Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
In press in Brain and Language Generalizations on variations in comprehension and production: A further source of variation and a possible account Naama Friedmann, Tel Aviv University, [email protected] Many types of variation can be found in agrammatic aphasia. Drai and Grodzinsky (in press) presented a convincing account for variations between syntactic structures, between individuals and across languages. However, there is one source of variation that is not accounted for by their approach: the variation that exists within individuals with agrammatism between different structures that are derived by movement. Such variation can be found in comprehension mainly between passive sentences and object relative clauses. Drai and Grodzinsky show that whether or not a sentence is derived by movement of a noun phrase is a critical factor in whether individuals with agrammatism understand it or not. However, both passive sentences and object relatives involve movement of noun phrases, and yet for some individuals, the comprehension of object relatives is impaired (showing chance performance) whereas the comprehension of passives is unimpaired or above chance. Some other (maybe most) individuals with agrammatism are impaired in both passives and object relatives. Interestingly, individuals who are above chance in object relatives but at chance in passives are not reported. This dissociation can be seen in several studies, for example, Sherman and Schweickert (1989) reported 5 individuals with agrammatism, two of them were above chance in the comprehension of reversible passives, but at-chance on reversible center-embedding object relatives. Berndt, Mitchum and Wayland (1997) reported 5 individuals with agrammatism, 3 of them were above chance on passives, and at chance on object relatives. This pattern can also be seen in studies that were not interested in the comparison of passives and object relatives. For example, Ballard and Thompson (1999) report the baseline comprehension results of 5 individuals with agrammatism prior to treatment. Two of them, participants 1 and 4, were above chance on passives and at chance on object relatives. The same was found in the baseline performance of two of the four participants reported by Thompson, Shapiro, Kiran and Sobecks (2003), RM and DL: they had above chance performance on passives, and chance performance on object relatives. Three of the participants reported in Love and Oster (2003) were above chance on passives and at chance on object relatives (only for one of them, BJ, object relatives significantly differed from the passives). 1 The rest of the participants in all these studied were at chance on both passives and object relatives (except for one participant in Berndt et al. who was above chance on both). Finally, Piñango (1999) reported an individual, RD, who showed a dissociation of a similar nature: he was above chance in the comprehension of alternating unaccusatives (which include movement of the subject from object to subject position) and at chance on object relatives. What can account for such pattern? It might have been a sheer statistical coincidence, but the clear asymmetric pattern of passives better than object relatives seems to call for a principled treatment, similar to the treatment Drai and Grodzinsky gave other types of variations. What we need is an account that will be able to account both for the dissociation found between passives and object relatives, as well as for the asymmetrical pattern of this dissociation: either both passives and object relatives impaired, or passives unimpaired and object relatives impaired, but not the opposite. One direction for such account is to try and apply the Tree Pruning Hypothesis, which was originally suggested for agrammatic production, also to agrammatic comprehension. The Tree Pruning Hypothesis proved useful in accounting for variation between individuals with agrammatism who show different degrees of severity in production (Friedmann, 2001, 2005, in press). Similarly to the pattern in comprehension, variation within individuals was found in production: Some showed a dissociation in production between good tense inflection and impaired production of questions and embedding, and some were impaired in both. In a study of 19 Hebrew- and Palestinian Arabic-speaking individuals with agrammatism, three distinct patterns of performance were found. Five individuals had a relatively mild deficit, which impaired their ability to produce Wh questions, embedded sentences and verb movement to second position, but not their ability to inflect verbs for tense and agreement; 13 were more severely impaired, and their deficit extended to both their ability to produce Wh questions, embedded sentences and V-C movement, and to inflect verbs for tense; One participant, who was in the early stages of her recovery, was even more severely impaired, and for her not only the abilities to produce Wh questions, embedded sentences, V-C movement, and tense inflection were impaired, but also agreement inflection. 2 I suggested that these differences in production can be accounted for by the height on the syntactic tree that a patient can access: the higher the patient can climb on the tree, the milder the impairment. A few words about the syntactic tree are necessary here before we proceed to the account. According to syntactic theories within the generative tradition (e.g., Chomsky, 1995), sentences are represented as phrase markers or syntactic trees (see Figure 1 and ignore for now the arches on the tree). These syntactic trees include functional nodes, which are responsible for the syntactic structure of the sentence. These functional nodes include, among others, inflectional nodes: an agreement phrase (AgrsP), the lowest functional node, which represents agreement between the subject and the verb in person, gender and number, and a tense phrase (TP) above it, representing tense inflection of the verb. Finite verbs move from V0, their base-generated position within the VP, to Agr0 and then to T0 in order to check (or collect) their inflection. The highest phrasal node in the tree is the Complementizer Phrase (CP). This node hosts complementizers such as “that”, Wh morphemes like “who” and “what” that moved from the base-generated position within the VP, and the auxiliary in yes/no questions or the verb, which in some languages moves to second position in the sentence. Thus, the ability to correctly inflect verbs for agreement and tense crucially depends on the AgrP and TP nodes respectively and the ability to move the verb to these nodes; The construction of embedded sentences, Wh-questions and movement of verbs to second position depends on the CP node being intact and accessible. Crucially, the nodes in the syntactic tree are hierarchically ordered – the lowest node is the Verb Phrase, the nodes above it are the Agreement Phrase and the Tense Phrase (in this order according to Pollock, 1989), and the Complementizer Phrase is placed at the highest point of the syntactic tree. Now, if we look at this theoretical construct, the syntactic tree, as a Guttman scale (Guttman, 1944), the hierarchy of functional nodes defines what are possible and impossible patterns of agrammatic deficit in production. It defines what can be a possible agrammatic deficit, and derives the three patterns of impairment in production that we found, and, perhaps more importantly, it defines patterns of agrammatism that are logically possible but that are not expected according to this hierarchy. 