3 A Pilot Study of Verbal Fluency in the Zulu Speaking Population with Preliminary Application to Traumatic Brain Injury Terri-Leigh Sperinck and Janet de Picciotto Department of Speech Pathology and Audiology University of the Witwatersrand ABSTRACT This study investigated the semantic verbal fluency (VF) abilities of non-neurologically impaired (NNI) Zulu speaking subjects in order to obtain normative data for this population. The data were analysed in terms of the total number of words generated in one minute, the number of words generated over four fifteen-second time periods and the strategies employed. Where possible, these results were compared to performance of South African English speaking NNI subjects. The responses of three traumatically brain injured (TBI) Zulu speaking subjects on VF tasks were obtained and compared to the NNI subjects. The NNI subjects performed significantly poorer than the current norms and poorer than the South African English speaking population. The TBI subjects generated fewer words than the NNI subjects. The strategies used by the TBI subjects were similar in type but were less efficiently used than the Zulu speaking NNI subjects. The results of this study are particularly relevant to the clinical role of speech pathologists in the multicultural and multilingual popu- lation of South Africa. OPSOMMING Hierdie studie is 'n ondersoek na die vermoe van nie-neurologies-gestremde Zoeloesprekers om woorde te herroep en produseer na aanleiding van semantiese kategoriee. Die doel van die studie was om norma tie we data te verkry vir hierdie groep sprekers. Die totale aantal woorde in een minuut geproduseer; die aantal woorde wat oor vier vyftien-sekonde periodes geproduseer is; en die strategiee wat deur die proefpersone gebruik is, is ontleed. Waar moontlik is die resultate met die van Engelsprekende proefpesone vergelyk. Data is ook verkry van drie Zoeloesprekers met geslote hoo/beserings, en met die data van die nie-neurologies-gestremde groep vergelyk. Die nie-neurologies-gestremde Zoeloesprekers het swakker gevaar as die Engelsprekende groep sowel as die gepubliseerde norme. Die prbefpersone met traumatiese hoo/beserings het minder w.oorde geproduseer as die nie-neurologies-gestremde groep, en alhoewel hulle dieselfde strategiee gebruik het as laasgenoerAde, het· hulle dit minder doeltreffend gebruik. Die resultate van hierdie studie is veral van belang binne die multikulturele, veeltalige Suid-Afrikaanse konteks. I KEY WORDS: verbal fluehcy, semantic verbal fluency, multicultural, multilingual I . Verbal fluency (VF) is a ,task commonly incorporated in the assessment batteries of Speech Pathologists and neuropsychologists in South Africa. In a population as culturally diverse as South Africa's, it is essential to rec- ognise the need for normative data pertaining to assess- ment tasks conducted on the non-English speaking popu- lation in order to provide a basis against which to com- pare the performance of our clients. This study investi" gated VF abilities in aspecific group from the Zulu speak- ing population of South Africa. In addition, a small sam- ple of Zulu speaking Traumatic Brain Injured (TBl) sub- jects was investigated. A discussion of VF and the provision of a rationale for the choice of Zulu speakers as the subject of investigation in this study, follows. Gruen, FrankIe & Schwartz (1990) define verbal fluency (VF) as "a language based task em- ployed to measure word associate generation and diver- gence" (p.163). A review of the literature revealed a number of studies on VF in different population groups including the elderly; non-neurologically impaired; Alzheimer's Dis- ease; Parkinson's Disease; Huntington's Disease; Demen- tia; Progressive Supranuclear Palsy and Head Injury populations (Adams, Reich & Flowers, 1989; Bolla, Lindgren, Bonaccorsy & Bleecker, 1990; Borkowski, Benton & Spreen 1967; Cauthen, 1978; de Picciotto, 1997; Gruen et aI., 1990; Hanley, Dewick, Davies, Playfer & Turnbull, 1990). Numerous studies have shown that re- sults ofVF tasks may be influenced by age, sex and verbal intelligence (Bolla et aI., 1990; Cauthen, 1978; Hanley et aI., 1990). The two VF tasks most utilised by researchers and clini- cians are phonemic a~d semantic category tasks (Adams et aI., 1989; Borkowski et aI., 1967; dePicciotto, 1997; Hanley et aI., 1990). In phonemic tasks, subjects are required to generate words that begin with a specific letter, whereas in semantic category tasks subjects are required to generate words that are part of a specific category. Phonemic VF is dependent on the "phonologically based word store" (Rosser Die Suid-Afrikaanse Tydskrif vir Kommunikasieafwykings, Vol. 46, 1999 R ep ro du ce d b y Sa bi ne t G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r ( da te d 20 12 ) 4 & Hodges, 1994, p. 1389). Semantic VF requires the indi- viduals to make an exploration of their conceptual