typical transitivity and intransitivity biases in their productive uses of a selected sample of verbs. In contrast to controls, IWA produced intransitive structures more often than transitive structures. This trend is consistent with studies showing that IWA tend to produce verbs with simpler argument structures (e.g., Kim & Thompson, 2004). However, our sample of verbs may be more likely to be produced intransitively in response to the AphasiaBank prompts. We are currently coding control data from AphasiaBank, to permit direct comparison of verb usage and types of structures across groups.