3 If access to a high node entails access to all nodes below it, and a deficit in a low node implies a deficit in all nodes above it, then we can explain the three patterns of deficit in production (see Figure 1) by postulating a single principle that distinguishes them from one another, the level in the syntactic tree at which the pruning occurs. The five individuals with impairment in Wh questions and embedding are only impaired at CP, but can access the nodes below it. The 13 more severely impaired individuals could not access TP, and therefore could not inflect verbs for tense and also could not access any higher node. The inaccessibility of the higher node resulted in an impairment also in structures that relate to the higher node CP. Finally, the most severely impaired individual could not even access the AgrP node, and therefore could not inflect verbs correctly for agreement, and also could not access the higher nodes TP and CP. Given that inaccessibility to a certain node prevents access to higher nodes, we do not expect to find individuals with impaired functions that relate to TP, but with intact CP functions such as Wh question production, and production of CP-embedded sentences. Similarly, if agreement-related structures are impaired, we expect both TP and CP to be impaired. And indeed, no agrammatic participant showed an impairment in TP without an impairment in CP, or impairment in AgrP without an impairment in TP and CP. CP Milder: 5 individuals C' (Wh-question) C TP (complementizer) Severe: 13 individuals T' T AgrP (tense) Agr' VP Agr (agreement) Severest: 1 individual V' NP V NP Figure 1. Degrees of agrammatic severity in production on the syntactic tree. 4 Returning to the variation in comprehension, a possible way to account for the within-subject variation of impaired object relatives and above-chance performance on passives is to relate the performance in comprehension to the position of each sentence type on the tree. The highest node that passive sentences, as well as SV sentences with unaccusatives, require is the TP node (termed IP by some analyses that do not split Agr and T). Relative clauses, on the other hand, require CP, and actually build a tree even above the CP of the embedded clause.1 If a limitation in access to a node on the tree holds not only for production but also for comprehension, the difference between individuals who understand neither object relatives nor passives, and individuals who fail to understand object relatives but who understand passives (and unaccusatives) will be how high they can get on the tree. Those who can access TP but not CP would understand passives but not object relatives; those who cannot even access TP, and therefore also not CP, would not understand both types of sentences. This also explains why the dissociation goes only this way: if passives are impaired it usually means that TP is impaired (morphological deficits that are not agrammatic aside), and this requires CP to also be impaired, and therefore causes a failure in object relatives as well. So it seems that this additional source of variation in comprehension can also be accounted for. This opens several questions - can Tree Pruning account in general for the deficit in comprehension as well as in production? Should we assume tree pruning in comprehension in addition to a deficit in movement? And finally - is there a relation between variation in comprehension and production – Namely, if a patient is unable to access TP in production, does it mean that she will not be able to understand passives? 1 A further interesting question relates to whether pruning can occur not only at AgrP, TP or CP, but also above CP, and therefore manifest only in embedded sentences. This would cause an additional source of variation between structures that involve nodes above CP and structures that involve only CP. This would cause a dissociation for some individuals between monoclausal topicalization sentences and Wh-questions on the one hand and relative clauses and embedded Wh questions on the other hand. 5 References Ballard, K. J., & Thompson, C. K. (1999). Treatment and generalization of complex sentence structures in agrammatism. Journal of Speech, Language, & Hearing Research, 42, 690-707. Berndt, R. S., Mitchum, C. C., & Wayland, S. (1997). Patterns of sentence comprehension in aphasia: A consideration of three hypotheses. Brain and Language, 60, 197-221. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT press. Friedmann, N. (2001). Agrammatism and the psychological reality of the syntactic tree. Journal of Psycholinguistic Research, 30, 71-90. Friedmann, N. (2005). Speech production in Broca's agrammatic aphasia: Syntactic tree pruning. In K. Amunts & Y. Grodzinsky (Eds.), Broca’s Region. Oxford University Press. Friedmann, N. (in press). Degrees of severity and recovery in agrammatism: Climbing up the syntactic tree. Aphasiology. Guttman, L. (1944). A basis of scaling quantitative data. American Sociological Review, 9, 139-150. Love, T., & Oster, E. (2002). On the categorization of aphasic typologies: The SOAP (A Test of Syntactic Complexity). Journal of Psycholinguistic Research, 31, 503-529. Piñango-Adames, M-M. (1999). Some syntactic and semantic operations and their neurological underpinnings. (Doctoral dissertation, Brandeis University, 1999). Dissertation Abstracts International, 60(2-B), 0851. Pollock, J. Y. (1989). Verb movement, Universal Grammar and the structure of IP. Linguistic Inquiry, 20, 365-424. Sherman, J. C., & Schweickert, J. (1989). Syntactic and semantic contributions to sentence comprehension in agrammatism. Brain and Language, 37, 419–439. Thompson, C. K., Shapiro, L. P., Kiran, S., & Sobecks, J. (2003). The role of syntactic complexity in treatment of sentence deficits in agrammatic aphasia: The complexity account of treatment efficacy (CATE). Journal of Speech, Language, and Hearing Research, 46, 591-607.