University of Groningen Verb and word order deficits in Swahili-English agrammatic speakers Abuom, Tom Onyango IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below. Document Version Publisher's PDF, also known as Version of record Publication date: 2013 Link to publication in University of Groningen/UMCG research database Citation for published version (APA): Abuom, T. O. (2013). Verb and word order deficits in Swahili-English agrammatic speakers Groningen: s.n. Copyright Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons). Take-down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim. Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum. Download date: 18-06-2017 Verb and Word Order Deficits in Swahili-English Bilingual Agrammatic Speakers Tom O. Abuom . CLCG The work reported in this thesis has been carried out under the auspices of the School of Behavioral and Cognitive Neurosciences (BCN) and the Center for Language and Cognition Groningen (CLCG). Publication of this thesis was financially supported by the University of Groningen, the Stichting Afasie Nederland (SAN) and BCN. Groningen Dissertations in Linguistics 119 ISSN 0928-0030 ISBN 978-90-367-6264-9 (printed version) ISBN 978-90-367-6265-6 (digital version) ©2013 by Tom O. Abuom Cover illustration by Ruggero Montalto, www.79s.co Document layout by Dörte de Kok, prepared with LATEX 2ε , typeset in pdfTEX Printed by Wöhrmann Print Service, Zutphen, The Netherlands RIJKSUNIVERSITEIT GRONINGEN Verb and Word Order Deficits in Swahili-English Bilingual Agrammatic Speakers Proefschrift ter verkrijging van het doctoraat in de Letteren aan de Rijksuniversiteit Groningen op gezag van de Rector Magnificus, dr. E. Sterken, in het openbaar te verdedigen op maandag 9 september 2013 om 11.00 uur door Tom Onyango Abuom geboren op 8 juni 1974 te Kisumu, Kenia Promotor: Prof. dr. Y.R.M Bastiaanse Beoordelingscommissie: Prof. dr. Ria de Bleser Prof. dr. David Howard Prof. dr. Loraine Obler ISBN: 978-90-367-6264-9 Acknowledgments I express my most sincere gratitude to God for the gift of life and good health and for having been my guide, light and wisdom throughout the study; and to all those who have helped me in any way in the course of this study. “I’ve learned that people will forget what you said, people will forget what you did, but people will never forget how you made them feel.” -Maya Angelou-. My heartfelt thanks to my supervisor and promoter, Prof. Dr. Roelien Bastiaanse, for her tireless scholarly and friendly guidance, timely suggestions and corrections that brought this project to fruition, I owe this great feeling of accomplishment to her. My sincere gratitude to Emmah Shah and her team at the Speech and Language Therapy department as well as the Research Ethics Committee at the Aga Khan University Hospital (Nairobi, Kenya) for helping me to identify the participants in this study and granting full access to all the facilities I needed for this study. Most sincere thanks to members of the reading committee for my thesis, Prof. Ria de Bleser, Prof. David Howard and Prof. Loraine Obler for taking their precious time to read this work. Special thanks to all members of the neurolinguistics research group of the University of Groningen, Roelien, Ben, Laurie, Gerard, Roel, Silvia, Dorte, Olga, Harwintha, Laura, Rimke, Fedor, Joost, Ellie, Seckin and Rui, who have journeyed with me during the four years of study in Groningen, for their moral support and immense contribution towards the success of this study through v vi their scholarly presentations, brotherly advice and suggestions. Special thanks to Dorte and Ruggero for being my paranimfs and for helping me with the thesis’ layout and designing of the cover. Heartfelt thanks to Trevor and Kelly for proofreading my thesis. BCN (Behavioral and Cognitive Neurosciences), CLCG (Center for Language and Cognition Groningen) and SAN (Stichting Afasie Nederland) financially supported the research presented in this thesis, for this I am deeply grateful to them. Finally, my heartfelt appreciation to all the members of my family, my mother Mary, my wife Irene and our two lovely children Reina and Ricky, all my siblings, relatives and friends who have been a great source of inspiration, moral support and encouragement throughout my life. I wouldn’t have accomplished this great work without your love and support. May God Bless You All! Groningen, September 9, 2013 Contents Acknowledgments v 1 Introduction and thesis outline 1.1 Linguistic accounts of Agrammatism . . . . . . . . . . 1.1.1 Representational deficit accounts . . . . . . . . 1.1.2 The Processing deficit accounts . . . . . . . . . 1.2 The Swahili Language . . . . . . . . . . . . . . . . . . 1.2.1 Characteristics of the Swahili Language . . . . 1.2.2 The Swahili verb complex . . . . . . . . . . . . 1.3 Bilingualism/multilingualism and aphasia . . . . . . . 1.3.1 Bilingual/multilingual aphasia . . . . . . . . . 1.3.2 Agrammatism in bilingual/multilingual aphasia 1.4 Verb and word order deficits in agrammatic aphasia . 1.5 The Research Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 3 4 8 12 14 15 23 25 27 28 30 2 Characteristics of Swahili-English Bilingual Agrammatic Spontaneous Speech 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.1 Cross-linguistic variations in aphasia symptoms . . . 2.1.2 Bilingual aphasia . . . . . . . . . . . . . . . . . . . . 2.1.3 Syntactic, lexical and morphological deficits . . . . . 2.1.4 Relevant features of Swahili . . . . . . . . . . . . . . 2.1.5 The research questions and expectations . . . . . . . 2.2 Methods and Procedures . . . . . . . . . . . . . . . . . . . . 2.2.1 Participants . . . . . . . . . . . . . . . . . . . . . . . 2.2.2 Materials . . . . . . . . . . . . . . . . . . . . . . . . 2.2.3 Procedures . . . . . . . . . . . . . . . . . . . . . . . 2.2.4 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 34 35 36 37 39 42 44 44 45 46 47 vii . . . . . . . . . . . . . . . . . . . . . . viii 2.3 2.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.1 Comparison of Kenyan and American English samples 2.3.2 Comparison of agrammatic and nbd speakers . . . . . 2.3.3 Verb inflection and time reference through verb forms Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.1 Comparisons with American English data . . . . . . . 2.4.2 Comparison of the English and Swahili samples . . . . 2.4.3 Verb inflection and time reference . . . . . . . . . . . 2.4.4 Linguistic complexity and frequency . . . . . . . . . . . . . . . . . . . 49 49 49 52 55 55 55 57 58 3 Production and Comprehension of Reference of Time in SwahiliEnglish bilingual agrammatic speakers 61 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 3.1.1 Time reference morphology in Swahili . . . . . . . . . . 63 3.1.2 Time reference in agrammatic monolinguals . . . . . . . 65 3.1.3 Time reference in agrammatic bilinguals . . . . . . . . . 68 3.1.4 The current study . . . . . . . . . . . . . . . . . . . . . 69 3.2 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 3.2.1 Participants . . . . . . . . . . . . . . . . . . . . . . . . . 70 3.2.2 Materials and procedure . . . . . . . . . . . . . . . . . . 72 3.2.3 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 3.3.1 Overall analysis: . . . . . . . . . . . . . . . . . . . . . . 77 3.3.2 Production in English and Swahili . . . . . . . . . . . . 77 3.3.3 Comprehension in English and Swahili . . . . . . . . . . 78 3.3.4 Individual results on production and comprehension of English and Swahili past . . . . . . . 79 3.3.5 Error types . . . . . . . . . . . . . . . . . . . . . . . . . 80 3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 3.4.1 Past time reference in agrammatic speakers . . . . . . . 81 3.4.2 Past time reference in English and Swahili . . . . . . . . 83 4 Sentence Comprehension in Swahili-English bilingual agrammatic speakers 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.1 Swahili syntax and verb morphology . . . . . . . . . . . 4.1.2 Theories on sentence comprehension in agrammatic aphasia 4.1.3 Swahili-English bilingual agrammatic aphasia . . . . . . 4.1.4 The predictions for the current study . . . . . . . . . . . 4.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 88 89 91 95 97 98 ix 4.3 4.4 4.2.1 Participants . . . . . . . . . . . . 4.2.2 Materials and procedure . . . . . Results . . . . . . . . . . . . . . . . . . . 4.3.1 Error analysis . . . . . . . . . . . 4.3.2 Swahili and English compared . . Discussion . . . . . . . . . . . . . . . . . 4.4.1 The problems with derived order 4.4.2 A working memory deficit? . . . 4.4.3 The effect of morphology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 99 102 104 105 105 106 107 107 5 Sentence Production in Swahili-English bilingual agrammatic speakers 109 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 5.1.1 Theories of sentence production . . . . . . . . . . . . . . 111 5.1.2 Swahili word order . . . . . . . . . . . . . . . . . . . . . 114 5.1.3 Swahili-English bilingual agrammatic speakers . . . . . 117 5.1.4 The current study . . . . . . . . . . . . . . . . . . . . . 118 5.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 5.2.1 Participants . . . . . . . . . . . . . . . . . . . . . . . . . 119 5.2.2 Materials & Procedure . . . . . . . . . . . . . . . . . . . 119 5.2.3 Scoring . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122 5.3.1 Error types . . . . . . . . . . . . . . . . . . . . . . . . . 123 5.3.2 Swahili and English compared . . . . . . . . . . . . . . . 126 5.3.3 Production and comprehension compared . . . . . . . . 126 5.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126 5.4.1 The DOP-H and linguistic complexity . . . . . . . . . . 127 5.4.2 Production in Swahili and English . . . . . . . . . . . . 129 5.4.3 DOP-H: an overarching processing theory . . . . . . . . 130 6 General Discussion and Conclusion 6.1 Major research findings . . . . . . . . . . . . . . . . 6.1.1 Language dependent variables in spontaneous 6.1.2 Representation vs. processing . . . . . . . . . 6.1.3 Bilingual aphasia and the bilingual brain . . 6.2 The clinical implications of the results of this study . 6.3 Scope for further study/research . . . . . . . . . . . . . . . speech . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 132 132 134 136 136 138 Summary 139 Nederlandse Samenvatting 143 x Appendices 2.A 2.B 3.A 3.B 3.C Swahili speech samples . . . . . English Speech samples . . . . Accuracy in English production English verb pairs in the TART Swahili verb pairs in the TART . . . . . . . . . . . . . . . . . . . . . . . . and comprehension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 148 154 156 156 157 References 159 Grodil 171 List of Figures 1.1 1.2 1.3 The syntactic tree according to Hagiwara . . . . . . . . . . . . The syntactic tree used by Friedmann & Grodzinsky (1997) . . Map of Swahili language areas . . . . . . . . . . . . . . . . . . 5 5 13 3.1 3.2 Samples of pictures used for production task . . . . . . . . . . . A sample of the pictures used for comprehension task . . . . . 73 75 4.1 4.2 Example of pictures in the sentence-picture matching task . . . Illustration of the interaction effects . . . . . . . . . . . . . . . 101 104 5.1 5.2 Example of pictures in sentence production task . . . . . . . . Illustration of interaction effects . . . . . . . . . . . . . . . . . . 121 124 xi xii List of Tables 1.1 1.2 1.3 1.4 1.5 1.6 A summary of Swahili noun classes (from Ashton, 1982). . . The information conveyed by the subject agreement marker. The summary of tense/aspect markers in Swahili. . . . . . . The six forms of object agreement in Swahili. . . . . . . . . The Structure of the Swahili verbal root. . . . . . . . . . . . The six most common types of Swahili verb suffixes . . . . . . . . . . . . . . . . 15 17 17 20 21 22 2.1 2.2 2.3 2.4 2.5 2.6 2.7 Demographic details and auditory comprehension results . . Kenyan English and American English agrammatic samples Comparison of MLU and speech rate . . . . . . . . . . . . . Comparison of noun and verb production . . . . . . . . . . Proportions of ungrammatical and embedded sentences . . Proportion of verb inflections . . . . . . . . . . . . . . . . . Production of time reference through tense and aspect . . . . . . . . . . . . . . . . . 45 49 50 51 52 53 54 3.1 3.2 3.3 3.4 Demographic details of the agrammatic speakers . . . Time reference production in Swahili and English . . . Time reference comprehension in Swahili and English Distribution of error types in production . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 77 79 81 4.1 4.2 4.3 4.4 4.5 Demographic details the agrammatic individuals Examples of target sentences . . . . . . . . . . . Percentages correct in the Swahili test . . . . . . Percentages correct in the English test . . . . . . Error types in Swahili and English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 100 102 103 105 5.1 5.2 Demographic details of the agrammatic participants . . . . . . Examples of target sentences in the sentence production task . 120 122 xiii . . . . . . . . . . . . . . . xiv 5.3 5.4 5.5 5.6 Raw scores on Swahili test . Raw scores on English test . Error types on Swahili task Error types on English task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122 123 124 125 CHAPTER 1 Introduction and thesis outline Aphasia, which literally means ‘not speaking’, is a language disorder caused by damage to areas of the brain involved in language processing, especially in the left hemisphere. Its etiologies are varied, but the condition most commonly results from lesions in the brain which are caused by cerebro-vascular accident (also known as stroke), intracranial hemorrhage, traumatic injury, tumor or brain infection. Many aphasic individuals usually have difficulties with both comprehension and production in all language modalities (listening, speaking, reading, writing and gesturing), to varying degrees depending on the locus and the extent of the lesion (Grodzinsky, 1990). The verbal expression at the sentence level of aphasics without global aphasia, where language is totally lost, is characterized either by errors generally classified as agrammatism (tendency to use “telegraphic speech” marked by grammatical errors) or by sentence interruptions due to anomias (word finding difficulties). Currently, a considerable amount of research on aphasia focuses on agrammatism –also known as agrammatic Broca’s aphasia– which is often associated with a lesion in Broca’s area, the foot of the third convolution of the dominant hemisphere (Huber, Poeck & Weniger, 2002), though it may also result from a lesion in other regions in the left hemisphere. Agrammatism, often accom- 1 2 CHAPTER 1. INTRODUCTION panied by slow, effortful and non-fluent speech, is characterized by a marked reduction in phrase length and syntactic complexity, but with relatively intact comprehension abilities. Individual differences in terms of severity ranging from mild to severe disruptions have been reported (Menn & Obler, 1990; Friedmann, 2005). A classic example of a severe case of agrammatism demonstrated by a participant in this study is illustrated in (1), the description of the ‘cookie theft’ (Boston Diagnostic Aphasia Examination, Goodglass & Kaplan, 1972). Problems with complex constructions such as passive sentences, relative clauses and other subordinate constructions have also been reported for agrammatic individuals (Menn & Obler, 1990). (1) AGRAMMATIC BROCAS PATIENT E.A. Examiner: Can you tell me what is happening in this picture? E.A.: Mum.... first... baby two... boy girl... climbing shelf after... wait wait... climbing to get cookies... after stool fall... boy fall... forget fall... Girl hold cookies boy fall. Mum washing... sink.... no no wipe plate.... tap pour sink...tap water flow sink. Agrammatism has dominated most of the recent linguistic research in aphasia, largely due to the fact that the application of linguistic theories as a tool for investigation has so far produced promising results. While these linguistic theories1 have contributed substantially to our understanding of agrammatism, they have focused almost exclusively on monolingual speakers of mainly 1 The theories and hypotheses are discussed in detail in the following sections of this proposal. 1.1. LINGUISTIC ACCOUNTS OF AGRAMMATISM 3 Indo-European languages. This thesis is unique for two reasons: it provides further evidence of features of agrammatism from a previously undescribed language, Swahili, and it also gives an insight into the patterns of manifestation of agrammatism in bilingual individuals speaking two morphologically different languages (Swahili and English) by testing these linguistic theories further. The universality of these linguistic theories in relation to Swahili, a highly agglutinative Bantu language with a complex verb inflection paradigm (compared to Indo-European languages studied so far), is extensively examined in this thesis. The findings offer valuable insights into the processing of verb inflections and word order in the bilingual brain. The thesis also suggests ideas for effective assessment and rehabilitation of aphasia. In the next sections the linguistic theories of agrammatism are elaborated, followed by a description of relevant aspects of the Swahili language, an overview of the relevant studies on multilingualism/bilingualism and aphasia, and a review of the literature on verb and word order deficits in agrammatic aphasia. The research questions and an overview of the subsequent chapters conclude this chapter. 1.1 Linguistic accounts of Agrammatism Attempting to characterize and account for the production and comprehension deficits in agrammatic speakers has been a source of intrigue and controversy among researchers over the years. Several linguistic theories have been proposed which can be grouped into two main categories: representational deficit accounts (see Grodzinsky, 1986, 1990, 1995, 2000; Friedmann, 1998, 2002; Friedmann & Grodzinsky, 1997) and processing deficit accounts (see Kolk 1998; Haarmann & Kolk, 1991; Avrutin, 2000; 2006; Bastiaanse & Van Zonneveld, 2006; Bastiaanse, Koekkoek & Van Zonneveld, 2003; Thompson, 2003). However, whether agrammatism can be characterized fully as either a representational deficit or a processing deficit is still a matter of debate. Within the two main categories, there are a number of theories aiming to explain either production and/or comprehension deficits. The following sections present the two categories of linguistic accounts, focusing on the most influential theories of agrammatism, some of which are tested extensively in this study. 4 CHAPTER 1. INTRODUCTION 1.1.1 Representational deficit accounts Proponents of the representational deficit accounts assume that the speech production and comprehension difficulties generally experienced by agrammatic speakers reflect a partial loss of syntactic competence due to incomplete linguistic representation. Lukatela et al. (1995) used the term ‘structural deficit hypothesis’ to explain the same phenomenon, arguing that agrammatism reflects impairment in the knowledge of grammar, rather than a lack of processing abilities. According to these proponents, the grammatical representations of a language are assumed to be lost in agrammatism, and are therefore unavailable to the patients. By mainly focusing on the syntactic tree, they argue that parts of the sentence representation, such as CP-nodes or traces, are missing; hence the patients cannot use them. Such patients, for instance, will not be able to form wh-questions nor comprehend sentences in which elements are not in their base positions. We focus on two main theories of the representational deficit accounts: the Tree Pruning Hypothesis (Friedmann & Grodzinsky, 1997) and the Trace Deletion Hypothesis (Grodzinsky, 1984; 1995). The Tree Pruning Hypothesis Friedmann and Grodzinsky (1997) proposed the Tree Pruning Hypothesis based on the initial assumption of Ouhalla (1990) and Hagiwara (1995)2 that the functional categories at the top of the syntactic tree are unavailable to agrammatic speakers. Ouhalla (1990) assumed that agrammatic speakers cannot project above the Verb Phrase (VP); any node above the VP is missing in the syntactic tree. Therefore, the agrammatic individual speaks in ellipses in which free and bound morphemes are missing. Hagiwara (1995) propounded Ouhalla’s idea by proposing that functional heads and their projections in lower positions of the sentence structure are more accessible than those in higher positions for agrammatic speakers. She assumes that the Complementizer Phrase (CP) is unavailable (see Figure 1.1). Hagiwara’s assumption that only the CP is missing was based on data from Japanese agrammatic speakers as well as previously reported data from agrammatic speakers of Romance and Germanic languages. The TPH, meanwhile, includes the Tense Phrase (TP) in the missing part of the syntactic tree, based on Pollock’s (1989) theory that the IP node is split into separate Tense and 2 Note that Hagiwara, however, assumes a processing deficit: a deficiency in resources to contruct a full syntactic tree. 1.1. LINGUISTIC ACCOUNTS OF AGRAMMATISM 5 CP C' Whquestions IP Complementizer I' AgrOP Tense & Agreement AgrO' VP Object Agreement V Figure 1.1: The syntactic tree according to Hagiwara, from which the CP might be missing in agrammatic sentence representation. CP C' Whquestions TP Complementizer T' AgrP Tense Agr' VP Agreement V Figure 1.2: The syntactic tree used by Friedmann & Grodzinsky (1997), based on Pollock’s theory that the IP node is split into separate Tense and Agreement nodes. Agreement nodes (TP and AgrP). Friedmann and Grodzinsky (1997) assumed the TP is above the AgrP in the syntactic tree3 following Pollock’s (1989) 3 Tense Phrase is considered to be located above Agreement Phrase in the phrase structure (syntactic tree) of Hebrew. However, there are several languages in which TP is below AgrP, such as Dutch and German. 6 CHAPTER 1. INTRODUCTION theory, as reflected in Figure 1.2. The TPH predicts that the syntactic trees underlying agrammatic phrase structure are pruned from the Tense node up, including the CP. This implies that the Agreement node and all the nodes below it remain unavailable to the agrammatic speaker. Friedmann & Grodzinsky (1997) reported the case of an agrammatic speaker of Hebrew who showed intact agreement inflection, but impaired tense inflection, use of copulas, and embedded clauses. Subsequently, Friedmann (2000) examined the performance of 16 agrammatic speakers4 on tests of wh-question and yes/no question formation and found that higher nodes of the syntactic tree, necessary for wh-questions in English, Arabic and Hebrew, as well as for yes/no questions in English, are impaired. Generalizing, the author concluded that the agrammatic speakers had intact Agreement, but were impaired at the levels of Tense and Complementizer due to the syntactic tree being “pruned” between the Tense Phrase and the Agreement Phrase. The TPH, thus, proposes that agrammatic speakers can successfully produce agreement inflection, but they will have difficulties producing objectextracted sentences, complementizers or tense inflections in English and Hebrew. This hypothesis has been challenged by a number of studies. Lee, Milman and Thompson (2008) and Penke (2001) have found Tense, but not Complementizers, impaired; Bastiaanse (2008) and Lee et al. (2008) have found both Tense and Agreement impaired; whereas Burchert, Swoboda-Moll and De Bleser (2005) found selective impairments of both Tense and Agreement in agrammatic speakers. Also, the TPH suggests that all the syntactic representations under TP are intact, which, however, is not the case as demonstrated by other studies (Bastiaanse et al., 2003; Burchert, Meissner & De Bleser, 2008). While the Tree Pruning Hypothesis has mainly focused on production problems, the Trace Deletion Hypothesis has been introduced to explain comprehension problems in agrammatic speakers under representational deficit accounts. Trace Deletion Hypothesis The Trace Deletion Hypothesis (TDH) was proposed by Grodzinsky (1984) to account for comprehension deficits in agrammatic speakers. It has undergone a series of revisions over the years (Grodzinsky, 1986; 1995; Drai & Grodzinsky, 2006). The TDH basically assumes that all argument traces are missing from the sentence representations of agrammatic speakers. Traces are abstract 4 There were 13 Hebrew speakers, 2 Arabic speakers, and 1 English speaker. 