Linguistics Colloquium: Yuval Katz

02/11/2021 - 14:00 - 15:30

Yuval Katz, Tel Aviv University

Title: Verbal alternations in language impairment: from practice to theory


Although there is a growing body of research in cognitive neuropsychology that takes theoretical linguistics into account and vice versa, it is still the case that the two fields have very different views of how language works. Architectural models from cognitive neuropsychology of language tend to focus on a central theme, lexical retrieval, the process of getting from a concept to a spoken word. Accordingly, assessment of lexical retrieval abilities in the lab and in the clinic is usually carried out by tasks that examine the production of single words based on pictures, definitions or reading aloud of written words. However, as every linguist knows, words in the wild rarely appear in isolation. The sound and meaning of words may be affected by their role and their position in a sentence. Architectural models from theoretical linguistics, on the other hand, deal with linguistic units of all sizes, but rarely take findings from cognitive neuropsychology into account. In this study of verbal alternations in language impairments we position ourselves at the interface between cognitive neuropsychology and theoretical linguistics: We design linguistically informed experimental material, test individuals with a wide variety of language impairments, and construct a neuropsychological model, not confined to single words, based on patients' diagnosis and their performance in the tasks. Finally, we discuss how the results and the neuropsychological model can constrain the set of possible models of the same phenomena in a different discipline – theoretical linguistics.


We tested 34 Hebrew speaking individuals with acquired and developmental language impairments with a test battery consisting of seven tasks designed to elicit the production of verbs: Each patient was also tested with a variety of other task in order to determine their exact functional loci of impairment in the production process. The test battery included alternating verbs (transitive-unaccusative, transitive-unergative, experiencer and reflexive alternations), which, in Hebrew, share a root but are usually differentially marked by a morphological pattern (e.g. nisgar - 'closeINTR' ↔ sagar - 'closeTR'; rakad - 'dance' ↔ hirkid - 'make-dance'), in all possible pattern combinations. We also tested non-alternating verbs, which in Hebrew are also marked by a morphological pattern, including what we call pseudo-alternating verbs, which are pairs of non-alternating verbs in different morphological patterns that share a root, but are not alternating. Some of those verbs may share some meaning component (maca – 'find' ↔ himci – 'invent'), whereas some do not (cava – 'paint, color' ↔ hicbi'a – 'vote, raise-hand').


We report the following error patterns, based on which we constructed a neuropsychological model: AI, a patient with an impairment in the conceptual system (low performance in non-verbal as well as verbal tasks) who has difficulty understanding "who did what to whom", has selective errors with alternating verbs. He selects the incorrect alternant and substitutes the roles in the event in many occasions, but has full knowledge of the verb's lexical-syntactic information, as evident by his almost complete lack of argument structure errors. For example, in the sentence completion task, AI used the intransitive alternant (ka'as – 'was angry') instead of the transitive alternant (hix'is – 'anger') and substituted the thematic roles (instead of the subject of the sentence being angry, someone is angry at the subject of the sentence). However, he used the correct preposition for the intransitive alternant (al – 'on'). Patient SN (as well as three other patients) has an impairment in the syntactic lexicon, which also results in incorrect selection of alternants. However, contrary to AI, since SN's representations in the syntactic lexicon are impaired, he not only selects incorrect alternants but also produces sentences with argument structure errors. For example, in a sentence completion task, when asked to complete the sentence (Ha-delet ___ ba-laila – 'the door ___ at night), his response was sagra ('closedINTR'), which the incorrect alternant, yielding an ungrammatical sentence due to argument structure. Patient RR has an impairment in the phonological output lexicon, and specifically in knowing which morphological patterns are compatible with which roots. This causes him to select incorrect alternants (e.g., hirkid – 'danceTR', 'make someone dance'  rakad – 'danceINTR'), but crucially, since his impairment is in a morpho-phonological post-syntactic stage, he also selects the wrong morphological pattern in non-alternating verbs (e.g., kodxim – 'drill'  *mekadxim) and pseudo-alternating verbs. Patient HH also has an impairment in the phonological output lexicon, but his impairment is manifested in errors in roots only, with no errors in patterns with any type of verb. Patient SH (and 15 other patients) has an impairment in the phonological output buffer, which stores pre-assembled morphemes and assembles them with the root before production. SH showed similar errors as RR in alternating and non-alternating verbs, but since her impairment is in a post-lexical output stage, and not in the representation of idiosyncratic morphological information, she also showed many errors in regular inflectional morphology, namely, in tense and agreement (e.g., da'ag – 'worryPAST' --< do'eg - 'worryPRESENT'). We propose a neuropsychological for the production of alternating verbs within a sentence, and discuss its possible implications for linguistic theory.





Subscribe to our Telegram channel to get notified about upcoming talks and events