Producing and inflecting verbs with different argument structure: Evidence from Greek aphasic speakers

Abstract

Background: Verb argument structure affects language production in aphasia. Unaccusative verbs have been found to be more difficult than transitive or unergative verbs across languages. Transitivity is also a factor with variable influence across aphasia types. Verb inflection production, and in particular tense and aspect, has been found to be vulnerable in aphasia, too. These domains have been studied independently of each other, and their potential interaction has been scarcely addressed.

Aims: The present study explores the effects of argument structure and past tense on verb production in Greek aphasia. Additionally, we address the question of the interrelation between inflection and verb and argument structure production.

Methods & Procedures: Two tasks were administered to 10 aphasic participants (six anomic, two Wernicke’s and two agrammatic Broca’s). The first task (sentence elicitation) consisted of video presentation and description of the video. The second task (tensed sentence elicitation) consisted of presentation of a video preceded by an adverbial prompting for a specific inflection marking. Participants had to produce a sentence with a verb in a specific past form. In both tasks, three verb classes were tested (unergative, unaccusative and transitive verbs with one object) with six items each.

Outcomes & Results: The sentence elicitation task showed that unaccusative verbs are the most difficult class for the aphasic participants whose predominant error was the production of +agentive verbs instead of unaccusatives. The tensed sentence elicitation task showed that while for most patients there was no effect of inflection on verb and verb argument production, a decremental effect was noticed for an agrammatic participant.

Conclusions: We suggest that +/− agentivity ranks above transitivity concerning verb production. To account for the unaccusativity effect, we provide a novel account based on the distinction between thematic and aspectual dimension and argue that unaccusative verbs are more difficult to produce, because they involve a mismatch between the two dimensions of argument structure (thematic and aspectual dimension). Concerning the effect of past tense on verb production we show that such pattern is possible but it does not emerge at the group level. Finally, our group data do not suggest any interrelation between inflection, on the one hand, and verb and argument structure production, on the other hand.

Keywords:

• Argument structure
• Inflection
• Unaccusativity
• Modern Greek
• Aphasia

This work was supported by the I.K.Y. (Greek State Scholarship Foundation) with a grant to the first author for doctoral research. We thank the aphasic and control participants of the study. Moreover, we thank the speech therapists Thalia Zolota, Georgia Kolintza, Flora Koutsi and Erifyli Pouliou and the neurologists Panagiotis Ioannidis and Vasiliki Tsirka for helping us with the data collection. We are also grateful to Roelien Bastiaanse, Roel Jonkers and an anonymous reviewer for suggestions on earlier versions of the manuscript. We also thank Markus Bader for consultation on statistic methods. All errors remain, of course, of our own.

Notes

¹ Notice that some of these assumptions are not totally uncontroversial. Controversial issues are: what is assigned during function assignment (theta-roles or grammatical functions), the possibility of grammatical role shift and the viability of movement operation within this model. Theses controversies are beyond the scope of this article, but see Bock and Levelt (1994) and Thompson and Faroqi-Shah (2002) for a thorough discussion.

² The specific demands imposed by unaccusative verbs to Greek-speaking individuals with aphasia were further confirmed by Peristeri, Tsimpli, and Tsapkini (2013) who employed an online task to assess processing of unaccusatives in aphasia.

³ The unaccusativity effect was reported for the control group, too, for the naming task. A similar effect in the control group is also reported by Stavrakaki et al. (2011).

⁴ With respect to the BDAE, the performance of each patient was compared to that of their age and educational group and, if found more than 2 standard deviations below the mean, it was judged as impaired (cf. Borod, Goodglass, & Kaplan, 1980).

⁵ The target verbs were chosen from an original list of 36 verbs (12 for each verb class matched for length and frequency) on the basis of the success rate of unimpaired younger adult speakers (henceforth pilot control group, which was different from the control group used for comparison with the patients) to produce the target verb (or a semantically equivalent one) and verb argument structure when presented with a video. The cut-off criterion for each video was 90% success of the pilot control group.

⁶ Frequencies were obtained from the Hellenic National Corpus (http://hnc.ilsp.gr).

⁷ The proportions refer to the total number of the errors for each verb class. The absolute number of semantic errors was similar across verb classes, as shown by the statistics.

⁸ Recall that in the first experiment, unaccusative verbs were produced mostly in past perfective. See also Dragoy and Bastiaanse (2013) for analogous findings from Russian, indicating an interaction between aspectual characteristics of the verb and the preferred verb form.

⁹ The two tasks (sentence elicitation task and tensed sentence elicitation task) were administered to all participants within the same session.

¹⁰ As an anonymous reviewer pointed out to us, the lack of a task effect might be due to the fact that the two tasks are not contrastive enough (notice that past forms were produced also in task 1). We investigated it in the following way: we applied a mixed model analysis to find out factors determining the tense produced by the aphasic participants (past vs. non-past). Agentivity, number of arguments, task and aphasia type were put in the model as fixed effects, patient and verb as random effects. We found an effect of agentivity (p = .0001) (not surprisingly, as in the first task the majority of the verbs produced in past tense were unaccusatives) and an effect of task (p = 0). Consequently, we can argue that the two tasks were contrastive.

¹¹ We should acknowledge that these hypotheses have been formulated to explain data mainly from agrammatic populations; however, we think their predictions can be fruitfully extended for other categories of aphasic patients.

¹² It remains an open question whether such an effect of inflection on lemma retrieval/function assignment appears only in pathological speech. Vigliocco, Vinson, Indefrey, Levelt, and Hellwig (2004) provide evidence that under time pressure there is a feedback of the syntactic frame on lexical retrieval in non-pathological speech, as well. In particular, German healthy speakers preserve the gender of the target noun in a continuous picture naming task, even if they produce semantic errors.