1.1. LINGUISTIC ACCOUNTS OF AGRAMMATISM 7 markers left by moved constituents in their original D-structure; they ensure the correct interpretation of thematic information. In sentences, verbs assign thematic roles to arguments in their original positions. Moved arguments receive their thematic roles through traces connecting them to their original positions. According to the TDH, agrammatic speakers have great difficulty understanding grammatically complex sentences such as passives, object clefts and object relatives, because elements in such sentences have moved from their original positions. Grodzinsky argues that the traces linking the moved constituents to their original positions are missing from the sentence representations of agrammatic individuals, hence thematic roles cannot be assigned to the moved elements. This is illustrated in (2), where the indexations ‘i’ indicate the relation between the moved constituent and its orginal position. (2) (a) The girl pushed the boy. (b) The boyi was pushed i by the girl. (c) The girli who i is chasing the boy is tall. (d) It is the girli who i is chasing the boy. In a passive sentence such as (2b), the NP (‘the boy’) has been moved from its original position to the first position in the sentence, but it does not function as the agent. It is assigned a correct thematic role by being co-indexed to its trace at the original position. Agrammatic speakers have been found to have difficulties in correctly interpreting the NP (‘the boy’) as the patient of the pushing event. Grodzinsky argued that when agrammatic speakers encounter such difficulties, they may resort to non-syntactic ways of interpreting the sentence, that is, a default or linear strategy, which assigns an agent role to the first NP in the sentence. However, they can simultaneously interpret the second NP (‘the girl’) as the agent because its thematic role is assigned by the “by-phrase”. As a result, they are faced with a representation consisting of two agents (‘the boy’ and ‘the girl’), leading to guessing, as shown by scores at chance level on binary choice tests. The default strategy has also been suggested to explain the above chance level performance of agrammatic speakers in sentences such as subject relatives (2c) and subject-clefts (2d). These sentences contain traces as well, but the default strategy (the first NP is the agent) results in the correct interpretation. The TDH and the default strategy proposed by Grodzinsky, however, cannot account for all comprehension deficits in agrammatic speakers. Fur- 8 CHAPTER 1. INTRODUCTION thermore, several researchers have disputed the TDH. For example, Berndt, Mitchum and Haendiges (1996), after a meta-analysis of a series of studies on the comprehension of passive and active constructions, claimed that there is no homogeneous pattern of comprehension errors in agrammatic speakers. Beretta and Munn (1998) showed that agrammatic representations do not involve double agent representations in passive sentences, as suggested by Grodzinsky’s default strategy. Pinango (2000) suggested that agrammatic individuals can represent subject-NP-traces as shown by their excellent performance on comprehension of unaccusative sentences. The recent revised version of the TDH by Drai and Grodzinsky (2006), which includes data from Dutch and German agrammatic speakers, has also been disputed by Bastiaanse and Van Zonneveld (2006) through their Derived Order Problem-Hypothesis (DOP-H), which is a processing account. Unlike the TDH which is restricted to comprehension deficits, the DOP-H is an overarching theory that includes both comprehension and production deficits in agrammatic speakers. We discuss the DOP-H in the next section under the processing deficit accounts. 1.1.2 The Processing deficit accounts Proponents of the processing deficit accounts claim that the linguistic competence of agrammatic individuals is unaffected by their lesions, and that the observable difficulties are instead the result of reduced capacities that are needed for computing complex linguistic operations. Simply put, they argue that agrammatic speakers have their linguistic representations intact, but that they have limited computational resources to exploit them fully (Avrutin, 2000; 2006; Burchert, Swoboda-Moll & De Bleser, 2005; Bastiaanse & Van Zonneveld, 2005; Thompson, 2003). They dispute the proponents of representational accounts’ suggestion that agrammatism is “an all or nothing” phenomenon, based on the variability in agrammatic speakers behaviour. The processing deficit accounts predict that agrammatic speakers will show variation in their use of particular functional elements based on the complexity of the clause. We focus on four theories of processing accounts that are relevant to our project: the Derived Order Problem-Hypothesis (Bastiaanse & Van Zonneveld, 2005), the Tense Underspecification Hypothesis (Wenzlaff & Clahsen 2004, 2005), the Tense and Agreement Underspecification Hypothesis (Buchert et al., 2005) and the Past Discourse Linked Hypothesis (Bastiaanse, Bamyaci, Hsu, Lee, Yarbay Duman & Thompson, 2011). 1.1. LINGUISTIC ACCOUNTS OF AGRAMMATISM 9 The Derived Order Problem-Hypothesis The Derived Order Problem-Hypothesis (DOP-H) was formulated by Bastiaanse and Van Zonneveld (2005) based on data from Dutch and English agrammatic speakers. According to the DOP-H, all languages have a base order in which constituents that naturally belong together are adjacent in their representations. The base word order is not the same for all languages: English, for example, has a base word order of Subject-Verb-Object (SVO), whereas both Dutch and German have Subject-Object-Verb (SOV) as a base order. However, other word orders – known as ‘derived orders’– in which the constituents have overtly moved from their base positions are also possible. The DOP-H proposes that agrammatic speakers will have difficulty comprehending and producing all sentences with derived orders, but that sentences in base order are relatively spared. Several studies have supported this hypothesis. For Dutch, Bastiaanse et al., (2002a) reported that agrammatic speakers showed greater difficulty with finite verb production in the matrix clause (derived order) than in embedded clauses (base order). For English, Bastiaanse and Thompson (2003) reported that agrammatic speakers had more problems with auxiliaries in derived position in yes/no questions than with auxiliaries in base position in declarative sentences. Recently, two studies were performed on the production of sentences with scrambled objects. Burchert et al. (2008) and Yarbay Duman et al. (2008) have reported that agrammatic speakers have more difficulties producing sentences with object scrambling than sentences with objects in their base positions in German and Turkish, respectively. Bastiaanse and colleagues (2002a) concluded that agrammatic speakers are better at producing and comprehending sentences in which the constituents are in base order than sentences with derived orders, and that movement of constituents low in the syntactic tree is impaired as well, contrary to the predictions of the TPH. The DOP-H can only account for and is only meant to describe word order problems. It cannot account for all agrammatic phenomena, such as the specific difficulties with verb inflection that have been reported in several languages (see: Wenzlaff & Clahsen, 2004; 2005, for German; Friedmann & Grodzinsky, 1997, for Hebrew; Stavrakaki & Kouvava, 2003, for Greek; Yarbay Duman & Bastiaanse, 2009, for Turkish). 10 CHAPTER 1. INTRODUCTION The Tense Underspecification Hypothesis Wenzlaff and Clahsen (2004, 2005) proposed the Tense Underspecification Hypothesis (TUH) after analysis of data from German agrammatic speakers. In this account, the authors argue that the difficulties with tense evident in the performance of German agrammatic speakers are the result of the tense features being underspecified. In their first study (Wenzlaff & Clahsen, 2004), both tense and agreement were examined through multiple choice and grammaticality judgment tasks. They found both past and present tense to be more impaired than agreement. Grammatical mood, which was tested in their subsequent study (Wenzlaff & Clahsen, 2005), was found to be relatively intact in agrammatic speakers. Wenzlaff and Clahsen, therefore assume that tense features are underspecified in agrammatic aphasia, resulting in difficulties with expressing reference to the past, present and future. They argue that it is not the syntactic category of tense per se nor its position in the syntactic tree that is responsible for the difficulties of agrammatic speakers, but the interpretability of the features of tense used for time reference. The tense features express an extrasentential relationship between verb inflection and a timeframe outside the sentence, while agreement features express an intrasentential relationship between the subject and the finite verb. The TUH, therefore, predicts intact agreement features and underspecified tense features in agrammatic speakers’ production and comprehension. This theory was disputed by Burchert et al. (2005), who proposed a slight adaptation: the Tense and Agreement Underspecification Hypothesis. The Tense and Agreement Underspecification Hypothesis The Tense and Agreement Underspecification Hypothesis (TAUH), formulated by Burchert et al. (2005), assumes that tense and agreement can be affected independently in agrammatic speakers. The authors examined both tense and agreement in German agrammatic speakers, applying a similar test design as used by Wenzlaff & Clahsen (2004; 2005). Their findings revealed a bidirectional individual variation: some agrammatic speakers had problems with tense, whereas others had problems with agreement. Burchert et al. (2005), therefore, argued that a selective underspecification of tense or agreement leads to impaired tense or agreement morphology, whereas a concurrent underspecification of both tense and agreement results in problems with both tense and agreement morphology in agrammatic production. The assumption of the 1.1. LINGUISTIC ACCOUNTS OF AGRAMMATISM 11 TAUH, that both tense and agreement can be impaired independently has been confirmed in other studies, such as Bastiaanse (2008). Bastiaanse has argued that agreement may not be intact or better preserved than tense in every language if the current cross-linguistic data are considered; however, she attributes the tense problems to time reference rather than to its position in the syntactic tree. Bastiaanse (2008) argues that it is not a ‘tense problem’ per se, but a problem of time reference. Reference to the past is especially difficult for agrammatic speakers because untensed forms referring to the past, such as participles, are impaired as well. The time reference problem is accounted for by the Past Discourse Linked Hypothessis (PADILIH; Bastiaanse et al., 2011). The Past Discourse Linked Hypothesis Avrutin (2000; 2006) introduced the idea of two levels of syntactic processing into aphasiology: the “narrow syntax” processing which takes place within the sentence level, and the “discourse syntax” processing needed for interpretation of elements that need extrasentential information in order to be interpreted. Discourse linked elements have to be processed by discourse syntax, which, according to Avrutin (2006), is impaired in agrammatic aphasia since the extra processing resources required for discourse linking (‘discourse syntax’ but not ‘narrow syntax’) are not sufficiently available for agrammatic speakers due to their brain damage. Avrutin (2000; 2006), therefore, argues that the general difficulty with tense for agrammatic speakers arises from the fact that the processing of tense requires discourse linking. Zagona (2003; in press), however, argues that only past tense requires discourse linking, whereas both present and future tenses do not. She explains that since speech time and event time do not coincide when referring to a past event, a discourse-linked relation has to be established between the speech time and an earlier event. For reference to non-past events (present and future), such linking is not needed. Bastiaanse and colleagues (2011) adapted Zagona’s theory and extended discourse linking to include not only past tense, but reference to the past through grammatical morphology in general. The proposed Past Discourse Linked Hypothesis (PADILIH) predicts that time reference to the past is difficult for agrammatic speakers because it is discourse-linked, whereas time reference to both the present and the future are not. The PADILIH was proposed by Bastiaanse and colleagues (2011) after considering the results of several cross linguistic studies showing impaired time reference to the past for agrammatic speakers (Stavrakaki & Kouvava, 2003; 12 CHAPTER 1. INTRODUCTION Nanousi, Masterson, Druks & Atkinson, 2006; Bastiaanse, 2008; Yarbay Duman & Bastiaanse, 2009; Bastiaanse et al., 2011). Stavrakaki and Kouvava (2003) analyzed the production of tense, aspect and agreement in the spontaneous speech of two Greek agrammatic speakers and found impaired production of past tense compared to the present tense, and past tense with perfective aspect was more impaired than past tense with imperfective aspect; subjectverb agreement was realtively spared. Nanousi et al. (2006) reported similar results in their study of Greek agrammatic speakers who showed impaired production of perfective aspect. The pattern was replicated in the performance of Dutch agrammatic speakers on a sentence completion task reported by Bastiaanse (2008): participants made more errors with past tense than with present tense and with past participles than with infinitives. Yarbay Duman and Bastiaanse (2009) found the same for past tense/perfect aspect versus future tense/imperfect aspect in Turkish agrammatic speakers. These findings were further confirmed by the results of a cross-linguistic study by Bastiaanse and colleagues (2011) in which the Test for Assessing Reference of Time (TART; Bastiaanse et al., 2008) was used to test both the production and comprehension of time reference to the past, present and future among agrammatic speakers of Chinese, English and Turkish. While both the production and comprehension of reference to the past time frame through verb inflection was selectively impaired among Turkish and English agrammatic speakers, the Chinese agrammatic speakers showed a similar pattern only on comprehension. The Chinese agrammatic speakers’ production of all three time frames (past, present and future) was equally affected, which the authors attribute to a general problem with aspectual adverbs (see Bastiaanse, 2013). In Chinese, time reference is expressed through aspectual adverbs rather than verb inflections as in other languages. This theory has been tested extensively in both languages of this study’s bilingual agrammatic speakers in Chapter 3. 1.2 The Swahili Language The Swahili language, also known as Kiswahili, is one of the most widely spoken languages in Africa. It belongs to the large family of Bantu languages spoken in the southern half of the African continent. It is spoken as a first language by approximately five million people who live on the east coast of Africa that stretches from the south of Somalia to the north of Mozambique, 1.2. THE SWAHILI LANGUAGE 13 including the islands of Pate, Lamu, Pemba, Zanzibar and Mafia. From east to west, the area of influence of Swahili extends from Tanzania and Kenya through the interior of Congo (ex-Zare), up to Uganda, Burundi, Zambia and Malawi (see Figure 1.3). Approximately two centuries ago, Swahili, which means “the coast”, became the language of contact and communication between the people who came from elsewhere (e.g., explorers, merchants, slave traders, colonizers, missionaries, etc.) and the populations of the coast and the interior of the continent. Developing as a lingua franca, it preserved its Bantu grammatical structure characterized by a system of nominal classes and agglutinative verbal constructions, while it was lexified by a substantial number of foreign words, for the greater part from Arabic and Persian, but also some Portuguese, German and, most recently, English. Perrot (1969) points out that Swahili has incorporated words from many other sources and “Bantuized” them to such an extent that most speakers do not recognize that they are foreign words. However, the structure of the Swahili language remains Bantu, notwithstanding the large number of borrowed foreign words. It was first written in Arabic characters and, more recently, in the Roman alphabet. Figure 1.3: A section of the Map of Africa showing countries in which the Swahili language is spoken (from Stanford.edu.) In the following sections, some of the general characteristics of the Swahili language and its verbal system, which are crucial to this project, are explored. 14 CHAPTER 1. INTRODUCTION 1.2.1 Characteristics of the Swahili Language The word order of Swahili, like that of English, is Subject-Verb-Object (SVO). It is an agglutinative language, with considerable prefixing and suffixing. Like other Bantu languages, Swahili has a derivational system, by which derivational affixes can be attached to the original stem in order to express an agent, “state” of a thing and totality. Derivation in Swahili takes place at nominal and verbal levels. At the nominal level, as illustrated in Mohammed (2001), a word like ongoz-a (‘lead’) can become u-ongoz-i (‘leadership’), ki-ongozi (‘leader’), ma-ongoz-i (‘direction’) and mw-ongoz-o (‘guideline’). At the verbal level, derivation subsumes six major grammatical categories: passive, stative, causative, applicative, conversive and reciprocal. Unlike derivational systems which cause a change of word class, Swahili has an inflectional system whereby the grammatical class of a word does not change with the addition of inflectional affixes. The inflectional system operates largely at three levels: nominal, pronominal5 and verbal. Inflectional affixes are normally attached to a word stem to signal grammatical relationships such as number and possession. At the nominal level, a class-prefix is attached to the nominal stem (see Table 1.1). For instance, the noun stem, toto (‘child’) becomes m-toto (‘one child’) and wa-toto (‘children’). At the verbal level, inflectional affixes are added to the verbal stem which normally comprises the root, with or without other suffixes and the final particle -a. For instance, the word chez-a (‘play’) can become a-na-chez-a (‘s/he is playing’) or wa-na-cheza (‘they are playing’), expressing person, number and tense. Moreover, the verbal system can be very agglutinative in nature, and in most cases can be regarded as a complete sentence consisting of the subject, verb and the object as shown in (3). Ni is the subject prefix indicating the first person singular; the particle na is the tense marker for present; wa is the pronoun functioning as a direct object, the third person plural; and heshimu ‘respect’ is the verb stem. U is a mandatory final suffix. The verb root heshim itself does not change, and can only be meaningful after acquiring the relevant affixes. (3) Ni-na-wa-heshimu sana I-PRESENT-them-respect a lot ‘I respect them a lot.’ 5 The nominal and pronominal levels are not the focus of this study; therefore, they will not be discussed further. 1.2. THE SWAHILI LANGUAGE 15 Table 1.1: A summary of Swahili noun classes (from Ashton, 1982). Noun class 1. 2. 3. 4. 5. 6. 7. 8. M-WA M-MI JI-MA KI-VI N-N U-U KU PA Example Singular Plural Mtu (‘a person’) Watu (‘persons’) Mlango (‘door’) Milango (‘doors’) Jiwe (‘stone’) Mawe (‘stones’) Kijana (‘a boy’) Vijana (‘boys’) Ndugu (‘brother’) Ndugu (‘brothers’) Ukuta (‘wall’) Kuta (‘walls’) Kusoma (‘reading’) Kusoma (‘reading’) — this class inherently contains one noun only (mahali / ‘place’), signifying location The Swahili language is also characterized by concordial agreement. The noun or pronoun governs the agreement of all the other words that are syntactically associated with it. Such agreement, according Mohamed (2001), is signaled by concordial prefixes and points to person, gender and number. (4) is an illustration of how the concept of agreement operates in Swahili. The concordial prefixes are underlined, and it is clear that they agree with the nouns. Traditionally, Swahili has eight noun classes (Loogman, 1965 & Ashton, 1982) and each concordial prefix is determined by its noun class. Table 1.1 gives a summary of the Swahili noun classes, a traditional classification of Swahili nouns according to their initial prefix (Ashton, 1982). (4) mrefu ule u-me-anguka leo. Mti A tree tall that it-PERFECT-fall today ‘That tall tree has fallen down today.’ The verbal morphology system is discussed in detail in the following sections. 1.2.2 The Swahili verb complex The distinguishing feature of the Swahili verb is its agglutinative form. Depending on the context, a certain number of prefixes and suffixes must be attached to the verb root for it to be functional. These affixes usually assume distinct 16 CHAPTER 1. INTRODUCTION positions and functions. The verb root cannot be further reduced morphologically. When a final suffix -a is attached to the verb root, it becomes a verb stem. With an exception of a few verbs borrowed from Arabic, all verb stems in Swahili end in -a. Illustrated below is the order of affixed and root elements in the verbal complex (5) and an example (6). The subject agreement, tense, root and the final vowel are obligatory in every affirmative Swahili utterance (Ashton, 1982). (5) Subject Agreement + Tense SA (6) + Object Agreement T + Root OA R + Suffixes + Final Suffix FS Alimpiga ‘S/he hit him/her’ A SA + + li T + + m OA + + pig R + + a FS The structure, occurrence and function of each of these elements are described in order in the following sections. Subject Agreement The subject agreement marker occurs in the initial position of the verb complex. It relates to the noun being the subject of the verb in the sentence and it is obligatory in almost all contexts in Swahili, whether the subject is overtly realized or not. It conveys almost all the information about the noun, including the class, the number and the person; therefore, it can also function as the pronoun in instances where the subject noun is not realized as shown in (6). The subject agreement markers vary depending on the class, person, number and role of the subject noun. See Table 1.2. The personal pronouns in the sentences in Table 1.2 are optional and are generally omitted since the subject agreement markers serve the same functions; however, they can be included for emphasis. Ashton (1982) and Myachina (1981) argue that the subject agreement markers are indispensable components of the verbal complex. Krifka (1995), in his survey of Swahili syntax, added that subject agreement is obligatory in all cases with the exception of certain verb tenses such as the habitual ‘hu’ (eg. Mimi huanguka: ‘I usually fall’). 1.2. THE SWAHILI LANGUAGE 17 Table 1.2: The information conveyed by the subject agreement marker. Optional Pronoun st 1 Singular 2nd Singular 3rd Singular 1st Plural 2nd Plural 3rd Plural Mimi Wewe Yeye Sisi Ninyi Hawa SA T Root FS subject past (to fall) final niuatumuWa- -li-li-li-li-li-li- -anguk-anguk-anguk-anguk-anguk-anguk- -a -a -a -a -a -a ‘I fell’ ‘You fell’ ‘He fell’ ‘We fell’ ‘You (all) fell’ ‘They fell’ Tense/Aspect The Swahili tense/aspect marker is obligatory in every indicative utterance. It is ungrammatical to omit the tense/aspect marker whether or not the temporal reference is clear from the discourse or other sources. A summary of various tense /aspect markers occurring in Swahili are presented in Table 1.3 below. Table 1.3: The summary of tense/aspect markers in Swahili. Tense/Aspect morpheme Meaning -li-na-/-ahu-ta-ka-me-sha-ki-nga-ku- past present, on-going habitual future narrative, resultative present perfect present perfect completive conditional hypothetical infinitival The -li- morpheme is used as a simple past tense marker to refer to a past activity without reference to a specific time. According to Comrie (1976), the -li- morpheme is an absolute tense, and it can be used as an anchoring tense in discourse independent of the surrounding context. The -na- and a- morphemes are mutually interchangeable according to Mohammed (2001), 18 CHAPTER 1. INTRODUCTION and both can be used to refer to actions taking place at the present moment. The -a- tense is considered merely as a dialectical variant of the -na- tense marker; however, the -a- is sometimes used to state facts regardless of a time dimension or to state scientific truths that are not necessarily connected to the present (Ashton, 1982). Also related to the present is the hu- prefix for the habitual present tense. The hu- tense marker is used to speak about habitual or recurrent actions without making reference to any specific time. It does not take subject prefixes/agreement. -Ta- is a future tense marker that describes events assumed to follow the present time. It is regarded as a relative tense marker that takes its reference from the immediately preceding context or the matrix tense when in an embedded clause. The morphemes: -ka-, -me-, and -sha- are also considered as relative tense markers. The -ka- tense marker is generally associated with the notion of consecutiveness. Steere (1976) and Ashton (1982) point out that the -ka- tense is used to express an action or state which follows another action, and it is most frequently used in narratives and story-telling in order to drive the story line forward in time. It is also referred to as a continuative marker that takes anaphoric interpretation from the previous action (see example (7)). (7) A-li-ruka chini a-ka-kimbia S/he-PAST-jumb down s/he-CONSEC-run ‘He jumped down, he (then) ran off.’ Both the -me- and -sha- are present perfect tense markers. They both refer to a past activity which has relevance to the present time, however, the difference is that -sha- carries a more emphatic sense of completion. The English translations of -me- is “have done X”, whereas -sha- is “have already done X”. The subsequent tense morphemes -nga- and -ki- are less commonly used in Swahili. The -nga- is a hypothetical morpheme, and is by far the least used of all the tense affixes. The -ki- is a habitual or conditional marker comparable to ‘if’ and “when” in English. Finally, the -ku- is an infinitive marker that serves two functions. It functions as the verbal infinitive which corresponds to the English infinitive usually signaled by use of ‘to’ before the verb. It also functions as a verbal noun or gerund. See examples (8) and (9). (8) A-na-penda ku-cheza S/he-PRESENT-like to-play ‘S/he likes to play’ 1.2. THE SWAHILI LANGUAGE (9) 19 Ku-cheza sana ha-kufai To-play excessively not-good ‘Playing excessively is no good’ Some of the tense morphemes mentioned here are significant in this study. The three tense prefixes: -li- (for past), -na- (for present) and -ta- (for future) are the most commonly used for the reference of time, and will therefore be discussed further in our discussion of the time reference of agrammatic speakers of Swahili. The Object Agreement The next component of the verb complex is the object agreement. The object agreement marker must agree with the person and noun class of the object in the sentence. According to Ngonyani (1996), object agreement in Swahili, like in other Bantu languages, is obligatory in simple transitive sentences when the object is specific and/or animate. However, in situations where the object of the sentence is inanimate and not specific, the object agreement marker is optional. See examples (10) for an animate and specific object and in (11) for an inanimate and non-specific object, according to Keach (1995). There are six forms of personal pronoun object agreement as shown in Table 1.4. However, there is an additional reflexive object -ji- used when the subject and the object are the same person in a sentence, as shown in (12). (10) (a) Ni-na-m-penda Juma I-PRESENT-him/her-like Juma ‘I like Juma’ (b) *Ni-na-penda Juma (11) (a) Ni-na-ki-soma kitabu I-PRESENT-it-read a book ‘I am reading the book’ (b) Ni-na-soma kitabu I-PRESENT-read a book ‘I am reading a book’ 20 CHAPTER 1. INTRODUCTION (12) A-na-ji-penda S/he-PRESENT-him/herself-like ‘S/he loves him/herself.’ Table 1.4: The six forms of object agreement in Swahili. Object agreement -ni-ku-m-tu-mu-wa-ji- person 1st singular 2nd singular 3rd singular 1st plural 2nd plural 3rd plural Reflexive It is important to note that a verb can be subcategorised for only one object. In situations where a sentence has both a direct (accusative) and indirect (dative) object, the indirect object is included in the verb complex, whereas the direct object is either stated in the sentence as a noun, or possibly understood by the context. See the example (13) below. (13) Wa-me-m-pa [mtoto] vitabu They-PERFECT-him/her-give [child] books ‘They gave [the child] some books’ The Verbal Root The Swahili verbal root is the core of the verbal paradigm. Similar to most Bantu languages, the verbal root is generally monosyllabic, with the exception of some verbs originally borrowed from other languages, particularly Arabic. The structure of the verbal root is predominantly CVC, but other structures are also possible as illustrated in Table 1.5. An obligatory final suffix is usually attached to the verbal root to form a verb stem. Optionally, one or more suffixes are attached to the verbal root to give varying shades of meaning to the verb, for example, passive, stative, causative, reciprocal, and prepositional, among others. Swahili has very few monosyllabic verbs whose roots contain a single consonant – although they happen to be the most commonly used verbs in the 1.2. THE SWAHILI LANGUAGE 21 Table 1.5: The Structure of the Swahili verbal root. Root structure Example + final suffix (verb stem) English Translation C VC CVC CVCC VCVC VCC VCCVC p-a iv-a pig-a shind-a azim-a amb-a anguk-a ‘give’ ‘become ripe, mature’ ‘hit’ ‘win’ ‘borrow’ ‘adhere’ ‘fall’ language. To be functional in a sentence, the syllable ‘ku’ must be inserted before these monosyllabic verb roots to make them disyllabic, since the stress in Swahili falls on the final syllable where a prefix cannot carry stress as in (14). However, in cases where certain prefixes that carry stress (such as ‘haku’) occur with the verb root, the ‘ku’ is not inserted as in (15). (14) A-me-ni-kup-a chai S/he-PERFECT-me-give-FINAL SUFFIX tea ‘S/he gave me some tea’ (15) Haku-ni-p-a chai Not-me-give-FINAL SUFFIX tea ‘(S/he) did not give me some tea.’ The Suffixes Swahili verbs may have up to three optional suffixes attached to the verbal root in addition to (and preceeding) the final suffix. The Swahili verbal root is considered rich enough to allow various forms of conjugations through suffixations such that numerous meanings are produced with subtle differences (Mohammed, 2001). The six most common types of Swahili verb suffixes are classified according to their functions: passive, stative, causative, applicative/prepositional, conversive and reciprocal, as illustrated in Table 1.6. For this study, the passive form, which is common in Swahili, will be discussed in detail in chapters 4 and 5. The passive in Swahili resembles that of 22 CHAPTER 1. INTRODUCTION Table 1.6: The six most common types of Swahili verb suffixes and their functions. Suffix Function Description Examples Passive Causative Causes the theme to be subject of the sentence Makes the verb causative Applicative/ prepositional Reciprocal Introduces an additional argument to the sentence Reciprocalizes the verb Stative Makes the verb stative Conversive Reverses the meaning of the verb chez-a (play) → chez-w-a (be played) chez-a (play) → chez-esh-a (cause to play) chez-a (play) → chez-e-a (play for) chez-a (play) →chez-ean-a (play with each other) chez-a (play) → chez-ek-a (played) pang -a (arrange) → pang -u-a (disarrange) other languages in that the subject of the sentence is the theme/patient of the action. The passive morpheme -w- is part of the verb as in the example below (16). The theme becomes the grammatical subject, whereas the logical subject appears optionally in the na (by) phrase at the end of a sentence. (16) (a) A-li-ni-piga ACTIVE s/he-PAST-me-hit ‘S/he hit me’ (b) Ni-li-pig-wa (na yeye) PASSIVE I-PAST-hit-PASSIVE (by him/her) ‘I was hit (by him/her)’ The Final Suffix The final suffix, also known as the verb-ending suffix, is an obligatory component of the Swahili verb paradigm. The final suffix has generally been described as a mood vowel in Swahili and other Bantu languages. There are three distinct forms of the final suffix in most Bantu languages, including Swahili: [a], [e] 1.3. BILINGUALISM/MULTILINGUALISM AND APHASIA 23 and [i]. In Swahili, this morphological distinction is associated with the corresponding semantic distinctions: the -a is associated with indicative/imperative mood, the -e- is associated with subjective mood, and the -i- associated with negation (Bresnan & Mchombo, 1987). Most Swahili verb stems (verbal root + final suffix), with the exception of loan verbs, are in indicative/imperative mood. The Swahili imperative verb form could function as either a simple command or an appeal directed to the second person singular wewe (you) or plural ninyi (both/all of you). When addressed to one person, the verb stem alone forms the imperative; however, when more than one person is addressed, a pluralizing suffix -ni- is added to the verb stem and the final -a- changes to -e- as shown in example (17). (17) Kula! (‘eat!’) SINGULAR Kule-ni (‘eat!’) PLURAL The Swahili subjunctive verb form may express a wish, desire, request, suggestion, obligation, purpose or intention. It usually consists of the subject prefix, the verbal root and the final vowel -e-(e.g. Ni-saidi-e! (‘help me!’) or A-end-e! (‘let him go!’). The negation final vowel -i expresses negation with no particular reference to time (Mohammed, 2001). It must, however, correspond to the negative subject prefix in accordance with the agreement principle as illustrated in (18) for the verb pik-a (cook). (18) Si-pik-i (‘I don’t cook’) Ha-tu-pik-i (‘we don’t cook’) Hu-pik-i (‘You don’t cook’) Ha-m-pik-i (‘You (pl) don’t cook’) Ha-pik-i (‘S/he doesn’t cook’) Ha-wa-pik-i (‘They don’t cook’) The following sections discuss the bilingual/multilingual situation in Kenya from which the participants in this study are drawn, and provide as an overview of relevant studies in bilingual/ multilingual aphasia. 1.3 Bilingualism/multilingualism and aphasia The two terms bilingual and multilingual are used interchangeably in this study to refer to speakers who use two or more languages or dialects in their everyday lives (Grosjean, 1994). Non-standard dialects of languages were previously 24 CHAPTER 1. INTRODUCTION not considered to be distinct languages for purposes of discussing bilingualism. However, this has turned out to be mere cultural bias. Fabbro (1999) clarifies that the distinction between language and dialect is at times only political, hence not relevant to linguistic or neurolinguistic research and practice. Presently, most books on bilingualism and multilingualism point out that more than half of the world’s population is bilingual or multilingual (e.g., Albert and Obler, 1978; Grosjean, 1994; Fabbro, 1999). In Kenya, the country in which this project was conducted, the majority of the citizens are either bilingual or multilingual speakers of Swahili and English in addition to one or more of the 42 native languages. Like bilinguals in other societies, Kenyans acquire and use their languages for different purposes, in different domains of life and with different people. Both Swahili and English share the same status as second languages, acquired after the age of four, after the native language acquisition by the majority of Kenyans. The English language serves as the official language of instruction in all institutions of learning from primary school to university; it is also the language of news broadcasts, parliamentary proceedings and business. The Swahili language, on the other hand, is the national language taught as one of the subjects from kindergarten to university; it is the language of politics, business and daily interactions of ordinary people from different ethnic backgrounds. Both languages are obligatory and examinable subjects in all primary and secondary schools in the country. The additional 42 native languages are acquired and used in rural regions within ethnic groups spread across the eight provinces of Kenya. Speech and language therapy for aphasia patients in Kenya poses a great challenge for clinicians who unfortunately can usually only communicate effectively in English. This is due to the fact that the clinicians are usually from other countries; hence they only focus on the English language of the patients during therapy sessions. This is contrary to a declaration of the first international symposium on bilingual aphasia captured in a book that was edited by Michel Paradis (1995): “Given what we know, it is no longer ethically acceptable to assess aphasic patients on the basis of the examination of only one of their languages” (p.219). Fabbro (1999) further argued that it is wrong for clinicians engaged in diagnosis and therapy of language disorders to disregard the principles of bilingual aphasia, given the reality that about half of the world’s population is bilingual. 1.3. BILINGUALISM/MULTILINGUALISM AND APHASIA 25 For researchers, bilingual / multilingual individuals with aphasia provide a unique opportunity to investigate the effect of brain damage in different linguistic systems simultaneously. Currently, a considerable number of studies on aphasia in bilingual populations exist (Albert & Obler, 1978; Paradis, 1995; 2001; Fabbro, 1999). However, no study has examined Swahili-English bilingual aphasic speakers in Kenya. The next section gives an overview of some relevant studies on bilingual aphasia. 1.3.1 Bilingual/multilingual aphasia The studies on bilingual aphasia reported in the aphasiology literature have shown that bilinguals with brain damage affecting language areas: (1) may lose the ability to use all the languages they mastered and exhibit the same type of aphasia in all languages; (2) may exhibit language disorders selectively affecting one language, known as ‘selective aphasia’; (3) may exhibit different types of aphasia affecting their different languages, known as ‘differential aphasia’; or (4) may exhibit pathological switching and mixing of their known languages. A crucial finding in bilingual aphasia studies is that for a vast majority of multilingual individuals, the patient will suffer from the same type of aphasia in all languages, to a degree proportionate to their pre-morbid proficiency (e.g., Charlton, 1964; Karanth, Ahuja, Nagaraj, Pandit & Shivashankar, 1991). When it comes to the recovery pattern, it has been mentioned that the language to return first is most often the one used around the time of the aphasiaproducing incident or the first-learned language (Obler & Albert, 1996; Obler & Mahecha, 1991). Prins and Bastiaanse (2006) described one of the first cases of bilingual aphasia as reported by Johannes Wepfer, a German army surgeon, in 1683. Wepfer’s stroke patient, R.N.N., was a 53-year-old bilingual speaker of Latin and German diagnosed with anomic aphasia that lasted between 2 to 3 days. After a month, however, R.N.N., exhibited other aphasic symtoms that were diffult to classify. He could not recall names of others nor express his own name, he had word finding difficulties (especially “little words”), his words were incomplete, and most often his utterances were “alien and incoherent” (paraphasias) and violated the rules of grammar and word order by “pre-posing one word to another”, especially in German. Based on these symptoms, Whitaker (1998) considers this patient to have suffered from agrammatic aphasia, wheareas Prins and Bastiaanse (2006) argue that this might have been a case of fluent 26 CHAPTER 1. INTRODUCTION aphasia with paragrammatism. The first case of qualitatively differential aphasia was reported by Albert and Obler (1978) who described a 35-year-old bilingual aphasia patient exhibiting Broca’s aphasia in Hebrew and Wernicke’s aphasia in English. The authors claim that the patient understood but hardly spoke Hebrew, and that it was the reverse for English. They further reported that the patient’s other languages, Hungarian and French, were mildly impaired. The authors therefore concluded that there must be a separate cerebral localization of English and Hebrew in this patient, suggesting the possibility of separate neuroanatomical representations for languages known by bilingual/multilingual speakers. A case of selective aphasia was reported a few years later by Paradis and Goldblum (1989). They presented a bilingual aphasia patient who exhibited selective aphasia in one of his languages after neurosurgery. The patient’s first post-operative assessment indicated he had Broca’s aphasia in Gujarati, whereas both French and Malagasy were intact. A follow up assessment some years later showed full recovery of all languages. The authors concluded that the most reliable interpretation of selective aphasia proceeds from neurofunctional grounds rather than neuroanatomical ones, as proposed by Albert & Obler (1978). Paradis and Goldblum (1989) likewise argued that the languages known by bilinguals separate primarily at the functional level, with each language using a distinct and independent neurofunctional system. The first case of pathological mixing of the languages known by a bilingual aphasia patient was reported by Stengel and Zelmanowicz (1933). They presented the case of a 57-year-old Czech patient named M.M. who was competent in both Czech and German prior to her aphasia onset. M.M. was admitted to a neurological clinic due to episodes of ‘cephalea’ (pain in the head) and word finding difficulties. On assessment, she showed very poor verbal expression, marked by grammatical errors and numerous phonemic and verbal paraphasias. She was also mixing German and Czech, in which neither of the two languages prevailed over the other. When M.M. could not find a word in one language, she used the corresponding word in the other language without hesitation. M.M.’s comprehension of both languages was found to be intact, even during complex instructions. Having briefly highlighted a few cases of aphasia in bilingual/multilinguals, an account of agrammatism in bilingual aphasia is provided in the following section. 1.3. BILINGUALISM/MULTILINGUALISM AND APHASIA 1.3.2 27 Agrammatism in bilingual/multilingual aphasia Grammatical deficits in aphasia are known to depend on the structure of each language system (Paradis, 1988). Grodzinsky (1991; 1999) argues that the pattern of omissions and substitutions of grammatical morphemes that characterises agrammatism is dependent on the morphological structure of the language. In languages where omission of bound grammatical morphemes results in a word, agrammatic speakers are likely to omit the grammatical morphemes, whereas in languages where omission of bound grammatical morphemes results in a non-word, the agrammatic speakers are likely to substitute the morpheme. Alajouanine (1963) also argued that aphasic speakers of a language with vulnerable inflectional morphology show outstanding signs of agrammatism. According to Fabbro (2001), the differences in manifestation of agrammatism in the different languages of bilinguals are insignificant and occur largely at the surface level. He argues that the general characteristics of agrammatism are universal and that most often, aphasia affects the grammatical aspects of a language at different levels of severity. He reported a study in which 20 bilingual agrammatic speakers of Friulian (morphologically-rich) and Italian were investigated. 13 patients (65%) showed similar impairments in both languages, 4 patients (20%) showed greater impairment of L2, while 3 patients (15%) showed a greater impairment of L1. The parallel impairment in both languages confirms Paradis’ (1988) hypothesis that grammatical disorders in aphasia depend on individual language structure. Fabbro’s further analysis revealed that agrammatic Friulian and Italian speakers had similar grammatical errors in both languages, with only minor differences in areas where the two languages differ. The patients frequently omitted free-standing grammatical morphemes and substituted bound grammatical morphemes in both languages. The most remarkable difference was found to be the proportion of omissions of obligatory pronouns in Friulian (38.25%) vs. Italian (1.25%) which was attributed to the particular grammatical structures of the two languages. For instance, in Friulian the subject is always accompanied by an obligatory pronoun (Il frut al bef : ‘The boy he drinks’), whereas in Italian the subject pronoun is dropped (Il bambino beve: ‘The child drinks’ or beve: ‘ he/she drinks’). Consequently, the bilingual agrammatic speaker can easily avoid structures with pronouns in Italian, but not in Friulian (Fabbro & Frau, 2001). The present project focuses on grammatical aspects in Swahili-English bilingual agrammatic speakers to determine how agrammatism manifest itself in two 28 CHAPTER 1. INTRODUCTION morphologically different languages. The next section provides an overview of the relevant studies on verb deficits and word order in agrammatism, with corresponding linguistic accounts. 1.4 Verb and word order deficits in agrammatic aphasia Verb and word order deficits have frequently been reported in aphasia literature, however, there is no consensus the underlying deficit in agrammatic aphasia. In general, verbs have been found to be more vulnerable than nouns in Broca’s aphasia (Berndt, Mitchum, Haendiges & Sandson, 1997; Miceli, Silveri, Villa & Caramazza, 1984; Semenza, Luzzati & Carabelli, 1997; Jonkers & Bastiaanse, 1998; Kim & Thompson, 2000) although other studies have found that both nouns and verbs may be equally impaired (Basso, Razzano, Faglioni & Zanobio, 1990; Berndt & Haendiges, 2000; Caramazza & Hillis, 1991), at least in some patients. One of the key findings in recent studies focusing on verb and word order deficits in agrammatic aphasia is that verb inflections for tense and/or time reference in both spontaneous speech and controlled experiments are difficult, and that derived word order sentences, especially in controlled experimental studies, are impaired. Studies in different languages have shown tense to be particularly vulnerable in sentence production and sentence anagram tasks for agrammatic speakers (Dutch: Bastiaanse, 2008; Kok, van Doorn & Kolk, 2007; English: Arabatzi & Edwards, 2002; Dickey, Milman & Thompson, 2005; Faroqi-shah & Thompson, 2007; German: Burchert et al., 2005; Wenzlaff & Clahsen, 2004; Greek: Nanousi et al., 2006; Stavrakaki & Kouvava, 2003; Varlokosta, Valeonti, Kakavoulia, Lazaridou, Economou & Protopapas, 2006; Hebrew: Friedmann & Grodzinsky, 1997; Turkish: Bastiaanse et al., 2011; Yarbay Duman & Bastiaanse, 2009). However, agreement inflection has also been reported to be affected in studies of some languages (Dutch: Bastiaanse 2008, English: Lee et al., 2008, German: Burchert et al., 2005). From this body of research, the authors provide several explanations for the tense problems which are captured in aforementioned linguistic theories of agrammatism. The first group of authors relates the problem to the position of tense in the syntactic tree, which is assumed to be pruned from the tense node upwards 1.4. VERB AND WORD ORDER DEFICITS IN AGRAMMATIC APHASIA 29 (Friedmann, 2002; Friedmann & Grodzinsky, 1997). Several other studies, however, have challenged this view by showing different patterns of impairments involving different parts of the syntactic tree. The second group of authors argues against the missing parts of a syntactic representation, and suggests that it is instead the interpretable features of tense that are underspecified (Burchert et al., 2005; Nanousi et al., 2006; Wenzlaff & Clahsen, 2004). Burchert and colleagues (2005) point out that both tense and agreement may be underspecified for patients with agrammatism. The final group of authors suggests that the problem with tense is neither due to its position on the syntactic tree nor an issue of underspecification, but is rather related to time reference (Bastiaanse, 2008; Bastiaanse et al., 2011; Yarbay Duman & Bastiaanse, 2009). This is because there is evidence that not all forms of tense (past, present and future) are equally affected, and that nonfinite verb forms are affected as well. The problem arises mainly when expressing time reference to the past through tensed verbs or untensed participles. They argue that the selective problem with tense results from affected discourse syntax processing, rather than its position in the syntactic tree or being underspecified. This theory is further explored extensively in chapters 2 and 3. Studies in a number of languages have also shown that derived order sentences are problematic for agrammatic speakers (English: Caramazza & Zurif, 1976; Grodzinsky, 2000; Bastiaanse & Thompson, 2003; Bastiaanse & Edwards, 2004; Dutch: Bastiaanse et al., 2002; 2003; Bastiaanse & Van Zonneveld, 2005; German: Burchert, Meisner & De Bleser, 2008; Turkish: Yarbay Duman & Bastiaanse, 2007; Yarbay Duman, Altınok, Özgirgin & Bastiaanse, 2011). Sentence structures such as passives, object relatives, and sentences with scrambled objects or moved verbs have been found to cause problems in both comprehension and production. There are different explanations as to why agrammatic speakers have difficulties with derived order more than base order sentences. One set of researchers attributes this difficulty to compromised mapping of thematic roles (agent, theme) onto grammatical roles (subject, object, byphrase: Linebarger, Schwartz & Saffran, 1983; Schwartz, Linebarger, Saffran & Pate, 1987). Another group of researchers suggest that the problem arises from missing traces in agrammatic language representation (Grodzinsky 1995; Drai & Grodzinsky 2006). Traces are crucial for assigning thematic roles to sentence arguments which are not in their base positions. The authors assert that these traces are not available to agrammatic speakers; the speakers thus apply a de- 30 CHAPTER 1. INTRODUCTION fault strategy by assigning thematic roles to noun phrases strictly on the basis of their position in the sentence. However, several studies have contradicted this theory by demonstrating that traces are actually intact in agrammatic linguistic representation (Dickey, Choy & Thompson, 2007; Thompson & Choy, 2009; Choy & Thompson, 2010). Finally, some researchers argue that the derived order sentences are a problem for agrammatic speakers, not only in comprehension but also in production, because constituents that naturally belong together are separated from one another in derived order sentences (Bastiaanse & Van Zonneveld, 2005; Bastiaanse & Thompson, 2003; Yarbay Duman et al., 2007; 2011). In chapters 4 and 5, we provide two studies that tested this theory and also present an extensive discussion on this topic. In summary, it is clear from this brief review that there is a consensus among researchers that verb inflection and word order pose great challenges to agrammatic speakers. The linguistic accounts reviewed earlier in this chapter make several predictions regarding agrammatic speakers’ performance based on data largely gathered from monolingual populations. While the TPH, TUH and TAUH predict a general problem with tense inflection, the PADILIH predicts a selective problem with past time reference through verb inflection, irrespective of whether it is expressed through tense or aspect or whether finite verbs or periphrastic verb forms are used. Moreover, both the TDH and DOP-H predict problems with sentence structures whose arguments are in derived order. While TDH predicts problems only in comprehension, DOP-H predicts problems in both comprehension and production of sentences with a derived order of arguments. The question is: what kind of pattern will emerge in bilingual speakers of two morphologically different languages? Despite all that has been learned in recent years, the linguistic account that gives the exact nature and a precise characterization of the underlying verb and word order deficits in agrammatic aphasia is still a matter of controversy. The next section presents the research questions that guided the current project. 