CROSSLINGUISTIC GENERALIZATION OF SEMANTIC TREATMENT IN APHASIA: EVIDENCE FROM THE INDIAN CONTEXT

• ¹ Manipal University, Department of Speech and Hearing, India
• ² Boston University, United States
• ³ All India Institute of Speech and Hearing, India

Abstract In the current study, we examined the nature of crosslinguistic generalization of treatment for Indian bilinguals with aphasia. We recruited three bilingual (Kannada-English) persons with aphasia and used the treatment protocol described by Edmonds and Kiran (2006). Our findings showed a striking similarity with the previous study, thus providing further empirical evidence for crosslinguistic generalization of semantic treatment in aphasia, especially from an unexplored language pair. Introduction The last two decades witnessed several novel treatment approaches to aphasia therapy. Semantic feature-based therapy is one of such treatment approaches that gained considerable research attention (Boyle & Coelho, 1995). More importantly, this treatment approach has been found effective in bilingual persons with aphasia. For instance, Edmonds and Kiran (2006) administered semantic feature based therapy in Spanish-English bilingual persons with aphasia and reported of crosslinguistic generalization of treatment effect to untreated language. This promising research, however, needs to be replicated and extended to novel language pairs. Research on crosslinguistic generalization of treatment effects is of paramount importance to multilingual countries like India. For instance, with several hundreds of languages and dialects spoken across India and with the pervasive use of English as second language, speech language pathologists (SLPs) in the country are often baffled on the selection of language for treatment in bilingual persons with aphasia. Empirical evidence from Indian languages would add confidence to the SLPs while selecting language for treatment in person with aphasia. In this context, the current study aimed to replicate and extend the earlier findings on crosslinguistic generalization of treatment effects in bilingual persons with aphasia to the Indian context. Methods This investigation was carried out in Karnataka, a southern state in India, where the primary spoken language is Kannada. We recruited three bilingual (Kannada-English) persons with aphasia for the current study. Their premorbid proficiency was calculated with a rating scale rated either by the participants or with the help of an intimate family member. While participant 1 (P1) was equally proficient in both languages, P2 and P3 were more proficient in Kannada. Table 1 provides the demographic data as well as language proficiency of the three participants. Prior to as well as following the administration of semantic feature-based treatment, the language skills of the participants were assessed with Western Aphasia Battery (Chengappa & Kumar, 2008), Bilingual Aphasia Test (Paradis & Rangamani, 1989), and Boston Naming Test (Shanthala, 1997) (see Table 2) in Kannada and English. ________________________ Table 1 about here ________________________ Table 2 about here ________________________ We employed the methods of Edmonds and Kiran (2006) for preparing treatment probes as well as administering treatment in our participants. From an initial set of 300 pictures that were administered on each participant, 50 unsuccessfully named pictures were selected for the training purpose. The names of these pictures were neither cognates nor had 50% or more phonetic similarity in both languages. The selected items for each participant varied from that of other participants as the final items (n = 50) were selected based on the individual naming failures. All items belonged to various semantic categories like animals, fruits, household articles etc. For each participant, six sets of stimuli were developed with 10 items in each set, except in control set, which had only 5 stimuli. The stimuli were categorized into: a) English Set 1 (e.g., cat: N = 10); b) Kannada set 1 (translation equivalent of English set 1: e.g., /bekku/: N = 10); c) English set 2 (semantically related to items in set 1: e.g., dog: N = 10); d) Kannada set 2 (translation equivalent of English set 2: e.g., /na:ji/: N = 10); e) English set 3 (control: semantically unrelated: e.g., stone: N = 5); f) Kannada set 3 (control: translation equivalent of English set 3: e.g., /kallu/: N = 5). For each participant, two sessions of therapy, each lasting for 2 hours, were provided on a weekly basis. Unlike in Edmonds and Kiran (2006), the participants were assessed only twice (i.e., before & after therapy) in the current study. Results Participant P1, equally proficient in both languages, was trained using English set 1. Performance improved on both English set 1 (from 0 - 90%; X2 = 9; df = 1; p = 0.0027) as well as on English set 2 (from 0 - 80%; X2 = 8; df = 1; p = 0.0046). Further, crosslinguistic generalization to both translation equivalents (i.e., to Kannada set 1: 0 - 70%; X2 = 7; df = 1; p = 0.0081) and