know- ledge followed by a lexical search according to semantic cat- egories (Cardebat et aI., 1996). The lexicon involved in se- mantic categories is stored in the semantic memory; the system oflong term memory which contains general infor- mation about the world (Rosser & Hodges, 1994). Battig & Montague (1969), as cited by Adams et aI. (1989), found that the categories which are chosen for VF tasks have extensive effects on the scores of the subjects. There is a relationship between language and thought in the culture in which the language is spoken, therefore lan- guage and culture are reciprocal (Berry, Poortinga, Segall & Dasen, 1992). Consequently, when choosing categories for VF tasks it is imperative to ensure that the categories are culturally appropriate. When obtaining normative data on an assessment tool for a population on which it was not originally normed, it is not enough to simply translate the test into-the required language. As stated by Payne (1986), the roots of language are embedded in culture and as a re- sult the two cannot be separated and thus it is necessary to adapt the test to the culture for which it is being normed. When considering the applicability of the two types of VF tasks.in the Zulu speaking population, the semantic category task of VF may be more appropriate than the phonemic tasks owing to the complexity of noun class marker and noun class agreement exhibited in Zulu (Suzman, 1996). Therefore a phonemic VF task was not felt to be appropriate because Zulu speakers make use of noun class prefixes when speaking, thus words would not simply be named but rather they would be said with their class prefixes. For example, "sock" in Zulu is "isokisi" how- ever, the"i" which occurs at the beginning of the word is. the noun class marker and not in fact the initial letter of the word. VF tasks are performed over a 60-second time period. This performance can be further analysed in terms offour 15-second time periods. Adams et aI. (1989) found that the non-neurologically impaired (NNI) subjects or subjects with Dementia of the Alzheimer Type (DAT) generated the most words during the initial fifteen-second time period, however, the DAT subjects exhibited a distinctive decrease in the number of responses in the ensuing time periods. Studies conducted by Cardebat et aI. (1996) on right brain damaged subjects, and Crowe (1992) of TBI subjects, showed that there was a decrease in the number of words generated after the first 30 seconds. When performing VF tasks the individual tends to use strategies. The ability to switch strategies when the se- ries of associations runs out, as well as good organisational skills enhance performance in tasks ofVF (Cauthen, 1978). There is currently limited data on strategy use in differ- ent populations. A review ofliterature regarding tasks ofVF revealed a paucity of research on VF using non-Engiish speaking subjects. Benton & Hamsher cited by Lezak (1995) estab- lished letter ranking norms in VF tasks in French, Ger- ma:n, Italian and Spanish. There appears to be no pub- lished literature on VF research carried out in any Afri- can languages. Normative data for tasks ofVF in African languages are important in the South African context be- cause assessments currently being on non-English speak- ing clients are normed on English speakers. South Africa is a country with vast ethnic diversity and a total population of 38 million. Mother tongue Zulu speak- Terri.Leigh Sperinck and Janet de Picciotto ers comprise 22,4% (6,4 million) of the total population. Within the Gauteng province, 18,4% of the people speak Zulu as their home language (Central Statistical Services, 1998). Zulu is the predominant African language spoken in South Mrica. The most common language spoken South- ern Mrican townships is a hybrid of either Zulu or Xhosa (Maake, 1991). It can therefore be deduced that Zulu speakers form a large proportion of the Speech Patholo- gist's caseload. In order to treat populations of various languages it is necessary to understand their languages and the way in which they organise their thoughts. Estes (1974) suggests that tasks of VF represent an ideal method of examining how subjects organise their thinking. Thus, normative data of Zulu speakers is imperative in order to establish base- line measures for the purpose not only of diagnosis but also for intervention. The performance of South Mrican Zulu speakers as compared to South African English speakers on tasks of VF may provide valuable insight into cultural and linguistic differences in their approach to VF tasks. There are numerous problems facing Speech Patholo- gists working in culturally and linguistically diverse populations. One of the most critical problems that the Speech Pathologist is faced with is "the absence of adequate assessment tools needed to provide valid and reliable evaluations" of linguistically and culturally disparate