1.5 The Research Questions Apart from the fact that the studies on verb and word order deficits in agrammatic aphasia reviewed in the previous sections are predominantly monolingual aphasia studies, the languages examined so far are mainly Indo-European languages. It is against this backdrop that we proposed a study that involves bilin- 1.5. THE RESEARCH QUESTIONS 31 gual agrammatic speakers, who, moreover, speak English and Swahili; that is, an Indo-European and a Bantu language. These two languages, contrasting in terms of verbal morphology, provide a valuable site to investigate how agrammatism manifests itself in this bilingual population in light of the linguistic theories discussed. The following questions are addressed in this thesis: 1. Are there similarities or differences in the speech production of monolingual and bilingual aphasic speakers of English? 2. What are the features of Swahili agrammatic speech? Are they comparable with those of English agrammatism? 3. What are the agrammatic speakers’ patterns of production of verb morphology for tense and time reference in Swahili and in English? Are verb forms referring to the past impaired? 4. Is production of time reference to the past through verb inflection impaired in both languages of Swahili-English bilinguals? 5. Is comprehension of time reference to the past through verb inflection impaired in both languages of Swahili-English bilinguals? 6. Is reference to the past similarly impaired in both Swahili and English? 7. Is comprehension of sentences whose arguments are in a derived order impaired in both languages of Swahili-English bilinguals? 8. Is production of sentences whose arguments are in a derived order impaired in both languages of Swahili-English bilinguals? These questions are addressed in the four studies described in this thesis. The first study, presented in Chapter 2, focuses on the first three questions listed above. It identifes the features of Swahili agrammatic speech and compares the use of verb inflections for tense and time reference between agrammatic and non-brain-damaged individuals. The second study, addressing questions 4-6, focuses on time reference problems in both languages of SwahiliEnglish bilingual agrammatic speakers. Both the third and fourth studies examine word order problems in Swahili-English bilingual agrammatic individuals. While the third study addresses sentence comprehension difficulties 32 CHAPTER 1. INTRODUCTION (question 7), the fourth study addresses sentence production problems (question 8). The final chapter discusses the major findings of the thesis and issues arising from the studies. It concludes with a discussion of the clinical implications of the study and recommendations for future research. CHAPTER 2 Characteristics of Swahili-English Bilingual Agrammatic Spontaneous Speech and the Consequences for Understanding Agrammatic Aphasia Abstract Most studies on spontaneous speech of individuals with agrammatism have focused almost exclusively on monolingual individuals. There is hardly any previous research on bilinguals, especially of structurally different languages; and none on characterization of agrammatism in Swahili. The current study identifies the features of Swahili agrammatic narrative and spontaneous speech, and compares the use of verb inflections for tense and time reference in English and Swahili in six bilingual agrammatic speakers and 12 non-brain-damaged speakers matched on age, native language and education level to the agrammatic speakers. The results show a remarkable similarity between the agrammatic phenomena in both languages on the typical agrammatic features: utterance This study has been published as: Abuom, T. & Bastiaanse, R. (2012). Characteristics of Swahili-English bilingual agrammatic spontaneous speech and the consequences for understanding agrammatic aphasia. Journal of Neurolinguistics 25, 276-293. 33 34 CHAPTER 2. CHARACTERISTICS OF SWAHILI-ENGLISH AGRAMMATISM length and speech rate are reduced and the proportion of grammatical sentences and complex sentences is lower than normal. Analysis of verb inflection demonstrates that there are no qualitative differences between the Swahili and English samples: in both languages reference to the past is more impaired than reference to the present. However, the use of verb inflection in general is better preserved in Swahili than English. English verb inflections are frequently omitted, whereas Swahili verb inflections are substituted. The implications of these findings for theories on agrammatism are discussed. 2.1 Introduction In the neurolinguistic literature, several studies report how agrammatism manifests itself in the spontaneous speech of individuals with Broca’s aphasia (see for English: Menn & Obler, 1990; Saffran, Berndt & Schwartz, 1989; Thompson, 2003; for Dutch: Bastiaanse & Jonkers, 1998; for Italian: Miceli, Mazzucchi, Menn & Goodglass, 1983; Rossi & Bastiaanse, 2008; for Indonesian: Anjarningsih, 2012; Anjarningsih & Bastiaanse, 2011). An overview of the findings of these studies consistently reveals a pattern of linguistic deficits: slow, halting, labored, non-fluent, telegraphic speech with omissions and/or substitutions of bound and free grammatical morphemes, while comprehension is relatively spared. The usage of verb morphology is impaired: verb inflections are often omitted or substituted and non-finite verbs are overused (Saffran et al., 1989; Bastiaanse, Hugen, Kos & Van Zonneveld, 2002). While this body of research has contributed substantially to our understanding of the spontaneous speech of aphasic speakers, it has focused almost exclusively on monolingual individuals; there is hardly any previous research on bilinguals, especially of structurally different languages. The current study focuses on agrammatism in bilingual speakers of Swahili and English, two languages which possess contrasting morphological and syntactic properties. Considering that languages differ largely in terms of grammatical morphology, the same underlying deficit may cause different surface manifestations in the different languages of a bilingual (Paradis, 1988). Given this assumption, the aim of the current study is three-fold. First, we would like determine whether there are similarities or differences in English speech production in monolingual and bilingual aphasic speakers. This will determine whether the analysis of spontaneous speech in languages for which 2.1. INTRODUCTION 35 aphasia tests are not available is sufficient to identify agrammatism. Second, we want to identify the features of Swahili agrammatic speech and compare them with those of English agrammatic speech. The final goal is to compare the agrammatic speakers’ production of verb morphology and time reference in both Swahili and English. As these languages differ significantly in their morpho-syntax, these data provide a valuable site for testing the claim that reference to the past is impaired. We will determine whether reference to the past is impaired as has been shown in several experimental studies. We start by giving some background on cross-linguistic variation in aphasia symptoms. Next, we provide a brief overview of bilingual aphasia. This will be followed by a neurolinguistic characterization of agrammatic speech and an introduction to relevant features of Swahili. We conclude by providing the research questions and expectations of the current study. 2.1.1 Cross-linguistic variations in aphasia symptoms Aphasia research across languages has led to some important assumptions concerning the language-specific factors that influence how the same underlying deficit may cause different surface manifestations of agrammatism in different languages. Such factors are related to linguistic complexity, the semantic importance of a morpheme, and to whether or not uninflected forms are permitted. Menn and Obler (1990) presented data from a cross-linguistic study on agrammatic production in 14 different languages. Their key findings are that grammatical morphemes such as plural endings, past-tense endings, and auxiliary verbs are generally prone to errors. Free grammatical morphemes tend to be omitted, whereas bound grammatical morphemes are substituted in agrammatic production. In a language such as English, agrammatic speakers tend to omit word endings such as markers of plural, past tense, and verb 3rd person singular. However, in a language like Italian where every noun, verb, adjective has an ending, agrammatic speakers tend to make substitution errors rather than omission errors. They concluded that the presence or omission of grammatical morphemes is determined by their semantic importance; and that grammatical morphemes which are part of an extensive paradigm are likely to be prone to more errors. Grodzinsky’s (1991; 1999) and Paradis’ (1995), however, argue that grammatical deficits in aphasia, specifically the omission and / or substitution of grammatical morphemes that characterise agrammatism, depend on the structure of each language. In languages where omission of 36 CHAPTER 2. CHARACTERISTICS OF SWAHILI-ENGLISH AGRAMMATISM bound grammatical morphemes results in a word (as in the case of English, for instance ‘-ed’ in picked ), agrammatic speakers are likely to omit the grammatical morphemes, whereas in languages where bound grammatical morphemes results in a non-word the agrammatic speakers are likely to substitute the morpheme with another (as in Swahili for instance, ‘ku’ and ‘-a’ in ‘ku-pig-a’ ‘to hit’, ‘pig’ is a non-word in Swahili). In other words, the manifestation of grammatical errors depends on whether or not the uninflected forms are permitted. Fabbro (2001) proposes that the nature of agrammatism has a universal character and follows the rule whereby aphasia impairs all grammatical aspects of a language, even if at varying degrees of severity. 2.1.2 Bilingual aphasia The term ‘bilingual’ has been used in this study to refer to all those people who use two or more languages or dialects in their everyday lives (Grosjean, 1994). Presently, most work on bilingualism and multilingualism suggests that more than half the world population is bilingual or multilingual, and that the vast majority of bilingual aphasic individuals suffer from the same type of aphasia in all their languages mastered pre-morbidly (for example, Charlton, 1964; De Diego-Balaguer, Costa, Sebastián-Gálles, Juncadella & Caramazza, 2004; Fabbro, 2001; Faroqi-Shah & Waked, 2010; Knoph, 2011). The most relevant issue in bilingual aphasia concerns differences in the pattern of recovery across the languages mastered by the bilingual speakers (see Fabbro, 2001; Paradis, 2001). The common pattern of recovery following the aphasia incident is parallel recovery where all languages of a bilingual are equally affected and recovery is similar. Should there be a difference in recovery; the difference may only reflect a difference in premorbid proficiency level between the languages mastered. Another pattern is differential recovery where the language likely to return first is either one learnt first, or used most frequently prior to the insult (Albert & Obler, 1996). In our case, the bilinguals in this study acquired English and Swahili early (from around the age of four), and were pre-morbidly highly proficient in all their languages. With respect to manifestations of aphasia across the languages of a bilingual individual, Paradis (1988) argues that in languages which differ largely in grammatical morphology, the same underlying deficit may cause different surface manifestations. Some recent studies show that the rich morphology of one language may be better preserved than the poor morphology of the other 2.1. INTRODUCTION 37 language. Knoph (2011), for example, found the complex Farsi morphology better preserved than the comparatively simpler Norwegian morphology in a Farsi-Norwegian bilingual agrammatic speaker, pre-morbidly proficient in both Farsi and Norwegian. Alexiadou and Stavrakaki (2006) and Abuom, Obler and Bastiaanse (2011) showed the same pattern for pre-morbidly proficient GreekEnglish and Swahili-English bilingual agrammatic speakers respectively: the more complex the morphological paradigm, the better it is preserved. One should keep in mind, however, that morphologically rich paradigms are usually more regular than morphologically simple paradigms. It may thus be the case that it is not a matter of complexity but rather of regularity (Goral, 2011). Bybee (2007) proposes two morphological processes in relation to morphologically rich and complex paradigms: the regular morphological process, which is affixal and often agglitunative in nature; and the irregular morphological process, which often involves changes in the stem or high degree of fusion between stem and affix. Regularity is associated with affixation, whereas irregularity is associated with internal stem change. While affixation has been claimed to be more ‘natural’ and highly frequent in most languages, and hence easier to process in general, internal stem changes (irregularity) are considered less ‘natural’, lexically arbitrary and of low type frequency, and hence more difficult to master (Dressler 1985; Bybee, 2007). In situations of linguistic limitations, such as in children with SLI, bilinguals with (L1) attrition (language loss) and individuals with aphasia, high-frequency linguistic items have been found to be resistant to dissolution and are preferred in oral expression (see Centeno & Anderson, 2011). 2.1.3 The syntactic, lexical and morphological deficits in agrammatic aphasia The spontaneous speech of agrammatic speakers is non-fluent. At the syntactic level, agrammatic speech is characterized by short and / or fragmentary utterances with frequent omissions of function words, such as prepositions, pronouns, articles, particles, conjunctions, and determiners (Menn & Obler, 1990). As a result, significantly fewer words are produced per utterance, generally expressed by a shorter Mean Length of Utterance (MLU: see Rossi and Bastiaanse, 2008; Sanchez, 1996; Thompson, Shapiro, Li & Schendel, 1995; Vermeulen, Bastiaanse & Van Wageningen, 1989). Furthermore, a close scrutiny of the structural quality of the agrammatic productions in previous studies re- 38 CHAPTER 2. CHARACTERISTICS OF SWAHILI-ENGLISH AGRAMMATISM veals two prominent features: the limited use of embeddings in sentences and the production of large proportions of ungrammatical sentences (Saffran et al., 1989; Bastiaanse et al., 2002). Moreover, a clear preference for shorter sentences with verbs with fewer internal arguments than normal has also been found (see Thompson, Lange, Schneider & Shapiro, 1997; Bastiaanse and Jonkers, 1998; Rossi and Bastiaanse, 2008). At the lexical level, it has been reported that nouns are produced to a normal extent, but fewer verbs and / or a lower diversity than normal are produced (e.g. Bastiaanse & Jonkers, 1998; Menn & Obler, 1990; Saffran, Berndt & Schwartz, 1989). However, Lorch (1990) found that the production of verbs and verb-types was spared in some of the agrammatic speakers in their cross-linguistic study of Hindi, Icelandic and Finnish, while Crepaldi, Ingignoli, Verga, Contardi, Semenza, and Luzzatti (2011) did not report a noun-verb dissociation in Italian agrammatic speech. There are also reports of a normal use of verbs in Indonesian agrammatic spontaneous speech (Anjarningsih, 2012), but Indonesian is a language without verb inflection for tense and agreement, which may explain these results. At the morphological level, the problems are clearly visible in verb inflection: the proportion of inflected verbs is low (Saffran, et al., 1989, Thompson et al., 1994), although this inflectional problem may be restricted to finite verbs, that is, verbs that are inflected for Tense, Aspect and Agreement and excluding participles or gerunds (Bastiaanse et al., 2002). Experimental data show that not all verb inflections (Tense, Agreement, Aspect, Mood) are equally vulnerable. According to Friedmann and Grodzinsky (1997) and Wenzlaff and Clahsen (2004), agreement inflection is spared but tense inflection is impaired. Clahsen and Ali (2009) showed that Mood is spared and Nanousi, Masterson, Druks and Atkinson (2006) showed that in Greek Aspect is impaired, whereas Tense and Agreement are relatively spared. And recently, our group reported for several languages that it is particularly reference to the past, by Tense and Aspect, that is impaired in agrammatism (Abuom et al. 2011, for English; Bastiaanse 2008 and Jonkers & De Bruin 2009, for Dutch; Bastiaanse, Bamyaci, Hsu, Lee, Yarbay Duman & Thompson 2011, for Chinese, English and Turkish; Yarbay Duman & Bastiaanse 2009 for Turkish). Little is known about reference to the past in agrammatic spontaneous speech; however, a few studies specifically mention problems with the time 2.1. INTRODUCTION 39 reference of verb inflection. Simonsen and Lind (2002) described a Norwegian agrammatic speaker who used very few verbs and hardly any past tense forms in his spontaneous speech. Similarly, Stavrakaki and Kouvava (2003) reported on two agrammatic Greek speakers who have problems producing verb forms with perfect aspect. Beeke, Wilkinson and Maxim (2003a) analyzed the conversation data of a monolingual agrammatic speaker of English who, they report, frequently produced either a present tense verb or an infinitive in place of a future tense verb or past tense verb. Note that both past tense verbs and perfect aspect verbs refer to events taking place or performed in the past. These data underlie the PAst Discourse LInking Hypothesis (PADILIH). This hypothesis is based on Avrutin’s (2000; 2006) idea that linking at the discourse level is done by ‘discourse syntax’ and linking at the sentence level is done by ‘narrow syntax’. According to Avrutin (2000; 2006) processing by discourse syntax is difficult for individuals with agrammatic speakers compared to processing by narrow syntax. Discourse linking requires more processing resources which agrammatic individuals lack. According to Zagona (2003), present tense is locally bound (or in Avrutin’s terminology, is processed by ‘narrow syntax’), since the time of the event and the time of speaking coincide. Past tense, however, is discourse linked (in Avrutin’s terminology, is processed by ‘discourse syntax’). Bastiaanse et al. (2011) modified Zagona’s (2003) theory and state that it is not past tense that requires discourse linking, but all verb forms that refer to the past, including, for example, the present perfect. For instance, in a sentence such as ‘the boy has written a letter’; the event took place in the past, although ‘has’ is present tense. The PADILIH, thus, predicts that Kenyan bilingual agrammatic speakers will have more problems producing verb forms referring to the past than verb forms referring to the future. 2.1.4 Relevant features of Swahili Swahili is a Bantu language, highly agglutinative and classified as an SVO language (Ashton, 1982). It is widely spoken in most African countries, including Kenya where this study was conducted. Kenya is a multilingual society with an average person speaking at least three languages. The two most dominant languages across the population are Swahili and English, but most Kenyans speak one of the 42 languages linguists term “ethnic languages” at home as well. Each of the 42 languages is classified either as Bantu, Nilotic, or Cushitic. Swahili and English have the same status as second languages since both are ac- 40 CHAPTER 2. CHARACTERISTICS OF SWAHILI-ENGLISH AGRAMMATISM quired around the age of 4, after native language acquisition by the majority of Kenyans. Swahili is the national language taught as one of the subjects from kindergarten to university; it is the language of politics, business, and daily interactions of people from different ethnic backgrounds. English is used as the official language of instruction in all educational institutions from primary school to university; it is also the language of news broadcasts, parliamentary proceedings, and business. Therefore, an adult with over 12 years of uninterrupted education in Kenya is generally expected to be equally highly proficient in both languages. Swahili has a fixed word order (SVO) at the sentence level, where the subject precedes the verb and the object (see 1). Within constituent phrases, modifiers follow the head. Therefore adjectives, pronouns, determiners etc., follow the nouns they modify while adverbs come after the verb. As in English, Swahili clause structure generally consists of a subject and a predicate. The subject always precedes the predicate (Mohammed, 2001). Swahili has three different types of sentences: a simple sentence that consists of a single clause; a complex sentence that consists of one main clause and at least one subordinate clause which obligatorily follows the main clause, and a compound sentence that consists of at least two main clauses joined by a coordinating conjunction. However, the most distinguishing feature of Swahili is its verbal system. (1) Mama a-li-m-piga kijana Mother she-Past-him-hit boy “The mother hit the boy” The Swahili verbal system is distinctly more complex than that of English, consisting of numerous affixes, both inflectional and derivational morphemes, attached to the verb root. These affixes (prefixes, infixes and suffixes) occupy specific positions and perform specific functions. The general position scheme of the affixes in relation to the verb root is shown in (1a). Some illustrations from Abuom et al. (2011) are given in (2b-d). (2) (a) Pre-prefix (Pp) + Subject prefix (Sp) + Tense marker (T) + Object infix (Oi) + ROOT + derivation (d) + Suffix (s) + Post-suffix (Ps). 