semantically related items (Kannada set 2: 0 - 60%; X2 = 6; df = 1; p = 0.01431) was also noticed. The control sets, however, did not show any significant improvement following therapy. Participant P2, who was more proficient in Kannada, received treatment in the same language using Kannada set 1. Both trained set (from 0 - 80%; X2 = 8; df = 1; p = 0.0046) and semantically related untrained (i.e., Kannada set 2: 10 - 60%; X2 = 6; df = 1; p = 0.014) showed significant improvement. However, neither English set 1 nor English set 2 showed improvement, indicating poor crosslinguistic generalization. Subsequently, the treatment was provided in English using English set 2. Though the trained set showed improvements (from 10 - 90%; X2 = 8; df = 1; p = 0.0046), it failed to generalize to English set 1. However, interestingly, with treatment initiated in English, P2 showed crosslinguistic generalization to both translation equivalents (i.e., Kannada set2: 60 - 80%) as well as to the semantic related items (i.e., Kannada set 1: 80 – 100%). Neither Kannada (0%-20%; p - ns) nor English (0%-0%; p - ns) control sets showed any notable changes following the training. Participant P3, who was more proficient in Kannada like P2, received treatment in English using English set 1 and crosslinguistic generalization was monitored for Kannada set 1 and set 2. Scores on the trained set improved from 0 to 80% (X2 = 8; df = 1; p = 0.0046). Within language generalization was observed on semantically related untrained items (i.e., English set 2) as the scores improved from 0 to 40% (X2 = 4; df = 1; p = 0.045) and there was crosslinguistic generalization to Kannada set1 (from 10 to 70%; X2 = 6; df = 1; p = 0.014) and to Kannada set2 (from 0 to 60%; X2 = 6; df =1; p = 0.014). Both control sets (0 - 0%; p - ns) did not show any change after the provision of training. Discussion The results of this investigation are, in general, on par with earlier similar investigations (e.g., Edmonds & Kiran, 2006). In their study, these authors observed improvements in both languages subsequent to the provision of treatment in L2 in balanced (i.e., equally proficient) bilinguals. However, in dominant bilinguals, where there exists a difference in proficiency level between the languages, provision of treatment in weaker language resulted in crosslinguistic generalization and the results of this study corroborated their findings. This in turn, supports the Revised Hierarchical Model (Kroll & Stewart, 1994) of bilingual language system. Additionally, it may be noted that the pre- and post-therapy evaluations revealed overall improvement in languages skills of all three participants in the current study. Though unexpected, this observation warrants further exploration of the generalized effects of strengthening the semantic system on the overall language skills. Conclusion The current study replicated and extended the previous evidence on crosslinguistic generalization of semantic therapy to the Indian context. The outcomes from the current study provide empirical evidence for selecting language of treatment in bilingual persons with aphasia in the pervasively bi-/multilingual Indian context. Table 1: Demographic and linguistic proficiency data of the participants Participant Gender Age (years) Education Etiology MPO Self-ratings: Premorbid proficiency in English/Kannada (1- least proficient; 7- most proficient) Bilingual proficiency rating P1 Male 60 Graduate CVA 15 Speech:7/7 Comp: 7/7 Reading:7/7 Writing:6/7 1.00 P2 Male 58 Graduate CVA 11 Speech:5/7 Comp:7/7 Reading:6/7 Writing:5/7 1.16 P3 Male 24 Graduate CVA 13 Speech:5/7 Comp:6/7 Reading:6/7 Writing:5/7 1.27 Table 2: Performance of participants on various test of language in Kannada and English Participant 1 Participant 2 Participant 3 Western Aphasia Battery (Scores in %) English Kannada English Kannada English Kannada PRE POST PRE POST PRE POST PRE POST PRE POST PRE POST Spontaneous speech 70 80 65 70 60 75 55 60 30 45 35 55 Auditory comprehension 75 80 75 80 60 70 65 75 36.6 50 40 50 Repetition 68 76 73 81 64 73 60 80 30 47 34 46 Naming 65 75 60 70 45 60 40 55 26.6 45 28.3 48.3 Auditory word recognition 66.7 71.6 61.6 68.3 53.3 61.6 58.3 65 35 46.6 31.6 51.6 Sequential commands 72.5 78.75 55 62.5 55 63.7 55 66.2 31.2 42.5 28.7 48.7 Word fluency 40 50 40 55 35 50 40 55 25 40 30 45 Sentence completion 80 80 60 80 40 60 60 60 40 60 40 60 Responsive speech 80 80 80 80 40 60 40 50 50 60 40 60 Reading 70 80 75 80 37.5 52.5 45 60 30 45 37.5 52.5 Reading commands 60 65 70 75 50 70 55 70 30 45 35 50 Written word stimulus - object choice matching 100 100 83.3 100 66.6 66.6 66.6 83.3 66.6 83.3 66.6 66.6 Written word stimulus-picture choice matching 100 100 83.3 100 66.6 83.3 50 83.3 66.6 83.3 50 83.3 Picture stimulus written word choice matching 100 100 100 100 50 66.6 66.6 66.6 50 83.3 50 66.6 Spoken words written word choice matching 100 100 100 100 50 50 50 75 50 66.6 50 66.6 Spelled word recognition 66.6 66.6 83.3 83.3 83.3 100 83.3 83.3 50 66.6 66.6 83.3 Spelling 50 66