populations (Vaughn-Cooke, 1986, p. 23). The need for culturally appropriate tests is vital as the employment of invalid assessment tools has direct impact OIi the client's results which in turn have far reaching implications for ·diagnosis, intervention and outcome (Vaughn-Cooke, 1986; Watt, Penn & Jones, 1996). One population typically seen in Speech Pathologist's caseloads is tpat ofTBI clients. The National Head Injury Foundation (NHIF) defines TBI as "an insult to the brain that is caused by an external force that may produce di- minished or altered states of consciousness which results in impaired cognitive abilities or physical functioning" (Ponsford, 1995, p.1). Statistics show that 70% of TBl's are caused by motor vehicle accidents (MVA) and that it is mostly men between the ages of 15-35 years that are af- fected (Gillis, 1996). Most of the men that are affected by TBI are from lower socio-economic classes who may have limited educational attainment (Ponsford, 1995; Watt, Penn & Jones, 1996). I The mechanisms that occur as a result ofTBI are com- plex and involve both focal and diffuse injuries (Gil:lis, 1996; Ponsford, 1995). Research has shown that there kre cognitive, behavioural and language consequences ofTBI (Borkowski, Benton & Spreen, 1967; Gillis, 1996; Gruen et aI., 1990; Ponsford, 1995; Schmitter-Edgecombe, 1996). The phenomenon of TBI is common worldwide although the pattern of causation differs from country to country. While. the largest cause of TBI in South Mrica is attrib- uted to motor vehicle accidents the statistics show that TBI in the African populations is mostly as a result ofvio- lence (Brown & Nel, 1991) and this etiology is adding to the statistics of TBI. Verbal fluency abilities have been explored in the TBI population. TBI subjects generate fewer words on tasks of VF than NNI subjects (Gruen et aI., 1990). The aforemen- tioned result may be attributed to the disruption in or- ganisational skills in TBI subjects, which are important requirements for tasks of VF (Gruen et aI., 1990). Furthermore, results of studies of TBI subjects indi- The South African Journal of Communication Disorders, Vol. 46, 1999 R ep ro du ce d b y Sa bi ne t G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r ( da te d 20 12 ) A Pilot Study of Verbal Fluency in the Zulu Speaking Population with Preliminary Application to Traumatic Brain Injury 5 cate that while they make use of similar strategies as NNI subjects in VF tasks, the injury appears to influence the speed and effectiveness of the strategies that are employed (Gruen et al., 1990). Anatomically the frontal lobes ap- pear to playa vital role in VF tasks and the well-docu- mented vulnerability of the frontal lobes to injury in TBI may account for the above finding (Crowe, 1992; Rosser & Hodges, 1994). The purpose of this study was therefore to gather nor- mative data on VF tasks for the Zulu speaking population and compare this performance to that of South . African English speakers in order to facilitate culturally and lin- guistically appropriate assessment, diagnosis and inter- vention. Furthermore, this study aimed to examine a sam- ple of Zulu speaking TBI clients in order to obtain an indi- cation of their performance in comparison to the NNI popu- lation. METHODOLOGY The research methodology followed the framework of de Picciotto (1997) with adaptations for the specific popu- lation of this study. AIMS This study aimed to explore VF performance in Zulu speaking subjects in order to gather normative data. The sub-aims were fourfold: » Firstly, to determine the number words generated . by Zulu speaking NNI subjects and compare them to South Mrican English speaking NNI subjects. » Secondly to determine the performance of Zulu speaking TBI subjects. » Thirdly to compare the results obtained by the Zulu speaking NNI and Zulu speaking TBI subjects. » Finally to describe any strategies used by the Zulu speaking NNI and Zulu speaking TBI subjects dur- . 'ing VF tasks. // RESEARCH DESIGN i A structured survey design was used to achieve the aims of this research as the data' are quantified and managed by means of descriptive statistics. j I TABLE 1: Description of Zulu speaking TBI subjects. SUBJECT # 1 AGE 18 ETIOLOGY assault COMA 1 week TIME SINCE INJURY 6 months EDUCATION LEVEL Std.8 PREMORBID EMPLOYMENT STATUS unemployed CURRENT I EMPLOYMENT STATUS unemployed SUBJECTS The subjects were required to meet the following crite- ria. All subjects had to be male Zulu speakers in the age range of 18-30 years with a minimum of 10 years of for- mal education. To control for the large diversity of educa~ tional standards within the South African geographical context (Hosking, 1992), subjects were required to have been raised and educated in the urban areas of Gauteng to achieve uniformity of the dialects of Zulu that the pa- tients spoke and ensure that the patient's educational backgrounds were