2.1. INTRODUCTION (b) 41 A + li + m + pig + a Sp + T + Oi + ROOT + d “S/he hit him/her” (c) Ha + tu + ta + m + pig + a Pp + Sp + T + Oi + ROOT + d “We will not hit him/her” (d) Tu + na + pig + a + n + a Sp + T + ROOT + d + S + Ps “We are hitting each other” A Swahili verb complex, due to its agglutinative feature unlike that of English, can function as a complete sentence as illustrated in (2b-d). In the verbal complex, the subject prefix (subject-verb agreement), the tense marker (includes tense and aspect) and the verb root are generally obligatory in every grammatical Swahili sentences. Both the subject prefix and object infix must agree in number with subject and object of the sentence respectively, for instance in (1).Tense and aspect inflections are critical to time reference in Swahili. As illustrated in (2b-d), Swahili has three marked tenses: past, present and future. The past tense marked by ‘li’ describes a past activity without reference to a specific time as in (2b). The future tense marked by ‘tu’ describes events predicted to follow the present time as in (2c). English, on the contrary, has no verb inflection for future tense. The Swahili present tense is marked by either ‘-a-’ or ‘-na-’ : both can used interchangeably to describe actions that take place at the time one is speaking as in (2d). Other inflections that are considered to be ‘neutral’ tense markers are the infinitive, which is generally marked by the prefix ‘ku’ attached to verb root, and the narrative or consecutive marker ‘ka’, which generally express an action or state which follows another action as in (2e). For aspect in Swahili, the present perfect marker ‘me’ is also frequently used to describe a past activity which has relevance to the present time and is comparable to the present perfect in English as in (2e). (2e) A-me-ruka chini a-ka-kimbia ku-cheza mpira He-Pres.perf.-jump down he-Cons.-ran off INF.-play football “He has jumped down, he (then) ran off to play football” 42 CHAPTER 2. CHARACTERISTICS OF SWAHILI-ENGLISH AGRAMMATISM Other types of verbs are the copulas and auxiliaries. Copulas in Swahili, as in English, are known to have little independent meaning, and mainly function to relate sentential elements of clause structure especially subject and complement (3a). Just like in English, Swahili modal verbs accompany lexical verbs (3b), but, unlike in English, Swahili modal verbs are inflected for Agreement. Auxiliaries such as ‘to have’ and ‘to be’ that are used in combination with a participle, do not exist in Swahili. (3) (a) Kenya ni nchi kubwa Kenya is country big “Kenya is a big country” (b) Mgonjwa a-na-weza kunywa maziwa patient he-Pres.-can drink milk “The patient can drink milk” 2.1.5 The research questions and expectations The current study contributes to aphasiology as a field of research in two ways: Firstly, it provides further evidence of features of agrammatism from a previously undescribed language; secondly, it gives an insight into the patterns of manifestation of agrammatism in bilingual individuals speaking two structurally different languages. This has been achieved by addressing three main research questions: 1. Are there similarities or differences in the English speech production of monolingual and bilingual aphasic speakers? 2. What are the features of Swahili agrammatism? Are they comparable with those of English agrammatism? 3. What are the agrammatic speakers’ patterns of production of verb morphology for tense, and time reference in Swahili and in English? Are verb forms referring to the past impaired? First, we compared the English samples of narrative speech of Kenyan bilingual agrammatic speakers with those of monolingual American agrammatic speakers from Thompson, Choy, Holland & Cole (2010). This was done to 2.1. INTRODUCTION 43 determine whether the analysis of spontaneous speech in languages for which aphasia tests are not available is sufficient to identify agrammatism. We expected similar manifestation of agrammatism in English language between the two groups irrespective of number of languages mastered pre-morbidly. To answer the second question on features of Swahili agrammatism, the Swahili and English samples of the bilingual Kenyan agrammatic speakers were analyzed on a number of variables that are relevant to quantify agrammatic aphasia: the mean length of utterances (MLU), speech rate (words per minute), number and diversity of nouns and verbs, the use of copulas and auxiliaries, and the percentages of ungrammatical and embedded sentences. We first made comparisons between samples of agrammatic speech and those of non-braindamaged speakers in Swahili and English. We expected differences between the two groups on each of these variables. We then compared the agrammatic performance on these variables between Swahili and English. If the underlying disorder is the same in both languages, which is likely because we compared within individuals (Fabbro, 2001; Faroqi-Shah & Waked, 2010), it was expected that the reduction in utterance length and the delay in speech rate, as well as the degree in which nouns and verbs are produced would be similar in both languages. Also, we expected the agrammatic speakers to produce more ungrammatical sentences and fewer embedded sentences than non-brain-damaged control speakers in both languages. The final question is how the bilingual agrammatic speakers perform on a variable that is fundamentally different in the two languages. Comparisons were made between the use of verb inflections and time reference morphology. Abuom et al. (2011) showed that, on a production experiment with two bilingual English-Swahili agrammatic speakers, verb inflection was better preserved for Swahili than for English. It was argued that this was caused by the fact that verb inflection is more anchored into the system in Swahili, with its very rich verb inflection paradigm for tense and agreement. In English, the tense and agreement paradigm is poor (restricted to -s, -ed), sometimes irregular and less anchored in the system and, therefore, more vulnerable. Hence, we predict that production of finite verbs (those verb forms that are inflected for tense and agreement) will be relatively spared in Swahili (cf. Goral, 2011). It has also been reported that agrammatic speakers produce fewer verb forms referring to the past than non-brain-damages speakers, in spontaneous speech data (Simonsen & Lind, 2002; Stavrakaki & Kouvava, 2003), conversation data (Beeke, 44 CHAPTER 2. CHARACTERISTICS OF SWAHILI-ENGLISH AGRAMMATISM Wilkinson and Maxim (2003a) and experimental data (Bastiaanse, 2008; Bastiaanse et al., 2011; Faroqi Shah & Dickey, 2009). We, therefore, predict that production of verb forms referring to the past will be reduced, whereas production of present verb forms will be relatively spared. However, as mentioned, in Abuom et al.’s (2011) experiment, the two bilingual English Swahili agrammatic individuals showed intact production of past tense in Swahili, but not in English. This relatively better preservation of verb forms referring to the past in Swahili may be reflected in spontaneous speech. Substitutions of verb inflections are expected to occur in Swahili, whereas omissions are expected in English (Grodzinsky, 1991; 1999; Paradis, 1995). 2.2 2.2.1 Methods and Procedures Participants The participants recruited for this study include 6 non-fluent aphasic/agrammatic speakers and 12 non-brain-damaged speakers (NBDs). Each participant spoke either Bantu or Nilotic language as native language, but all spoke English and Swahili as second languages. They were matched on age, native language, and education (a minimum of O-Level qualification, equivalent to high school diploma). In the Kenyan school system, these are graduates who have gone through kindergarten, elementary (primary school) and high school tiers of the education system, which means 12 years of uninterrupted exposure to English and Swahili. The agrammatic speakers were, therefore, premorbidly highly proficient in the two second languages. All participants were right-handed and without a history of psychiatric or developmental speech or language disorders or any other neurological conditions. The aphasic speech produced was judged as ‘telegraphic’ by a speech therapist: it was perceived as being slow and effortful with short phrases consisting of mainly content words. Unfortunately, in Kenya there are no tests available to establish the aphasia syndrome. However, all agrammatic speakers had good comprehension in both languages on an adapted version of the subtask for auditory comprehension of single words (nouns, verbs, colours, shapes, letters, numbers) from the Boston Diagnostic Aphasia Examination word comprehension test (Goodglass & Kaplan, 1972). A few pictures of this task were substituted, because some items were unknown in Kenya and, hence, no Swahili word 2.2. METHODS AND PROCEDURES 45 was available (for example, a hammock was changed to a swing). It is however important to note that the BDAE cannot be used to classify the aphasia type in Kenya because of its cultural bias. The demographic details of the participants and their scores on the BDAE subtest for auditory word comprehension are shown in the Table 2.1. Table 2.1: The demographic details and the results on the test for auditory comprehension of words (on an adapted version of the BDAE-test in Swahili and English) of the agrammatic speakers and non-brain-damaged speakers. Participant Age Agrammatic LA 43 MM 46 EA 40 HJ 45 JK 49 SW 20 NBD IA 41 DM 49 MK 40 JN 46 NK 45 BK 20 GN 43 GK 44 MC 40 PK 39 BS 21 JR 48 2.2.2 Gender Handed- Education Years Native ness post language stroke F F M F M M R R R R R R 16 16 16 12 17 12 F M M F F M F F M M F M R R R R R R R R R R R R 17 16 16 15 16 12 16 15 17 12 14 16 1 16 17 10 1 2 Word comprehension (maximum score: 72) English Swahili Bantu Nilotic Nilotic Nilotic Bantu Bantu 71.5 72 72 72 72 71.5 72 72 72 72 72 72 Nilotic Bantu Nilotic Bantu Bantu Nilotic Bantu Bantu Nilotic Nilotic Bantu Nilotic 72 72 72 72 72 72 72 72 72 72 72 72 72 72 72 72 72 72 72 72 72 72 72 72 Materials For elicitation of narrative speech, the methods of Olness (2006) were used. Two pictures have been used, the Pulitzer Prize winning photograph by Annie Wells of a girl being rescued from the water by a fireman and the cookie theft picture from the BDAE (Goodglass & Kaplan, 1972). The participant was first asked to describe the picture and then to make a story around it with a beginning, a middle and an end. Olness (2006) showed that when aphasic speakers are asked to tell a story rather than to describe a picture, more variation in 46 CHAPTER 2. CHARACTERISTICS OF SWAHILI-ENGLISH AGRAMMATISM verb forms occurs. Since our main interest was in the use of verbs and verb inflection for time reference, this seemed to be an appropriate method. Additionally, a spontaneous speech interview was held with one question asking for information from the past and one for information in daily life. The questions were: The flood rescue and the cookie theft (a) Can you tell me what is happening in this picture? (b) Can you create a story with the beginning, the middle and the end about what happened in the picture? Spontaneous speech Reference to the past: (a) Can you tell me about how your speech problems started (agrammatic speakers)/ about your last illness (for NBDs)? Reference to the present: (b) Can you tell me about your current work / hobbies? 2.2.3 Procedures Recording sessions were held in a quiet setting for each of the participants using a digital audio recorder. For the agrammatic speakers, the interviews in English and Swahili were done on different days, for the NBDs at different times on the same day. To avoid code switching during recording sessions, the participants were explicitly asked to speak only Swahili or only English. The experimenter was very cautious to use only one language without intrusions from the other. Half of the participants were first tested in English, half first in Swahili. The order of the elicitation was similar for each participant: first the ‘flood rescue’ picture, followed by the ‘cookie theft’. The interview was done last. The questions were always in the same order, that is, the order given above. From each participant all the data were recorded and orthographically transcribed in English and Swahili. For each question, the participants were encouraged to say as much as possible. All the audio-recorded samples were orthographically transcribed 2.2. METHODS AND PROCEDURES 47 verbatim. The sparse number of intrusions from the other language was ignored. After transcription the samples were segmented into utterances. Each utterance was considered a ‘sentence’ based on Hartmann and Stork’s (1972) definition of a clause as a grammatical unit that includes, at the minimum, a predicate and an explicit or implied subject and expresses a proposition. 2.2.4 Analysis 300 words is a reliable sample size for spontaneous speech analysis for languages like English, Italian and Dutch (see Brookshire & Nicholas, 1994). However, Swahili, unlike English, is a highly agglutinative language and much information that is expressed in single words in English is expressed by suffixes in Swahili. Therefore, a comparison of samples of the same size would be inappropriate. A first analysis of both the NBD samples showed that for speech rate and MLU the ratio of English and Swahili is around 3:2, meaning that the information conveyed in 300 words in English is given in around 200 words of Swahili. Therefore, 200 words of the Swahili samples were analyzed. To determine whether speech which was characterized as ‘agrammatic’ for English in bilingual participants by the SLTs was similar to agrammatic speech of monolingual speakers of English, a comparison was made between the American English agrammatic data of eleven monolingual speakers from Thompson et al. (2010) and the Kenyan English samples. The following variables were analyzed: • speech rate (number of words produced per minute) • mean length of utterances (total number of words divided by the number of utterances). • noun to verb ratio • percentage grammatical sentences Subsequently, the English and Swahili samples were then compared on the following variables: • mean length of utterances and speech rate • number and diversity of nouns and verbs and number of copulas and (modal) auxiliaries 48 CHAPTER 2. CHARACTERISTICS OF SWAHILI-ENGLISH AGRAMMATISM • percentage grammatical sentences and proportion of embedded / complex sentences Finally, for the analysis of verb inflection and time reference through verb forms, we analyzed: • number of verb inflections for tense and agreement on the lexical verbs • time reference through verb inflection (number of verb forms referring to the present and the past, including copulas, auxiliaries and modals, taking both tense and aspect into account6 ) We ensured a balanced representation of all the questions by extracting an equal number of words from every section (flood rescue, cookie theft, interview). For statistical analysis, we used non-parametric tests (Mann-Whitney U test was used for between-group comparisons) because of the small group sizes. Transcription procedure and the inter-transcriber reliability All collected spontaneous speech samples were orthographically transcribed verbatim and then segmented into utterances using prosodic features as well as syntactic structure to determine utterance boundaries, and analyzed for lexical, morphological, and structural measures. Minor utterances (e.g., ‘yes it is’, ‘no it isn’t’ for English, ‘sawa sawa’, ‘ndio’ in Swahili), voiced and unvoiced starters, repetitions of words or sentence chunks, non-meaningful words, and fillers or interjections were excluded. For analysis (300 words for English and 200 words for Swahili), the focus was on clauses that contained a verb: utterances containing verb, copula, modal, or auxiliary were analyzed. Initial transcription and segmentation was done by a student of clinical linguistics, a bilingual speaker of Swahili and English, who was trained on this method of analysis. The transcriptions and segmentations were checked for reliability by a second transcriber, the first author (TA), also a bilingual speaker of Swahili and English, and verified by the second author (RB) for English. The same procedure was used for the analysis. The results of the two raters were compared and the disagreements were resolved. 6 For both languages, present tense-imperfective aspect (for English, for example, writes and is writing) was taken as referring to the present and past tense-imperfective aspect, present tense-perfect aspect or past tense-perfect aspect (for English: wrote, has written, has been writing, was writing, respectively) was considered to be reference to the past. 2.3. RESULTS 2.3 2.3.1 49 Results Comparison of the Kenyan English samples with those of American English agrammatism Since we had no test data that characterized Broca’s aphasia, agrammatism in Kenyan English, we compared the English agrammatic data with the pretreatment data of eleven American agrammatic speakers from Thompson et al. (2010). As can be seen from Table 2.2, the mean scores are very similar and so are the ranges. The Kenyan English aphasic speakers talk slowly, in short and simplified sentences which are often incomplete and / or ungrammatical. We, therefore, consider these samples as typically agrammatic. This, in combination with the well-preserved comprehension of single words, justifies the assumption these agrammatic speakers suffer from Broca’s aphasia, at least in English. In the next sections, we will compare their data for both English and Swahili. Table 2.2: The means and ranges compared for the Kenyan English agrammatic samples and the American narrative speech data of Thompson et al. (2010). Kenyan English agrammatic Mean Range MLU Words / minute Noun verb ratio % Grammatical sentences 2.3.2 American English agrammatic Mean Range 6.02 51.5 1.52 2.9-7.7 14-70 0.82-1.63 5.18 46.37 1.34 2.3-7.2 11-83 0.76-2.10 49.5 13-65 46.1 22-75 Comparison of the English and Swahili samples of the agrammatic and non-brain-damaged speakers Mean Length of Utterances and Speech Rate The results reveal a serious fluency reduction in the agrammatic speakers. Their speech rate is low and the utterance length is considerably reduced (see Table 2.3). The differences between the NBD and agrammatic group were significant for both MLU (for English: z = -3.37, p< .001, for Swahili: z = -3.37, p< .001) and speech rate (for English: z = -3.37, p < .001, for Swahili: z = -3.37, p < .001). In fact, for both variables and in both languages, each of the agrammatic 50 CHAPTER 2. CHARACTERISTICS OF SWAHILI-ENGLISH AGRAMMATISM speakers scores outside the normal range. The data reveal that the agrammatic speakers performed very similar in both languages: their MLU was 60.4% of that of NBDs in both English and Swahili and their speech rate was reduced with 61.1% in English and 58.5% in Swahili. The corresponding patterns evident in the two languages confirm that the brain-damaged participants had a quantitatively and qualitatively similar aphasia in both languages. Table 2.3: The comparison of the Mean Length of Utterances (MLU) and speech rate (words per minute). MLU LA MM EA HJ JK SW NBDs’ range English Speech Rate 7.692 6.818 2.941 5.172 6.977 6.522 8.11-10.71 70 63 14 46 61 55 119-140 MLU Swahili Speech Rate 4.255 4.082 2.778 3.636 4.348 3.846 5.41-7.14 65 40 9 50 56 60 75-100 Nouns, Verbs, Copulas and Auxiliaries The analysis of the noun and verb production showed that the agrammatic speakers fell within the normal range in both English and Swahili with only a few exceptions. There were no significant differences between the groups: noun tokens (for English: z = -.33, p >.05, for Swahili: z =-.56, p > .05); noun diversity (for English: z = -.7, p >.05, for Swahili: z = -.37, p > .05); verb tokens (for English: z = -.28, p >.05, for Swahili: z = -.56, p > .05); verb diversity (for English: z = -.23, p >.05, for Swahili: z = -.14, p > .05); copulas (for English: z = -.52, p >.05, for Swahili: z =-.66, p > .05); and auxiliaries (for English: z = -.09, p >.05, for Swahili: z =-.9, p > .05). Some individual agrammatic speakers showed behavior outside the normal range: EA overused nouns in the English sample and produced a reduced number of copulas and auxiliaries in both English and Swahili. MM’s use of auxiliaries was outside the normal range in Swahili. Overall, the production of nouns, verbs, copulas and auxiliaries was quite normal, however. It can also been seen from Table 2.4 that the number of nouns and verbs is very similar in English and Swahili and comparable between the two groups 2.3. RESULTS 51 (except that the mean number of nouns in the agrammatic group is a bit higher for the agrammatic speakers when speaking English, due to the large number of nouns produced by EA). This is interesting, because the English sample size was 300 words and the Swahili sample size 200 words. It shows that, as expected, in English many more function words are used than in Swahili, the latter being a far more agglutinative language. This not only holds for the NBDs, but also for the agrammatic speakers, who are supposed to be poor in the production of function words. Table 2.4: The comparison of noun and verb production between the agrammatic and nonbrain-damaged speakers in English and Swahili. Languages English Swahili Noun-tokens Noun-types Verb-tokens Verb-types Copulas Auxiliaries Noun-tokens Noun-types Verb-tokens Verb-types Copulas Auxiliaries LA MM 37 26 45 28 10 16 35 23 44 28 7 11 57 31 35 20 11 19 58 20 33 15 9 3 Agrammatic speakers EA HJ JK SW 99 45 56 32 0 3 48 26 26 18 2 0 54 27 47 25 9 11 41 23 57 38 1 10 43 26 43 28 12 20 41 24 53 32 8 9 47 26 32 22 21 22 32 16 34 20 21 8 mean NBD range mean 56.2 30.2 43 25.8 10.5 15.2 42.5 22 41.2 25.2 8 6.8 28-63 16-45 31-51 20-32 5-18 8-32 27-58 11-32 23-49 18-34 3-14 5-14 47.6 28.4 41.8 26.3 9.9 16.4 44.8 22.7 38.5 25.2 8.6 9.4 Grammaticality and syntactic complexity Analysis of the grammaticality and complexity of the produced sentences revealed that all the agrammatic speakers produced a significantly larger proportion of ungrammatical sentences (for English: z = -3.37, p <.001, for Swahili: z = -3.37, p < .001) and a significantly smaller proportion of embedded sentences (for English: z = -3.37, p <.001, for Swahili: z = -3.37, p < .001) than the NBDs in both languages (see Table 2.5). Again, the data in English and Swahili are very similar: the agrammatic speakers produce 10 times as many ungrammatical sentences in English and 8.6 times as many ungrammatical sentences in Swahili than the NBDs. The NBDs produce 4.5 times as many embeddings in English and 4.8 times as many in Swahili than the agrammatic speakers. The ungrammatical sentences produced by the agrammatic speakers in both 52 CHAPTER 2. CHARACTERISTICS OF SWAHILI-ENGLISH AGRAMMATISM English and Swahili were mainly incomplete structures (missing internal arguments; frequent omissions of free morphemes such as articles, prepositions, determiners; substitutions of grammatical morphemes, such as inflections, leading to tense and agreement errors). Such frequent omission of elements within a sentence, coupled with limited use of embeddings largely account for the reduced length of utterances as mentioned above. Table 2.5: Proportions of ungrammatical and embedded sentences of the agrammatic and non-brain-damaged speakers in English and Swahili Participants LA MM EA HJ JK SW Mean NBD 2.3.3 mean Range English Ungrammatical Embeddings sentences 18 (46%) 17 (39%) 89 (87%) 32 (55%) 15 (35%) 19 (41%) 31.7 (51%) 1.58 (5.1%) 0%-13% 7 (18%) 6 (14%) 0 (0%) 2 (3%) 4 (9%) 4 (9%) 3.8 (8%) 10.83 (36%) 27%-46% Swahili Ungrammatical Embeddings sentences 15 (32%) 16 (36%) 66 (92%) 20 (36%) 14 (31%) 17 (35%) 24.2 (44%) 1.83 (5.5%) 0%-14% 5 (11%) 3 (7%) 0 (0%) 1 (2%) 4 (9%) 4 (8%) 2.8(6%) 8.75 (29%) 20%-52% Verb inflection and time reference through verb forms Verb Inflections on Lexical Verbs As shown above, the number of lexical verbs produced by the agrammatic speakers was similar to that of the NBDs in both languages. However, in English the agrammatic speakers produce a larger proportion of these verbs as an infinitive and gerund than the NBDs (see Table 2.6). 