The contribution of frequency, regularity and phonological complexity to the production of verb inflections in Greek aphasia

Presented at the “Language Disorders in Greek” Conference, Patra, Greece, 10–11 June 2006

Alexandra Economou, 1 Athanassios Protopapas,2 Spyridoula Varlokosta,3 & Maria Kakavoulia4
1 Department of Psychology, University of Athens
2 Institute for Language & Speech Processing / Athena
3 Department of Mediterranean Studies, University of the Aegean
4 Department of Communication, Media, and Culture, Panteion University

Background: Verb inflection errors are commonly observed in the speech of different types of aphasic patients. Working models of word production (e.g., Levelt, 1999; Levelt et al., 1999) posit two distinct steps in the production of inflected verbs: selection of diacritical (lexical-semantic) features such as tense, aspect and person, which is a pre-phonological process; and concatenation of the verb stems with affixes, which occurs during phonological encoding. A deficit in the selection of diacritical features is predicted to result in semantically and syntactically inappropriate diacritical features. Failure to apply the appropriate features may result in the affixation of verb inflections being influenced by factors such as ease of lexical access, revealed in effects of frequency of occurrence or familiarity of the affix or verb. Predictions are unclear with respect to verb regularity. In English, greater accuracy in irregular than in regular verbs of similar frequency can be explained by a difficulty in affixation during phonological encoding. In Greek, the interpretation of impaired affixation in terms of regularity is unclear because both regular and irregular verbs are affixed; moreover, regularity may refer to the affix, to aspectual stem formation, or to both.

Aims: The present study examines the level of breakdown in the production of verbal inflection by seven native Greek speaking aphasic persons using a sentence completion task, contrasted with the performance of seven nonaphasic participants matched for age, gender, and education. Because the target productions in such a task are known, it is possible to compare the accuracy in the production of the inflections of different types of verbs.

Method & Procedures: The task employed eight transitive, two-syllable verbs, half of which were regular (in aspectual formation) and the other half irregular, with regularity crossed with word form familiarity. Furthermore, half of the verbs contained a consonant cluster in the stem and half did not; presence of consonant cluster was also crossed with both regularity and familiarity. Using these eight verbs, sentences were constructed to test for agreement, tense, and aspect (see Varlokosta et al., in press). Errors were categorized into (a) errors in morphological suffixation; (b) word form errors; and (c) mixed errors, that is, errors involving both morphological suffixation and word form.

Results: The aphasic participants made more morphological suffixation errors (71% of all errors) than either word form (15% of all errors) or mixed errors (14% of all errors). The majority of word form errors were productions of an incorrect verb. Agreement inflection was the least impaired. The three patients who accounted for almost all of the word form and mixed errors were the most impaired ones in terms of total number of morphological suffixation errors, and showed evidence of agrammatism on picture description tasks. For these patients, word form and mixed errors combined accounted for a large proportion of their total number of errors (38% versus 62% for morphological errors). Analyses of variance showed no differences in overall accuracy of production of agreement, tense, or aspect as a function of regularity, familiarity, or the presence of consonant cluster, with the only exception of significantly more agreement errors for the less familiar verbs. The latter was confirmed by regression analyses showing that verb familiarity was the only variable that predicted the total number of agreement errors.

Conclusions: The predominance of morphological suffixation errors together with the selective nature of the inflectional errors indicate breakdown at the level of selection of diacritical features. Nevertheless, the high percentage of word form and mixed errors for the three most impaired patients likely reflects additional difficulty in lexical selection. These errors are all the more notable given the constraints of a sentence completion task where the verb is provided. Lack of a selective effect of regularity, familiarity, or presence of consonant cluster in the accuracy of the production of most verb morphology indicates that the observed difficulty in suffix selection is not influenced by these factors, with the exception of an effect of familiarity on subject-verb agreement. The findings are thus partly in agreement with those of Faroqi-Shah ∓ Thompson (2004), who found that word form frequency was a significant predictor of the accuracy and substitution errors of verb inflections in agrammatic aphasic participants using a picture description task. The results are discussed in terms of Levelt’s model of word production and theories concerning morphological operations in subject-verb agreement in a richly inflected language.

References:
Faroqi-Shah, Y., & Thompson, C. (2004). Semantic, lexical, and phonological influences on the production of verb inflections in agrammatic aphasia. Brain & Language, 89, 484–498.
Levelt, W. J. M.. (1999). Models of word production. Trends in Cognitive Sciences, 3, 223–232.
Levelt, W. J. M., Roelofs, A., & Meyer, A. S. (1999). A theory of lexical access in speech production. Behavioral and Brain Sciences, 22, 1–75.
Varlokosta, S., Valeonti, N., Kakavoulia, M., Lazaridou, M., Economou, A., & Protopapas, A. (in press). The breakdown of functional categories in Greek aphasia: Evidence from agreement, tense, and aspect. Aphasiology.