similar. In addition, the three TBI sub- jects had to be at least 3 months post head injury in order to allow for neurological stabilisation (Lezak, 1995). The Zulu speaking NNI group was represented by forty subjects and three Zulu speaking TBI subjects were se- lected. The Zulu speaking NNI subjects ranged in age from 18-30 years (mean age = 23.58 years; SD = 4.03 years). Formal education of the NNI subjects ranged from 10-15 years (mean education = 10.15; SD = 0.99 years). Owing to the small sample size ofthe Zulu speaking TBI subjects statistical inferences could not be made, however, their details are provided in Table 1. TESTING In order to ascertain performance of the Zulu speaking NNI subjects as well as Zulu speaking TBI subjects on tasks of semantic VF, three categories were chosen, namely: "ani- mals'" "food" and "transport", as they are most commonly used by researchers in studies of VF (Gruen et aI., 1990; Hanley et al., 1990). These three categories were the .ones that were least likely to be affected by culture (Berry et al., 1992). The subjects were given a trial category of "clothes" before beginning the task. The procedure employed required that each subject generate as many words as possible over a one-minute time interval for each of the three categories. The categories were presented in the following order: "ani- mals", "food" and "transport". The interpreter verbally re- quested the subjects to continue if they stopped before the full minute (Joannette et al., 1988). At no time during the task were the subjects provided with cues. A tape recorder was used to record the subjects' responses and a stopwatch was used to time each task. The number of words generated in one minute for each category was totalled. The one- minute time interval was SUBJECT#2 SUBJECT#3 25 19 MVA stab wound unknown unknown 15 months 18 months Std.8 Std.8 storeman Unemployed unemployed Unemployed Die Suid-Afrikaanse Tydskrif vir Kommunikasieafwykings, Vol. 46, 1999 R ep ro du ce d b y Sa bi ne t G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r ( da te d 20 12 ) 6 then divided into four time intervals of fifteen-seconds each. The number of words generated in each of the four fifteen-second time intervals were then summed for each category. The strategies used by the subjects were ana- lysed in terms of type and efficiency. A strategy was pre- sumed if a subject's response was related to the word im- mediately preceding it (Gruen et aI., 1990). Once the instructions for the semantic VF task had been translated they were. piloted on a small group of Zulu speaking males. A number of alterations were made and these included a more explicit explanation of the 60-sec- ond time period for each category as well as clearer in- structions not to make use of proper nouns. Owing to the fact that all testing was conducted in Zulu, a translator was required to translate the instructions of the VF tasks into Zulu. The translator was chosen accord- ing to the criteria suggested by van Ede (1996), that is, he was required to be proficient in English and Zulu, in or- der to translate the instructions for the task into Zulu from English. Once the material had been translated, the process of back translation was employed. This is a process whereby a second person, proficient in both languages, translates the material back into the original language. The second translator does not have any previous knowledge of the original material, thereby ensuring that he is not biased. Back translation is conducted to guard-against any errors that the first translator may have made. These errors in- clude grammatical structures, word complexity and over- all resemblance of the meaning (van Ede, 1996). The original translator was used as an interpreter be- cause he was proficient in both English and Zulu and he had knowledge of the testing material. He was then in- structed in the administration of the VF tasks and worked with the researcher.in transcribing the responses. DATA ANALYSIS In order to address the main aim, normative data were depicted by tabulating the results ofthe 40 Zulu speaking NNI subjects. The means for the number of-words for each category were established as well as for each time period within each category. According to von Fraunhofer & Murray (1976), the most common method of ascertaining 'normal' is to determine the mean. Standard deviation (SD), for the number of responses in each category and the number of responses for each time period, was then calculated. SD was used to determine how much the indi- vidual values deviated from the mean values. The percent- age of SD to the mean was calculated to obtajn an indica- tion of the relative differences between the SD. This cal- culation was carried out for each category and for each time period of each category. These calculations were used to compare the results between the three categories and to compare. the results of