89% of the lexical verbs produced by the agrammatic speakers were in the infinitive or gerund (ing) form, compared to 74% in the NBD population (z = 3.32, p < 0.0001). For Swahili, a similar comparison cannot be made. Unlike in English, uninflected verb forms in Swahili do not exist. Nonfinite verb forms, that is, verb forms that are not marked for tense and agreement, also bear inflection. These are the infinitives and the consecutive forms, marked with ku- and ka-, respectively. The agrammatic speakers use as many of these non-finite inflected lexical verbs 2.3. RESULTS 53 as the NBDs (z = 1.55; p > 0.05). Table 2.6: The proportions of verb inflections on lexical verbs in English and Swahili produced by agrammatic and non-brain-damaged speakers. English verb inflections: Tense Agreement Infinitives & inflections: inflections: v+ing verb -ed(irr) -s forms LA MM EA HJ JK SW Mean NBD: Mean Range Swahili verb inflections: Tense Agreement Infinitives & inflections inflections Consecutive Markers -li, -na, -a -ni, -a, -wa ku-, ka- 4 (4) 1 (2) 1 (2) 3 (4) 5 (2) 4 (1) 3 (2.5) 0 2 0 2 1 1 1 41 32 53 42 37 27 38.7 26 13 0 15 28 20 17 29 13 1 42 40 31 26 17 12 1 31 22 15 16.3 8.8 (6.1) 6-14(>5) 1.83 1-5 30.9 23-41 25.83 17-42 29.0 17-43 10.50 4-17 (Irr = irregular verbs) In English lexical verbs, only the third person singular, present tense is inflected for agreement (-s) and only the past tense is marked on the verb (-ed). In the NBD samples, 4.5% of the lexical verbs are inflected for the 3rd person singular, comparable to the percentage in the agrammatic speakers (z = -1.36, p > 0.05). However, the agrammatic speakers used significantly fewer tense markers in English than the NBDs (7.4% vs 20.95%; z = -3.32, p < 0.001). In Swahili, all finite verbs are bearing markers for agreement and tense. Therefore, in Swahili the agrammatic speakers (and the NBDs) produce many more verbs inflected for tense and agreement, but the pattern is the same: the agrammatic speakers use a normal percentage of agreement markers (z = -0.98, p > 0.05), but the use of tense markers is significantly reduced (z = -2.39, p < 0.02). In summary, when agrammatic speakers produce lexical verbs in the English condition, they overuse forms that are not inflected for tense and agreement (infinitives and gerunds). The number of finite verbs inflected for third person singular, present tense is normal, meaning that agreement inflection, as far as measurable in English, is intact, although there were only a few instances of agreement inflection in the data set. However, the overt tense marker in English, -ed, in combination with the irregular past tense forms, is used with a lower frequency than normal. In Swahili, uninflected forms of verbs do not 54 CHAPTER 2. CHARACTERISTICS OF SWAHILI-ENGLISH AGRAMMATISM exist and are not produced when the agrammatic speakers produce Swahili. The number of infinitives and consecutive forms is not statistically different from normal. Just like when they are speaking English, agreement inflection is intact, but tense inflection is produced less frequently than normal. Time Reference through verb inflection Tense and aspect inflections are used to refer to a time frame, that is, past, present or future. Since the majority of tense inflections in both languages and in both populations were to present and past, reference to the future was ignored in the current analysis. Not only the tense and aspect inflections of lexical verbs were tallied, but also copulas and auxiliaries. Table 2.7: Production time reference through tense and aspect in English and Swahili by agrammatic and non-brain-damaged speakers. English time reference Past Present (tense & aspect) (tense & aspect) LA MM EA HJ JK SW Mean NBD: Mean Range Swahili time reference Past Present (tense & aspect) (tense & aspect) (-li- & -me-) (na-, -a-) 19 6 6 18 10 16 12.5 27 36 26 25 30 24 28 6 14 0 12 13 5 8.3 26 12 0 10 23 15 14.3 24.5 20-29 15.8 10-21 18.3 16-30 11.9 8-19 As shown in table 2.7, both in English and in Swahili, the individual agrammatic speakers’ use of verb forms referring to the past was outside the normal range and significantly lower than in the NBDs (for English: z = -3.37, p <.001, for Swahili: z = -3.37, p < .001). This was not the case for verb forms referring to the present. In fact there was a substantial overuse of present verb forms by the agrammatic speakers in English (z = -3.37, p = <.001), but not in Swahili (z = -.84, p > .05 ) compared to the NBDs. A further comparison of the use of verb forms referring to the past and those referring to the present within the agrammatic speakers group reveals a preference for verb forms referring to the present over those referring to the past, a difference that is significant for 2.4. DISCUSSION 55 English (z = -2.88, p = <.01), but not for Swahili (z = -1.12, p = > .05). 2.4 Discussion This section discusses three research questions outlined in section 1.5, namely (1) Are there similarities or differences in the English speech production of monolingual and bilingual aphasic speakers? (2) What are the features of Swahili agrammatism? How comparable are they with the features of English agrammatism? (3) What are the agrammatic speakers’ patterns of production of verb inflection for tense and time reference in Swahili and in English? Are verb forms referring to the past impaired? The section will end with a discussion on linguistic complexity and frequency. 2.4.1 Comparisons with American English data The results of the analysis of the Kenyan-English samples were compared to those provided by Thompson et al. (2010) to determine whether analysis of spontaneous speech in languages for which aphasia tests are not available is sufficient to identify agrammatism. The similarities were remarkable: the Kenyan-English samples had the same speech rate, mean length of utterances, noun-to-verb ratios and percentages of grammatical sentences as the American English agrammatic samples of Thompson et al. (2010). This, in combination with the good performance on the BDAE subtest for auditory word comprehension, shows that the spontaneous speech could be characterized as ‘agrammatic’ and that the agrammatic speakers could be classified as suffering from Broca’s aphasia, at least in English, although no formal Kenyan tests are available. Knowing that these aphasic individuals were agrammatic in English allowed us to do additional analyses and to make comparisons with the Swahili samples, to see whether the spontaneous speech in this language can also be qualified as agrammatic. 2.4.2 Comparison of the English and Swahili samples The analysis of Swahili agrammatic samples compared to those of NBDs shows features that generally quantify agrammatism: reduced spontaneous speech out; low speech rate; shorter utterances; limited use of embeddings in sentences; and production of large proportions of ungrammatical sentences. For cross- 56 CHAPTER 2. CHARACTERISTICS OF SWAHILI-ENGLISH AGRAMMATISM linguistic comparisons, we first compared the mean length of utterances (MLU) and speech rate (words per minute). The results showed similar pattern in Swahili and English: a considerable reduction in utterance length and the delay in speech rate is similar in both languages. Second, the use of nouns and verbs (including copulas and auxiliaries) is very much alike in both languages. Somewhat surprisingly, the scores on these variables are not different from normal. On the basis of the literature (e.g., Bastiaanse & Jonkers, 1998; Thompson et al., 1997) it was expected that the agrammatic speakers would produce fewer lexical verbs and / or lexical verb types, as well as fewer copulas and auxiliaries. However, this pattern was not observed in our agrammatic speakers in either language. There are at least two other studies that report the same results. Lorch (1990) studied verb production in three highly inflectional languages (Hindi, Icelandic and Finnish) and showed that agrammatic speakers’ production of verbs and verb-types was not impaired in all speakers. More recently, a study by Crepaldi, Ingignoli, Verga, Contardi, Semenza, and Luzzatti, (2011) did not report a noun-verb dissociation in Italian agrammatic speech. However, looking at the individual data of the current study, we see that EA, who was the most severely aphasic, had difficulties producing both copulas and auxiliaries in the two languages. MM had difficulties with auxiliaries only in Swahili. Our interpretation of the current data is that the agrammatic speakers in this study may not have specific retrieval problems for lexical verbs, at least not in their spontaneous speech. Their ability to retrieve nouns, lexical verbs and other categories of verbs appears to be intact. Interesting is the finding that even though lexical verbs and other verb categories are retrieved normally by these agrammatic speakers, the proper use of the retrieved verbs in sentences is generally impaired leading to higher proportions of ungrammatical sentences. The agrammatic speakers often omitted complements (obligatory and implicit arguments). Finally, as per our expectation, the agrammatic speakers produce more ungrammatical sentences and fewer embedded sentences than NBDs in both languages. Again, there is a similar pattern in Swahili and in English. The agrammatic speakers’ ability to produce grammatically correct sentences that are typically longer and/or complex (with embeddings) is clearly disrupted, confirming the results of previous studies on agrammatism (see section 2.1.3). In sum, the agrammatic symptoms in both languages are very similar on variables that were not expected to yield different results in the two different 2.4. DISCUSSION 57 languages: reduced speech rate, short utterances, high proportion of ungrammatical sentences, and fewer embedded sentences. Unexpectedly, the production of verbs is well-preserved in both languages, which is not common in agrammatic spontaneous speech. Still, the noun-verb ratio of our English samples is comparable to those of Thompson et al. (2010) for American English agrammatic speech. 2.4.3 Verb inflection and time reference The final question addressed the production of verb inflection. Two subquestions were raised. First, whether there would be differences in the use of verb inflections in these typologically different languages and second whether there was a selective disorder for verb forms referring to the past. With regard to the first sub-question, several theories were addressed in the Introduction. In spontaneous speech it is hard to point to obvious errors, but it was clear that those verb forms that were produced were usually correct, in both languages. Thus, it is not the case that more errors are made in Swahili than in English. In English, the agrammatic speakers overused the infinitive and gerund forms of lexical verbs. In Swahili, the infinitive and narrative markers (ku- and ka-) were produced slightly more often than normal, although the difference is not significant. When the finite verbs are compared, we see that in both languages agreement inflection is normal, but the production of tense inflection is compromised, although less in Swahili than in English (English 35% of the normal number of Tense inflections, in Swahili 57%). This is in line with the findings of Knoph (2011) and Alexiadou and Stavrakaki (2006) who also reported that inflection of the morphologically most complex language is best preserved. It also supports the findings of our earlier study in which the English-Swahili agrammatic speakers’ production of past tense forms were impaired in English but not in Swahili (Abuom et al. 2011). These authors explained this by the assumption that inflection for Tense is much more anchored in the Swahili language system, with its very rich and informative paradigm, than in English, in which the lexical verb is not even marked for two time frames (present and future). In the final section we will return to the issue why there is a slightly better performance on Tense in Swahili than in English, regardless of the morphological complexity of the two languages. Grodzinsky’s (1991; 1999) and Paradis’ (1995) claim that omission of grammatical inflections occurs in languages that allow bare forms. They also claim 58 CHAPTER 2. CHARACTERISTICS OF SWAHILI-ENGLISH AGRAMMATISM that substitutions will be produced in languages where omissions result in nonwords. Both claims are supported by our data. Speaking in English results in an overuse of infinitives, whereas speaking in Swahili results in a more than normal number of verbs with ku- and ka- markers. However, in English gerunds are also overused and these forms are inflected (-ing). Thus, the results may also be explained in a different way: in both languages the agrammatic speakers show a preference for non-finite verbs, whether or not they are inflected. This is more in line with the idea of Fabbro (2001), who expects the same errors in both languages. This seems, indeed, to be the case: in both languages, the agrammatic speakers refrain to non-finite verbs, that is, verbs that are not inflected for tense and agreement. Why would agrammatic speakers use these forms? Obviously, it is not because they have problems with agreement markers, as their use of this affix is normal. However, inflecting a verb for Tense is difficult for them. They produced fewer tense inflections in both languages compared to the non-brain damaged speakers. The data suggest that this is not because they have problems with tense in general. The use of verbs in the present tense is normal in both languages. However, if we take both tense and aspect into account, and make a division between verb forms that refer to the present and verb forms that refer to the past, the real underlying deficit appears, a deficit that manifests itself in both languages. It is reference to the past through verb inflection that is impaired: in both languages the agrammatic speakers produce significantly fewer verb forms that refer to the past; and in English this is compensated by an overproduction of verb forms referring to the present. This selective deficit for verb forms referring to the past has been reported before for monolingual agrammatic speakers in studies: spontaneous speech (Simonsen & Lind, 2002; Stravakaki & Kouvava (2003); conversation analysis (Beeke et al., 2003a); and in experimental study (see Bastiaanse et al., 2011, for an overview). These results are in line with the prediction made by the PADILIH (Bastiaanse et al. 2011). Reference to the past is discourse linked, whereas reference to the present is bound within the sentence. 2.4.4 Linguistic complexity and frequency It has often been suggested that agrammatic (or telegraphic) speech is due to decreased processing abilities. The idea is that what is more complex requires more resources, hence, more difficult to produce or comprehend for individuals 2.4. DISCUSSION 59 with agrammatic aphasia. Although this sounds reasonable, the crucial question is what is complex and what is simple. According to some, complexity is based on frequency: highly frequent constructions (such as active sentences in English) require less processing load than infrequent constructions (such as passives), and are therefore easier to process for agrammatic individuals. Recently, Bastiaanse, Bouma and Postma (2009) showed that frequency is not the key factor in agrammatism, at least not in production: Dutch agrammatic speakers have less problems when asked to complete (a less frequent) embedded clause than (a high frequent) matrix clause. This can be explained in terms of linguistic complexity, but not in terms of frequency, although both factors often co-incide. Bastiaanse et al. (2009), therefore, suggest that the complexity hierarchy should be defined in linguistic terms. Avrutin (2000; 2006) offers a model how this works. With his competition model, he shows why certain linguistic categories (such as pronouns and which-questions) are harder to process for agrammatic individuals than other categories (reflexives and who-questions, respectively). The first category members require discourse linking that is processed by discourse syntax; the second are bound within the sentence and need to be processed by narrow syntax. The critical point in his argumentation is that discourse syntax requires more processing resources than narrow syntax and, therefore, narrow syntax is used to process all these categories, resulting in problems when elements are discourse-linked, like verb forms that refer to the past. However, the present data of bilingual agrammatic speakers present problems to this sort of processing model. It is obvious that the Swahili verb inflection paradigm is more complex than the English paradigm, both intuitively and from a linguistic point of view. Therefore, one expects the agrammatic speakers to have more problems in Swahili, which, however, is not the case. Just like in the study of Abuom et al. (2011), more problems arise with the simple paradigm. It is hard to see how or why the Swahili verb inflection paradigm requires less processing load than the English one. The theory raised by Abuom et al. (2011), that verb inflection is more anchored in the Swahili language system, may be on the right track, but it does not give an explanation in terms of processing load. Another problem is that even if the verb inflections are more anchored in the Swahili system, the selective problems with reference to the past are similar as in English (though less severe). Goral (2011), however, suggests that it is not a matter of morphological complexity but rather 60 CHAPTER 2. CHARACTERISTICS OF SWAHILI-ENGLISH AGRAMMATISM of regularity. She argues that the morphology of a language with complex but regular morphology is likely to be better preserved than the morphology of a language with simple but not completely regular morphology. This argument has been substantiated by Bybee (2007) who associates regular morphology with affixation and irregular morphology with internal stem change. Bybee (2007) argues that affixation (regularity) is rather more ‘natural’ and highly frequent in most languages, hence easier to process in general; whereas internal stem change (irregularity) is rather less ‘natural’, lexically arbitrary and of low type frequency, hence more difficult to master. Centeno and Anderson (2011) demonstrate that individuals with linguistic limitations (such as agrammatic speakers, children with SLI), in their oral expression, have a preference for highfrequency linguistic items over low frequency items in a language. This theory partly explains high preference for present verb forms in both languages, as well as the slightly better preserved Tense system in Swahili than in English regardless of the morphological complexity of the two languages. Swahili past Tense has only one regular form marked as an infix on the verb paradigm, whereas English past Tense has both regular and irregular forms. Furthermore, Swahili verb morphology has high frequency of affixation compared to English, which combines a few instances of affixations and internal stem changes. In sum, the current data and the discussion thereof do not offer an answer to the question on the underlying deficit. It rather challenges the theories that pose that it is a matter of processing load. We do not dispute that agrammatism is a processing rather than a representational deficit. However, we do think that more data and more refined experiments are needed to generate a theory that can explain the problems of agrammatic speakers in terms of complexity in relation to the effects of diminished processing abilities due to brain damage. The present study demonstrates that studies in bilingual populations can be very helpful, if not in solving the problem, then at least in revealing the weak points. CHAPTER 3 Production and Comprehension of Reference of Time in Swahili-English bilingual agrammatic speakers Abstract Background: Several studies on time reference show that monolingual agrammatic speakers have difficulty producing and/or comprehending verb forms referring to past events or actions. The PAst Discourse LInking Hypothesis (PADILIH) has been formulated to account for this phenomenon (Bastiaanse, Bamyaci, Hsu, Lee and Yarbay Duman, 2011). In the current study on bilingual aphasia, we examine whether time reference problems, especially reference to the past, extend to both languages of bilinguals with agrammatic aphasia. The two languages, Swahili and English, have different verb morphology for expressing reference of time. Aims: The current study tested the production and comprehension of reference of time through verb morphology in two languages of Swahili-English bilingual agrammatic speakers. This study has been published as: Abuom, T. & Bastiaanse, R. (2013). Production and Comprehension of Reference of Time in Swahili-English bilingual agrammatic speakers. Aphasiology 27, 157-177. 61 62 CHAPTER 3. TIME REFERENCE PRODUCTION AND COMPREHENSION Methods & procedures: Thirteen agrammatic speakers and thirteen nonbrain-damaged individuals were tested using an adaptation of the Test for Assessing Reference of Time (TART; Bastiaanse, Jonkers, & Thompson, 2008; Swahili version: Abuom & Bastiaanse, 2010). Reference to the past, present and future conditions were examined through a sentence-completion and a picture-sentence matching task. Outcomes & Results: While the non-brain-damaged participants performed at ceiling in both languages, the agrammatic individuals’ performance showed a selective deficit of reference to the past on both comprehension and production tasks. A similar pattern was observed in the two languages in spite of the structural differences. Conclusions: The PAst Discourse LInking Hypothesis (PADILIH) was supported by these results. Furthermore, it has been revealed that time reference deficits extend to both tested languages of bilingual speakers with agrammatic aphasia regardless of the structure of languages mastered pre-morbidly. The implications of these findings for the theories of bilingual agrammatism are discussed. 3.1 Introduction In daily life communication, it is often easy and rather effortless not only to talk about, but also to understand, events or actions that: happened in the past, is happening at the present, or will happen in the future. The use of verb inflections to communicate time reference is generally not a problem for most people. However, there is evidence that this may not be the case with brain damaged individuals suffering aphasia. Recent cross-linguistic studies on time reference indicate that individuals with agrammatic aphasia have problems with verb inflections for reference to the past time-frame (e.g. Stavrakaki & Kouvava, 2003 and Nanousi, Masterson, Druks and Atkinson (2006) for Greek; Bastiaanse, 2008; Jonkers & de Bruin, 2009 for Dutch; Yarbay Duman & Bastiaanse, 2009 for Turkish; Abuom, Obler, Bastiaanse (2011), Bastiaanse, Bamyaci, Hsu, Lee and Yarbay Duman (2011) for English; Abuom & Bastiaanse (2012) for Swahili and English). It is, however, not clear how these problems with time reference are reflected in both languages spoken by a bilingual agrammatic individual. This is particularly interesting when the languages have very different verb inflection paradigms, as in this case, Swahili and English. The verb inflection 3.1. INTRODUCTION 63 paradigm of Swahili is very large, whereas it is relatively small in English. The current study investigates the performance of pre-morbidly highly proficient Swahili-English bilingual agrammatic speakers on production and comprehension of time reference in both languages. The question is how time reference problems, especially reference to the past, are manifest in both production and comprehension of the Swahili-English bilingual agrammatic speakers. The study will determine whether the size of the verb inflection paradigm influences the performance of the agrammatic speakers quantitatively and qualitatively, and whether the language with a larger paradigm (Swahili) is more or less impaired than the one with a smaller paradigm (English). In the next sections we will first present a description of time reference morphology in Swahili; followed by an overview of the literature on time reference problem among monolinguals and bilinguals with agrammatic aphasia. The predictions for the current study will conclude the introduction section. 3.1.1 Time reference morphology in Swahili Swahili, a Bantu language spoken mostly in Africa, including Kenya, has Subject - Verb - Object as the basic word order (Ashton, 1982). Kenya is a multilingual society with an average person speaking at least three languages. Swahili and English are the two most dominant languages across the country, used both as official and as national languages, but most Kenyans speak an extra language from the 42 languages linguists refer to as “ethnic languages” (classified under four broad categories: Bantu; Nilotic; Indo-Aryan; and Cushitic languages) at home as well. Swahili and English have the same status as second languages since both are acquired around the age of 4, after acquiring the native language. Swahili’s most distinguishing characteristic is its agglutinative aspect; and of greatest interest to this study is the size of the verb structure consisting of prefixes and suffixes. Time reference for Tense and Aspect in Swahili is done through verb inflection as a prefix in the verb complex. The complete structure of the Swahili verb paradigm is illustrated in (1). It is important to note that the subject agreement (Sp), tense (T), root and the derivation (d) are obligatory in every affirmative Swahili utterance. The Tense system used for time reference in Swahili is given in examples (1a-e). 