each of the four time periods within the three categories. The results were then depicted graphically in order to ascertain whether any patterns or tr.ends were present. The results of the Zulu speaking NNI subjects on the categories of "animals" and "transport" were compared to the results of South African Englisp- speaking NNI sub- jects (de Picciotto, 1997). The category "food" was not in- cluded in this comparison since it was not used in the study on South African English speakers. The means for the number of responses for each cat- Terri-Leigh Sperinck and Janet de Picciotto egory were tabulated into frequency distributions and then represented graphically in the form of a distribution curve. These graphs were plotted in order to ascertain whether there was a Gaussian (normal) distribution about the mean or whether there was a Poisson (skewed) distribution about the mean (von Fraunhofer & Murray, 1976). To address the second aim of the study the result of each Zulu speaking TBI subject was calculated and ana- lysed descriptively. To address the third aim of the study the responses ofthe Zulu speaking TBI subjects were com- pared to those of the Zulu speaking NNI subjects. Percen- tile ranking was used for each of the Zulu speaking TBI subjects. This analysis provided information as to each Zulu speaking TBI subject's relative position in the nor- mative sample (Anastasi, 1982). Finally, the results of the Zulu speaking NNI subjects were examined to look for 'patterns' of strategies within the responses. Types of strategies that can be used are alphabetical or subordinate categories. However; these strategies may be language specific. When naming in the Zulu language all words begin with vowel such as 'i'; 'u'; 'a'. Thus to analyse the results in terms of the alphabet would not have been linguistically appropriate. Likewise; the use of subordinate categories can be subject to the in- fluence of culture. For example the category of food may be analysed in terms of food groups (vegetable, fruits, starches, meat etc.); meals (breakfast, lunch & supper); staple diet (the foods that are indigenous to the culture); and finally ceremonial foods (those foods that are eaten during important occasions). The subordinate categories used were analysed according to culture-specific organi- sation. Where possible the responses were arranged into groups for each subject. These groupings by subject were analysed numerically and separated according to perceived 'use' or 'non-use' of subordinate categories. The mean number of words achieved by the two groups was calculated and com- pared to assess efficiency of the subordinate categories. 'Efficiency of applying a strategy' was defined as the com- parative degree to which those subjects utilising a strat- egy generated ahigher mean number of words than the subjects who did not use a strategy. RESULTS AND DISCUSSION I Results are discussed according to each aim of the stlldy. i WORDS GENERATED BY ZULU SPEAKING NNI SUBJECTS IN TOTAL AND OVER FOUR FIFTEEN- SECOND TIME PERIODS. Table 2 depicts the results obtained on the VF tasks for the Zulu speaking NNI subjects. The mean number of words generated for the catego- ries "animals" and "food" was similar (means of 11.52 .and 11.65 respectively) however, for the category of "transport", fewer words were generated (mean of 7.75). The results for the category of "transport" may be attributed to the fact that there are a limited number ofpossible~esponses for the category. When the standard deviation (SD) was caJculated as a percentage of the mean for the total words, the results within each category ~ere relatively similar across the groups of "animals", ·"food"· and "transport" (27.51 %; 28.87% and 26.61% respectively). This result may imply that the word gener-ation abilities that are being / The South African Journal of Communication Disorders, Vol. .46, 1999 R ep ro du ce d b y Sa bi ne t G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r ( da te d 20 12 ) A Pilot Study of Verbal Fluency in the Zulu Speaking Population with Preliminary Application to Traumatic Brain Injury 7 used are working consistently across all three categories. Overall, the number of words generated by the Zulu speaking NNI subjects of this study is much lower than the normative data currently being used (Bayles & Tomoeda, 1991; Bayles & Tomoeda, 1993; Goodglass & Kaplan, 1972). According to the normative data the norm for "animals" is 22.5 (SD=6.8) and for "transport" is 13.4 (SD=3.2). When compared with the mean for "animals" (11.53) and "transport" (7.78) obtained in this study, the large difference suggests that a test with linguistically appropriate norms would be more appropriate. In addition when comparing the results of the Zulu speaking NNI subjects to that of South African English speaking NNI subjects obtained in a previous study (de Picciotto, 1997) there was a sigriificant difference (p