64 CHAPTER 3. TIME REFERENCE PRODUCTION AND COMPREHENSION (1) Pre-prefix (Pp)7 + Subject prefix (Sp) + Tense prefix (T) + Object prefix (Op) + ROOT + derivation (d) + Suffix (S)+ Post-suffix (Ps)8 (a) LEO [Wa + na + tak + a] ku + m + pig +a TODAY [Sp + T + ROOT + d] Pp + Op + ROOT + d ‘TODAY [they want] to hit him/her’. (b) JANA A + li + m + pig +a YESTERDAY Sp + T + Op + ROOT + d ‘YESTERDAY she/her hit him/her’. (c) LEO U + me + m + pig +a TODAY Sp + T + Op + ROOT + d ‘TODAY you have hit him/her’. (d) SASA Tu + na + pig +a +n +a NOW Sp + T + ROOT + d + S + Ps ‘NOW we are hitting each other’. (e) KESHO Ha + tu + ta TOMORROW Pp + Sp + T + m + pig +a + Op + ROOT + d ‘TOMORROW we will not hit him/her’. As illustrated in (1a-e), Swahili has three explicit morphological markers of Tense: past, present and future. The infinitive (1a) is also marked in Swahili by the prefix KU- attached either directly to the verb root as in ‘to hit’ (kupiga) or attached to the object prefix with the verb root as in ‘to hit him/her’ (kumpiga). This prefix KU Corresponds to the English infinitive generally signaled by the use of ‘to’ before the verb. The past Tense marker formed by the insertion of the prefix -LI, as in (1b), refers to a past activity without reference to a specific time. The prefix -LI corresponds to the simple past in English (Verb + ed: he painted). The present perfect Tense inserted as a prefix -ME- (1c) refers to a past activity which has relevance to the present time. The prefix -ME- is comparable to the present perfect in English (has/have + Verb+ ed : he has painted). The present Tense -NA- describes actions taking place at the present moment (1d) comparable to either the English present progressive 7 The pre-prefix is a negation marker found in all negative sentences. It has, however, not been included in the present test. 8 It is not possible for a verb to possess all these affixes at one and the same time. 3.1. INTRODUCTION 65 (Verb + ing: he is hitting him) or simple present (Verb + s: he hits him). The future Tense marker -TA- describes events assumed to follow the present moment (1e), corresponding to reference to the future through periphrastic form in English (will + infinitive: he will paint). It is generally considered ungrammatical to omit Tense markers in any Swahili utterance whether or not the temporal reference is clear from the discourse or any other sources (with the exception of imperative sentences such as ‘Nenda nyumbani’, a translation equivalent of ‘go home’ in English). The grammatical structure of time reference morphology for the two languages is clearly different: Swahili has a large verb inflection paradigm that can function as a complete sentence on its own; whereas, English has a rather small verb paradigm that requires other sentence constituents. In sum, verb morphology in Swahili appears complex due to its relatively large number of affixes on the verb paradigm, but it is completely regular. Also, time reference is simple: there are only three tenses expressed through affixes which are inserted in the same position for every single verb. No irregular verbs exist. In English, the finite verb itself is simpler than in Swahili, but there is a wider variation in inflection for time reference, since it is inflected for both tense and aspect. The inflection is done by both analytical verb forms (‘wrote’) and periphrastic verb forms (‘has written’). Additionally, English has regular and irregular verbs. This means that overall the Swahili finite verbs are more complex, because Swahili is an agglutinative language where the finite verbs incorporate more information, although inflection for time reference is simpler than in English. 3.1.2 Time reference problem in monolinguals with agrammatic aphasia Most languages express time reference through tense and aspect. A number of studies that have compared the processing of Tense and Agreement in individuals with agrammatic aphasia show that Tense inflections are more prone to impairment than Agreement inflections. Friedmann and Grodzinsky (1997) found impaired tense inflection, but intact agreement inflection in Hebrew aphasic speaker. They, as a result, proposed the so called Tree Pruning Hypothesis (TPH), which attributes the difficulty with tense to its higher position in the syntactic tree. Wenzlaff and Clahsen (2004) found similar results for German aphasic speakers. However, they proposed the Tense Under-specification Hy- 66 CHAPTER 3. TIME REFERENCE PRODUCTION AND COMPREHENSION pothesis (TUH), which relate the tense problem to the fact that tense carries extra-sentential information. Burchert, Swaboda-Moll and De Bleser (2005), however, found both tense and agreement can be difficult for German agrammatic speakers. They proposed the Tense and Agreement Hypothesis (TAUH), arguing that both tense and agreement can be affected independently in agrammatic speakers. However, what these studies fail to show is: whether Aspect is impaired as well; and whether past Tense inflections are impaired selectively or not. The findings by Stavrakaki and Kouvava (2003) and Nanousi, Masterson, Druks and Atkinson (2006) for Greek; Bastiaanse (2008); Jonkers and de Bruin, (2009), for Dutch; Yarbay Duman and Bastiaanse (2009), for Turkish; and Bastiaanse et al. (2011) for English, Chinese and Turkish, show a selective impairment of reference to the past through Tense and/or Aspect. Stavrakaki and Kouvava (2003) analysed the production of Tense, Aspect and Agreement in the spontaneous speech of two Greek agrammatic speakers and found that past Tense was more difficult to produce than the present Tense. Furthermore, the past Tense with the perfective aspect (‘’I wrote’) was more difficult to produce than the past Tense with the imperfective aspect (‘I was writing’). However, the production of subject-verb Agreement was relatively spared. Nanousi et al. (2006) also reported impaired perfective aspect in a group of Greek agrammatic speakers. Bastiaanse (2008) used a sentence completion task to assess the performance of a group of Dutch agrammatic speakers on production of time reference through Tense (past Tense versus present Tense) and periphrastic forms (non-finite participle versus infinitive). The results were in line with the findings of Stavrakaki & Kouvava (2003): past Tense forms were more difficult to produce than the present Tense forms. Interestingly, however, for the non-finite forms, the production of the participle (used, in combination with a present tense auxiliary, to denote perfect aspect; and not marked for gender or number) was more impaired than the production of the (uninflected: with no tense or agreement) infinitive. Yarbay Duman and Bastiaanse (2009) examined the performance of a group of Turkish agrammatic speakers on the production of Tensed finite verbs and participles referring to the past and the future through a pictured sentence completion task. Their findings revealed past Tense/perfect aspect to be more difficult than future Tense/imperfect aspect for the Turkish agrammatic speakers. In another study on time reference in Dutch agrammatic speakers, Jonkers 3.1. INTRODUCTION 67 & de Bruin (2009) tested both production and comprehension of the present and the past Tenses in seven individuals with Broca’s aphasia and five individuals with Wernicke’s aphasia using a sentence-to picture matching task for comprehension and a sentence completion task for production. Jonkers & de Bruin (2009) observed that past Tense forms were more impaired than present Tense forms in both production and comprehension. They noted, however, that past Tense deficit was a general problem that affects Broca’s and Wernicke’s aphasic individuals alike. The results of these studies raised the question why reference to the past is more difficult for agrammatic individuals. Bastiaanse et al. (2011) formulated the PAst DIscourse LInking Hypothesis (PADILIH) which is based on the idea of Zagona (2003) that when referring to a present event through Tense, speech time and event time are the same and hence, are locally bound. When referring to the past through Tense an extra-sentential reference is made, because speech time and the (earlier) event time are different. Therefore, discourse linking is required. For reference to the future, Bastiaanse et al. (2011) suggest that the relationship is neither discourse-linked nor locally bound since the event has not yet happened (see also Zagona, in press). Avrutin (2000; 2006), based on findings from a number studies, suggests that the interpretation of discourse-linked relationships is a problem for individuals with Broca’s aphasia. He distinguishes ‘narrow syntax’, that is, syntax processes that are taking place within the sentence structure, and ‘discourse syntax’, that is, processing syntactic information at the discourse level. According to Avrutin (2000; 2006) more processing resources are required to establish relationships with discourse syntax information than with narrow syntax. Bastiaanse et al. (2011), in proposing the PADILIH, expanded Zagona’s (2003) theory on tense and discourse linking to include not only past Tense, but all verb forms referring to the past, including periphrastic verb forms and verb forms in perfect aspect. The PADILIH, therefore, predicts that all verb forms referring to a past or finished event, regardless of Tense, are discourse linked and, therefore, require more processing resources, leading to more errors by grammatically impaired individuals. In a cross-linguistic study, Bastiaanse et al. (2011) tested the PADILIH. Agrammatic speakers of Chinese, English, and Turkish were assessed on both comprehension and production of time reference morphology (in three conditions: past, present and future) using the Test for Assessing Reference of Time 68 CHAPTER 3. TIME REFERENCE PRODUCTION AND COMPREHENSION (TART; Bastiaanse, Jonkers, & Thompson, 2008). Their results were in line with previous findings on time reference: both Turkish and English agrammatic speakers showed a selective deficit with reference to the past through verb inflections in production and comprehension. The Chinese agrammatic speakers’ performance on comprehension was relatively similar to those of Turkish and English. However, the Chinese agrammatic speakers’ performed poorly on production of all the three time-frames: reference to the past, present and the future. The authors attribute the Chinese production results to a general difficulty with aspectual adverbs: time reference is expressed through aspectual adverbs rather than verb inflections. These findings were taken as support for the PADILIH, attributing the problem with reference to the past, through tense and / or aspect, to impaired discourse linking in agrammatic aphasia. What all these studies on Tense and time reference have in common is that they have mainly investigated monolingual agrammatic speakers. In the present study with bilingual agrammatic speakers, the performance in both Swahili and English can be compared to evaluate the influence of the verb inflection paradigm on the severity of the time reference deficit. 3.1.3 Time reference problems in bilinguals with agrammatic aphasia In the current study, we use the term ‘bilingual’ to include both bilingual and multilingual persons who make use of two or more languages or dialects in their everyday lives (Grosjean, 1994). Most bilingual aphasic individuals suffer from the same type of aphasia in all their languages mastered pre-morbidly; and the recovery pattern commonly observed is parallel (Fabbro, 1999, 2001; Paradis, 2001; Miozzo, Costa, Hernandez & Rapp, 2010; Abutalebi, Cappa, & Perani, 2005). Non-parallel recovery patterns have also been reported and are usually attributed to differences in either the age of acquisition, frequency of use or proficiency level between the languages mastered pre-morbidly (Albert & Obler, 1978; Fabbro, 1999). In early balanced and proficient bilinguals, languages acquired early are generally assumed to be represented in shared processing regions (e.g., Miozzo et al., 2010; Abutalebi et al., 2005). When it comes to time reference problems among bilingual aphasic speakers, so far, only two studies have investigated the tense and time reference problem in bilingual agrammatic speakers. In the first study, Abuom et al. (2011) tested two Swahili-English bilingual agrammatic speakers on production of reference 3.1. INTRODUCTION 69 to the past and the future, using a sentence completion task. The English data confirmed the findings of the previous studies on Tense and reference to the past. However, reference to both the future and the past was intact for Swahili in both agrammatic speakers. For the second study, Abuom and Bastiaanse (2012) analysed the use of verb inflections for tense and time reference to the past and to the present in the spontaneous speech of six Swahili-English agrammatic speakers. The results showed that the use of verb inflections for reference to the present was normal in their spontaneous speech, but the number of verb forms referring to the past was significantly lower than normal. However, overall, the use of verb inflections was better preserved in Swahili than in English. While the results of both studies on bilingual Swahili-English agrammatic speakers concur with respect to the difficulty with producing reference to the past in English, it is not yet clear whether Swahili is similarly affected. Furthermore, the comprehension of time reference in Swahili-English agrammatic speakers has not been investigated. Therefore, it is unclear whether the predictions of the PADILIH apply similarly to the two languages of bilingual agrammatic speakers. 3.1.4 The current study The current study examined the production and comprehension of time reference to the past, the present and the future in the two languages, using the Test for Assessing Reference of Time (TART: Bastiaanse et al., 2008; bilingual English-Swahili version: Abuom & Bastiaanse, 2010). We assume that balanced bilinguals who acquired their languages from an early age use a shared grammatical device (Miozzo et al., 2010; Abutalebi et al., 2005). The key research questions this study addressed were as follows: (1) Is production of time reference to the past through verb inflection impaired in both languages of a bilingual? (2) Is comprehension of time reference to the past through verb inflection impaired in both languages of a bilingual? (3) Is reference to the past similarly impaired in both Swahili and English? According to PADILIH: (1) the production of time reference to the past through verb inflection is impaired, whereas, reference to the present and fu- 70 CHAPTER 3. TIME REFERENCE PRODUCTION AND COMPREHENSION ture are relatively spared; (2) the comprehension of time reference to the past through verb inflection is impaired, whereas, reference to the present and the future is relatively intact; (3) the impairment is language independent, hence, both Swahili and English will be impaired similarly. However, according to Bates, Friederici and Wulfeck (1987), in a language with a more complex inflectional paradigm, the cue cost is higher to produce the grammatical morphemes. This predicts that the bilingual agrammatic speakers will make more errors in Swahili, where the verb inflection paradigm is larger (more complex). However, Bates, Wulfeck and MacWhinney (1991) introduce the notion of cue validity: inflectional markers that are more informative are better preserved than inflectional markers that are less informative. With respect to the present study, this theory predicts that past, present and future morphology will be equally affected, since all three are used for reference to a time frame. Also, cue validity is the same in English and Swahili. In sum: the cue cost in Swahili is higher, so more errors are expected in Swahili; the cue validity in the two languages is the same, so no differences are expected. Another interesting idea was expressed by Goral (2011). She assumes that in bilingual agrammatic aphasia, regular paradigms are better preserved than paradigms with irregular members. Goral (2011) thus expects Swahili with its fully regular, though larger (complex) paradigm to be better preserved than English with its smaller (simple) paradigm for reference to the past, that nevertheless contains three different regular (-t, -d, -ed) allophones and irregular forms. 3.2 3.2.1 Method Participants There were a total of 26 participants in this study: 13 Swahili-English bilingual agrammatic speakers from Aga Khan University hospital Nairobi-Kenya and 13 non-brain-damaged speakers (NBDs) matched in age, gender, native language and education level to agrammatic speakers. All participants were aged between 20 and 49 years, with over 12 years of education, and none had any history of neurological, hearing or vision problems. The agrammatic speakers had all suffered from single stroke with the exception of one, EA, who had left hemisphere brain damage due to a car accident. They were early balanced bilinguals, equally proficient in English and Swahili 3.2. METHOD 71 pre-morbidly; any Kenyan adult with over 12 years of uninterrupted education is generally expected to be equally highly proficient in both Swahili and English. The demographic details of the agrammatic participants are shown in the Table 3.1. Table 3.1: Demographic details and the results on the test for auditory comprehension of words (on an adapted version of the BDAE-test in Swahili and English) of the agrammatic speakers Agrammatic participants SW HJ LA MM JK EA PN JA MW VK HS JN SS Mean Age (years) Gender Handedness Education (years) Years post stroke/ head trauma Native language Swahili BDAEsubtest English BDAEsubtest 20 45 43 47 49 42 36 46 50 25 64 50 30 42.1 M F F F M M F M F F M F F R R R R R R R R R R R R R 12 14 16 16 17 16 14 16 16 16 16 13 12 14.9 2 10 1 10 1 17 1 1 1.5 2 1 10 17 4.4 Bantu Nilotic Bantu Nilotic Bantu Nilotic Bantu Bantu Bantu Bantu Indo-Aryan Bantu Indo-Aryan 98.6% 100% 100% 100% 100% 100% 100% 100% 98.6% 100% 99% 100% 100% 99.7% 99% 100% 100% 100% 100% 99% 100% 100% 100% 98.6% 98.6% 100% 99% 99.5% Their spontaneous speech was judged to be telegraphic with reduced speech rate in both languages by a practicing speech therapist, and further confirmed by the experimenter based on the criteria of Menn and Obler (1990)9 . However, due to lack of most relevant standardized tests for establishing the aphasia syndrome in Kenya, we administered an adapted version of a sub-test from the Boston Diagnostic Aphasia Examination (Goodglass & Kaplan, 1972) to ensure their good comprehension abilities. They all had good comprehension in both languages on the subtask for auditory comprehension of single words (nouns, verbs, colors, shapes, letters, and numbers) from the Boston Diagnostic Aphasia Examination. Some pictures were substituted due to cultural differences, for instance, a hammock was changed to a swing since no Swahili word exists. Unfortunately, there is no information as to the possible extent of cognitive damage in agrammatic participants. The agrammatic speakers’ scores on the 9 There was spontaneous speech available of 6 of the patients, who also participated in Abuom & Bastiaanse, 2012. Their speech was characterized by significantly (1) lower speech rate; (2) shorter sentences; (3) fewer grammatical sentences; (4) fewer embeddings in both Swahili and English. 72 CHAPTER 3. TIME REFERENCE PRODUCTION AND COMPREHENSION BDAE subtest for auditory word comprehension are included in Table 3.1. 3.2.2 Materials and procedure The African version (an adaptation for English and Swahili) of the Test for Assessing Reference of Time (TART: Abuom & Bastiaanse, 2010) was used. The TART was developed for cross-linguistic investigation of comprehension and production of reference of time and it has been used in several other languages including: Dutch, English, Indonesian, Chinese, and Turkish. The test has two sections with pictures of 20 verbs depicting actions in three time-frames: the present (an on-going action), the past (a completed action) and the future (an action about to start). The production section consists of a sentence-completion paradigm with prompting to elicit the intended verb forms. There are 20 items for each of the three conditions: the simple past, the present continuous, and the future. The following forms were elicited in the two languages: simple past (an infix -li- for Swahili, and a suffix -ed for English); present continuous (an infix -nafor Swahili, and is + V -ing for English); future (an infix -ta- for Swahili, and will + infinitive for English). The participants were presented with a pair of pictures of contrasting actions in a particular time-frame (past, present and future; see Figure 3.1 and the examples in 2a-c below) with the same object. The infinitive form of the verb was printed above the pictures on the same page (e.g. ‘write’ and ‘read’). The experimenter constructed a sentence for the picture on the left using the appropriate temporal adverb with the correct form of the verb. The participant was expected to complete the second sentence for the other picture using the right form of verb after the experimenter has given the temporal adverb and the subject (See example 2a-c for eliciting: present continuous, past and future tenses in Swahili and English). Samples of pictures for the production task are given in Figure 3.1a-c. 3.2. METHOD 73 (a) WRITE READ (b) POUR DRINK (c) PEEL EAT Figure 3.1: Samples of the pictures used for production task: a) to elicit time reference to the present; b) to elicit time reference to the past; c) to elicit time reference to the future. 74 CHAPTER 3. TIME REFERENCE PRODUCTION AND COMPREHENSION (2) (a) Examiner: Sasa huyu Now this mwanamme man mwanamme a-na-andika barua. man he-Present-write a letter. ... ... Sasa huyu Now this ‘Now this man is writing a letter. Now this man...’ Participant: a-na-soma barua he-Present-read a letter ‘is reading a letter’. (b) Examiner: Hapo-mbeleni Previously Hapo-mbeleni Previously huyu this huyu this mwanamme a-li-mwaga maziwa. man he-Past-pour milk. mwanamme ... man... ‘Previously this man poured milk. Previously this man...’ Participant: a-li-kunywa maziwa he-Past-drink milk ‘drunk milk’. (c) Examiner: Hivi-punde Soon Hivi-punde Soon huyu this huyu this mwanamme man mwanamme man a-ta-menya chungwa. he-Future-peel orange. ... ... ‘Soon this man will peel an orange. Soon this man ...’ Participant: a-ta-kula chungwa he-Future-eat orange ‘will eat an orange’. The comprehension section consists of a spoken-sentence-to-picture-matching 3.2. METHOD 75 Figure 3.2: A sample of the pictures used for comprehension task task. The pictures of 20 verbs depicting actions in three time frames from production task were also used in comprehension task. Constructions in three time-frames (the simple past, the present continuous, and the future) were used with each time frame consisting of 20 items. A pair of pictures was presented to the participants and a sentence with the target inflected form of the verb was read aloud by the experimenter. The participants chose a picture from the given pair that corresponds to the sentence read by the experimenter (see example 3). The pictures depict actions in two contrasting time-frames (see Figure 3.2). (3) Experimenter: Mwanamme a-na-kula chungwa Man he-Present-eat an orange ‘The man is eating an orange’ The TART has been translated into more than 15 languages, and the same verbs are used for each language. Therefore, there were verbs with regular (n=13) as well as irregular (n=7) past tense included in the English version. The tests in each of the languages were administered on two different days with an interval of two weeks for each of the agrammatic speakers. The order of the tests was varied for each participant: either English first, followed by Swahili or vice versa, but all participants were tested on production first before 76 CHAPTER 3. TIME REFERENCE PRODUCTION AND COMPREHENSION comprehension in each language. The tests were conducted in a speech therapy room at the Aga Khan University Hospital in Nairobi, Kenya. Each test session began with six practice items to ensure the participants understood the task before starting the test. The participants were corrected and given feedback during the practice. No further feedback was given once the test was begun. Each test session of the agrammatic speakers lasted one and half hours, with a break. 3.2.3 Analysis The 13 non-brain-damaged Swahili-English bilingual speakers, matched in age, gender, native language and education level to the agrammatic speakers performed at ceiling on the two tests for both languages. We therefore assume that all errors made by the agrammatic speakers are a result of their aphasia. The agrammatic speakers’ data were analysed both quantitatively and qualitatively. For quantitative analysis, the number of both the correct and incorrect responses produced, and the number of correctly identified pictures, by each participant were tallied. Self-corrections were permitted during the test and the final response was scored. Produced responses were considered correct when the correct form of verb inflection was produced for the three time-frames in Swahili (-li for simple past, -na for present continuous, and -ta for future) and English (-ed for simple past, -ing for present continuous, and will + infinitive for future). Omission of the object was ignored. For the qualitative analysis, the three most frequent error types were distinguished in the production task: substitution (e.g. ‘anakula’‘he is eating’ instead of ‘alikula’‘he ate’), omission (e.g. ‘*sukuma toroli’‘pull the trolley’ instead of ‘alisukuma toroli’‘he pulled the trolley’), and other errors (e.g. ‘*he sweeped the floor’ instead of ‘he swept the floor’). For statistical analysis, we used repeated measures ANOVA (for main effect of language, sub-test and time reference); t- tests (for differences between the time-frames within each language, and between the two languages) and correlations (for relationship in performance within and between the two languages) 3.3. RESULTS 3.3 3.3.1 77 Results Overall analysis: A repeated measures analysis of variance was performed to investigate main effects of language (English and Swahili), sub-test (production and comprehension) and time reference (past, present and future) on agrammatic performance. There was a statistically significant effect for language: F (1, 12) = 19.340, p = .001; Swahili was better preserved than English. There was also statistically significant effect for Sub-test: F (1, 12) = 7.256, p = .020; production was worse than comprehension in both languages, which is to be expected, since there is a 50% chance level for the comprehension test. Furthermore, there was a statistically significant main effect for time reference: F (2, 12) = 48.230, p = .000; reference to the past was worse than reference to both the present and the future in the two languages. 3.3.2 Production in English and Swahili The production results are presented in Table 3.2. Table 3.2: The comparison of agrammatic speakers’ scores on production of reference of time in Swahili and English. SW HJ LA MM JK EA PN JA MW VK HS JN SS mean Swahili Production Past Present Future English Production Past Present Future 12 19 20 20 20 11 17 19 14 14 17 15 16 16.46 9 4 17 15 18 5 5 11 7 9 9 6 11 9.69 20 18 19 20 19 15 20 20 20 19 20 20 20 19.23 19 20 20 20 20 14 20 20 20 20 20 20 19 19.38 20 20 20 20 20 18 20 20 20 20 20 20 20 19.85 20 20 20 20 20 20 20 20 20 20 19 19 20 19.85 78 CHAPTER 3. TIME REFERENCE PRODUCTION AND COMPREHENSION On the English version of the test, the results showed a significant main effect for production of reference of time: F (2, 12) = 65.730, p = .000. Reference to the past was significantly impaired compared with reference to the present (t (12) = 8.23, p = 0.0001), and to the future (t (12) = 8.09, p = .0001). There was no difference between regular and irregular past tense forms (t (12) = 0.71, p = .49) (see Appendix 3.1). There was also no significant difference between reference to the present and to the future (t (12) = 0.00, p = 1.000). The results on the Swahili version of the test showed a similar pattern: there was again a significant effect for reference of time: F (2, 12) = 13.551, p = .002. The agrammatic speakers performed poorly on reference to the past compared with reference to the present (t (12) = 3.39, p = .0054) and to the future (t (12) = 4.28, p = .0011). However, no significant difference between reference to the present and the future was found (t (12) = 0.62, p = .5486). Overall, the production of English past (mean = 9.69) was significantly worse than the production of Swahili past (mean = 16.46) (t (12) = 6.60, p = .0001). 3.3.3 Comprehension in English and Swahili The comprehension results are presented in Table 3.3. The pattern of performance on comprehension of reference of time was quite similar to that on production. For English comprehension, a significant main effect was found: F (2, 12) = 16.434, p = .001. Comprehension of English reference to the past was significantly more challenging than reference to both the present (t (12) = 4.10, p = .0015) and the future (t (12) = 4.04, p = .0016). There was no effect of regularity of English past tense (t (12) = 0.55, p = .59) (see Appendix 3.1), meaning that the poor performance on English past tense was not due to the irregularity of some of the verbs. Again, no significant difference between reference to the present and to the future was found (t (12) = 1.00, p = .3370). The performance on Swahili showed a relatively similar pattern, there was a significant main effect for comprehension of reference of time: F (2, 12) = 19.648, p = .000). Comprehension of Swahili reference to the past was more challenging than Swahili reference to the present (t (12) = 4.57, p = .0006) and to the future (t (12) = 4.63, p = 0.0006). The difference in performance between Swahili reference to the present and to the future was not significant (t (12) = 1.48, p = .1654). The performance on comprehension of English reference to the past (mean = 14.92) was significantly worse than comprehension of the Swahili reference to the past (mean = 17.46) (t (12) = 3.3. RESULTS 79 2.75, p = .0177). Table 3.3: The comparison of agrammatic speakers’ scores on comprehension of reference of time in Swahili and English. SW HJ LA MM JK EA PN JA MW VK HS JN SS mean 3.3.4 Swahili Comprehension Past Present Future English Comprehension Past Present Future 19 17 17 20 20 19 15 17 15 18 19 15 16 17.46 20 7 17 20 20 18 9 16 15 11 15 12 14 14.92 20 20 20 20 20 19 20 20 20 20 20 20 20 19.92 20 18 19 20 20 20 19 20 20 20 20 19 20 19.61 20 19 20 20 19 19 20 20 20 20 20 20 20 19.77 20 19 19 20 20 19 20 20 19 20 20 20 19 19.62 Individual results on production and comprehension of English and Swahili past On further analysis of individual results of each agrammatic speaker, a discrepancy on individual performance is observed on production as well comprehension of time reference on both languages. The English production data show that all the agrammatic speakers had problems with reference to past. However, on the Swahili production data, three agrammatic speakers (LA, MM and JK) did not make any errors with past. This shows that with respect to production, reference to the past was more vulnerable in English than Swahili (See Table 2 above). For comprehension, the English comprehension data show that three agrammatic speakers (SW, MM, and JK) performed quite well on reference to the past, whereas, on Swahili comprehension, two of them (MM and JK) performed similarly quite well on reference to the Swahili past. It is clear that the two patients (MM and JK) had no difficulty comprehending past in both Swahili 80 CHAPTER 3. TIME REFERENCE PRODUCTION AND COMPREHENSION and English (See Table 3). In conclusion, all agrammatic speakers had difficulty with production of the English past, implying that reference to the past is a problem to all the agrammatic speakers, but not all agrammatic speakers are equally affected on both languages. However, further analyses all the data show a significant correlation in agrammatic performance between English and Swahili in production (R (37) = 0.68, p < .001 and in comprehension (R (37) = 0.80, p < .001; implying that agrammatic speakers who are poor in English are also poor in Swahili and the other way around. Again, at the subtest level, there was a significant correlation between production and comprehension10 in English (R (37) = .81, p < .001, and in Swahili (R (37) = .35, p < .029), indicating that agrammatic speakers who performed poorly in production also performed poorly in comprehension (and vice versa). 3.3.5 Error types The distribution of error types of the agrammatic speakers is presented in Table 3.4. Incorrect responses of the agrammatic speakers on the production task were distinguished into three error types: substitutions by other forms of the same verb, omissions of verb inflections and others. Substitution errors were equally prevalent in both Swahili and English tasks (t (12) = 0.73, p = .48). On the Swahili task, the past tense marker ‘li’ was frequently substituted with the present tense marker ‘na’. On the English task, the past time frame was substituted most frequently with the present time frame (either simple present or the present progressive). However, errors of omission were more prevalent on the English test than on the Swahili test (t (12) = 4.9, p = .0004). The agrammatic speakers quite often omitted the tense inflection in English (eg. ‘-ed’ for past). In a few instances in Swahili, they omitted both the subjectagreement prefix and the tense prefix. Other type of errors noted were a few regularization errors in English: some of the agrammatic speakers regularized the English past in irregular verbs (e.g. ‘sweeped’). 10 The comprehension task is a binary choice test, which makes a comparison between comprehension and production based on raw scores less valid. Therefore, the scores on the comprehension task were corrected for guessing, using the formula: Corrected score = #correct [# incorrect: (#alternatives-1)]. 3.4. DISCUSSION 81 Table 3.4: Distribution of error types on production of individual agrammatic speakers Swahili production errors Substitutions Omissions Others SW HJ LA MM JK EA PN JA MW VK HS JN SS mean 3.4 9 1 0 0 0 15 3 1 6 6 3 5 5 4.15 0 2 1 0 1 5 0 0 0 1 0 0 0 0.78 0 0 0 0 0 0 0 0 0 0 0 0 0 0 English production errors Substitutions Omissions Others 6 10 3 0 2 0 9 5 11 3 6 11 4 5.38 5 5 0 5 0 16 6 4 2 8 5 5 5 5.08 0 1 0 0 0 1 0 0 0 0 0 0 0 0.15 Discussion The present study examined the production and comprehension abilities of bilingual agrammatic speakers in regard to time reference. The results revealed that the agrammatic speakers were impaired in their production and comprehension of time reference to the past. However, their production and comprehension of time reference to both the present and the future are relatively spared. A similar pattern of impairment is reflected in both languages, Swahili and English, irrespective of their morphological differences. These findings are discussed in relation to the theories of agrammatism mentioned in the introduction. 3.4.1 Why time reference to the past is difficult for agrammatic speakers Friedmann and Grodzinsky’s (1997) TPH, argues that the agrammatic representation of the syntactic tree is pruned at the position of the Tense node. As 82 CHAPTER 3. TIME REFERENCE PRODUCTION AND COMPREHENSION a result, the tense node is inaccessible to agrammatic speakers. The authors, therefore, predict that Tense should be impaired for both languages irrespective of time-frame. We do not find this account compatible with the present data because it cannot explain the selective impairment we found for reference to the past in both languages. An inaccessible Tense node should lead to random distribution of errors over the tested categories (past, present and the future in English and Swahili); however, this was not the case in the present study. The TUH (Wenzlaff and Clahsen, 2004) and TAUH (Burchert et al., 2005) concur that it is not the position of tense in the syntactic tree that is causing problems, but the characteristics of Tense. The authors argue that the interpretable features of Tense that are used for time reference are underspecified in agrammatic speakers. Tense inflection must be checked against the time-frame the proposition is referring to. This requires an extra-sentential computation which is usually difficult for agrammatic speakers. However, this explanation cannot adequately account the specific problem with reference to the past evidenced in this study and in several other studies (Bastiaanse, 2008; Bastiaanse et al., 2011; Jonkers & De Bruin, 2009; Nanousi et al., 2006; Stavrakaki & Kouvava, 2004; Yarbay Duman & Bastiaanse, 2009). Generally speaking, all these theories (TPH, TUH and TAUH) restrict themselves to Tense; they do not explain the problems agrammatic speakers have with verb forms expressing time reference through Aspect inflection, nor through periphrastic forms. Reference to the future is relatively spared, both in production and in comprehension. For English, that may be explained by its simple form: an invariant auxiliary that is not inflected for person and number and an infinitive. However, this does not hold for Swahili, where reference to the past, present and future are marked by inflections within the verb. In other languages, performance on present and future are also relatively spared compared to reference to the past, for example, for Turkish, which is an agglutinative language, like Swahili. According to Zagona (in press), future is a variant of present. Together, they are ‘non-past’, and therefore, reference to the future is not discourse linked. The PADILIH of Bastiaanse et al. (2011) apparently accounts best for the current data. The PADILIH is based on the idea that reference to the past through grammatical morphology (Tense and Aspect) is discourse linked, reference to the past, therefore, requires extra processing load (Avrutin, 2000; 2006), These extra processing resources are not sufficiently available in the affected brain areas of the agrammatic speakers. Notice that ‘discourse link- 3.4. DISCUSSION 83 ing’ or ‘processing by discourse syntax’ does not mean that there should be a linguistic context. It means that a linguistic element (in this case the verb inflection for time reference) should be linked to an extra-sentential context, which may or may not be text. Thus, it is processing at the syntax-semantic interface. In the sentence completion test, there was linguistic discourse: there was an introductory sentence and a lexical adverb referring to past, present or future. In the comprehension test, the processing at the level of discourse context had to be done solely on the basis of the past tense inflectional marker. When hearing a past tense inflection, the hearer must process at the level of discourse syntax, since the time of speaking and the time of the event do not coincide. 3.4.2 Why reference to the past is worse in English than in Swahili One may propose the influence of the native language of the agrammatic speakers as an alternative explanation for the slightly better performance in Swahili past. However, we do not think this is the case. As mentioned, Swahili is a Bantu language, whereas English is not. Apart from the acquisition of both Swahili and English at an early age, the agrammatic speakers acquired another language from birth, which is classified either as Bantu, Nilotic or IndoAryan. Bantu languages, like Swahili, express time reference through a very rich and distinct Tense morphology. Similarly, Indo-Aryan expresses time reference through Tense. Nilotic languages, on the contrary, express time reference through Aspect. Furthermore, in Nilotic languages, both perfective and imperfective Aspect are distinguished based on tone rather than on grammatical morphemes (Dimmendaal, 2001). Of the thirteen agrammatic participants, eight acquired Bantu, while the rest acquired Nilotic and Indo-Aryan languages, respectively, as first languages. Although these first languages have different ways of expressing time reference, we still find a similar pattern of impaired reference to the past in Swahili in all the agrammatic speakers. Furthermore, we did not find any differences on performance on reference to the past between native speakers of Bantu and native speakers of other languages (Nilotic and Indo-Aryan). Therefore, we do not think that the structure of the earliest learned language explains the better performance in Swahili. The better performance on Swahili (compared to English) is shown consistently in our studies (Abuom et al, 2011; Abuom & Bastiaanse (2012). 84 CHAPTER 3. TIME REFERENCE PRODUCTION AND COMPREHENSION According to Bates et al. (1987)’ cue cost theory, morphologically complex paradigms are likely to yield more errors than morphologically simple paradigms. Since the Swahili verb morphology paradigm is large compared to the English paradigm, Bates et al. (1987) predict worse performance in Swahili than English. What we found is the opposite: Swahili is better preserved than English. The cue validity theory (Bates et al., 1991) predicted equal performance in both languages and no difference between past, present and future. This is also the wrong prediction. There is a selective deficit for Time Reference to the past, but past reference in Swahili is better preserved than in English. The reason may be that Swahili’s large and complex verb paradigm consists of several morphologically important affixes attached to the verb root. An omission of any single morpheme including tense prefix from the Swahili verb paradigm would render it a non-word, whereas an omission of past Tense suffix (-ed) does not render the English verb a non-word. Given their morphological importance, the affixes on the Swahili verb paradigm are generally more firmly anchored in the language system, than the affixes on the English verb. We, therefore, think Bates and colleagues’ (1987; 1991) assumption cannot explain the quantitative difference between the languages, nor can it explain the selective deficit for verb forms referring to the past. Interestingly, for this single category (past tense), English is more complex with its multiple regular allophones and irregular past tense forms. A better explanation is based on the morphological differences regarding the past Tense marker on both languages. While Swahili past Tense has only a single invariant, regular form marked as a prefix on the verb paradigm, the English past Tense has both regular (with different phonetic realizations: /d/, /t/, and /Id/) and irregular forms. It may thus be the case that English past Tense requires more processing resources compared to Swahili past Tense: the agrammatic speaker has to consistently monitor whether the English past Tense takes regular or irregular form; and appropriately select the correct phonetic realization for the regular past Tense. This assumption is in fact in line with the findings of Tsapkini, Jarema & Kehayia (2001): their Greek agrammatic speaker had particular difficulty with reference to the past. They attributed this difficulty to Greek verb morphology which is characterized by several allomorphs specifying Tense and Aspect. In Abuom & Bastiaanse (2012), we argued that apart from the differences in morphological complexity of the two languages, Swahili and English, the 3.4. DISCUSSION 85 regularity of the Tense affixes may also play a role. It has been suggested that bilingual agrammatic speakers are likely to perform better in a language with complex but regular morphology than in a language with simple but not regular morphology (Goral, 2011). In usage-based frameworks such as Bybee (2007) and Dressler (1985), regular morphological process is associated with affixation and considered to be more ‘natural’, highly frequent in most languages and generally easier to process. In contrast, irregular morphological processes are associated with stem change and considered to be less ‘natural’, infrequent in most languages and generally difficult to master. Although we did not find any dissociation between regular and irregular past Tense verbs in English, the fact that irregular forms exist may make the whole category more difficult to process than members of the very regular Swahili paradigm. Some studies have shown that individuals with impaired linguistic abilities, such as agrammatic speakers, prefer to produce highly frequent linguistic morphemes over less frequent linguistic items in their speech (see Centeno & Anderson 2011; Abuom & Bastiaanse, 2012). This suggests that affixation is preferable to stem change in linguistically impaired populations since affixes are more common within a single language and easier to process in general than stem changes. Swahili verb morphology for past Tense is regular (involves affixation) hence highly predictable and of high frequency compared to English past Tense which can be both regular and irregular, consisting of instances of infrequent affixations and internal stem changes. We, therefore, think Goral’s (2011) idea is a likely explanation for the different performance levels in Swahili and in English in the current study. In sum, reference to the past through verb morphology was impaired in these bilingual agrammatic individuals, both in production and in comprehension in both languages, as predicted by the PADILIH. The strong correlations in agrammatic speakers’ performance between the two languages show that there is one central deficit underlying the general performance patterns. However, the impairment was less severe in the language with the most complex, but also most regular paradigm, as suggested by Goral (2011), implying a compounding effect of the morphological system and reference to the past. 86 CHAPTER 4 Sentence Comprehension in Swahili-English bilingual agrammatic speakers Abstract For this study, sentence comprehension was tested in Swahili-English bilingual agrammatic speakers. The sentences were controlled for four factors: (1) order of the arguments (base vs derived); (2) embedding (declarative vs. relative sentences); (3) overt use of the relative pronoun ‘who’; (4) language (English and Swahili). Two theories were tested: the Trace Deletion Hypothesis (TDH; Grodzinsky, 1995), that assumes a representational deficit in agrammatic aphasia and the Derived Order Problem Hypothesis (DOP-H; Bastiaanse & Van Zonneveld, 2005), which is a processing account. Both theories have the same predictions for sentences in derived order. The difference is that the TDH predicts chance level performance for sentences in which the arguments are not in base order, whereas the DOP-H predicts poorer performance when processing This study has been published as: Abuom, T., Shah, E., & Bastiaanse, R. (2013). Sentence Comprehension in SwahiliEnglish bilingual agrammatic speakers.Journal of Clinical Linguistics & Phonetics, 27, 355-370. 87 88 CHAPTER 4. SENTENCE COMPREHENSION demands increase. The results show that word order influences performance, in that sentences in which the arguments are in derived order are harder to comprehend than sentences in which the arguments are in base order. However, there is a significant interaction with the factor ‘embedding’: sentences with an embedding are harder to comprehend than simple declaratives and this influence is larger in derived order sentences. There is no effect of language, nor of the use of a relative pronoun. These results are correctly accounted for by the DOP-H. 4.1 Introduction Individuals with agrammatic Broca’s aphasia are generally assumed not only to produce slow and effortful speech, with short phrases consisting of mainly content words, but also to have a relatively intact language comprehension ability. However, there is substantial empirical evidence to suggest that certain linguistically complex sentence types are difficult to comprehend for monolingual agrammatic speakers (cf. Caramazza & Zurif 1976; Grodzinsky 2000; Bastiaanse & Edwards, 2004). Particularly vulnerable are the semantically reversible sentences whose arguments are not in their base position, such as object clefts, passives, and object relative sentences. Several theories have been formulated to account for this phenomenon, some of which have contributed substantially to our understanding of language comprehension in individuals with agrammatic Broca’s aphasia. However, the focus has been almost exclusively on monolingual speakers, with hardly any attention given to comprehension patterns in bilinguals, especially to speakers of two morphologically different languages. The current study examines comprehension patterns in pre-morbidly highly proficient bilingual speakers of Swahili and English, two languages that possess contrasting morphological and syntactic properties. We start by giving some background on relevant Swahili syntax and verb morphology. Next, we provide an overview of comprehension theories in agrammatic aphasia. This will be followed by a review of some previous studies on Swahili-English bilingual agrammatic aphasia. We conclude the introduction by providing the aims and predictions for the current study. 4.1. INTRODUCTION 4.1.1 89 Swahili syntax and verb morphology Syntax (word order) According to Ashton (1982), Swahili, like many other Bantu languages, is a highly agglutinating and mostly prefixing language with a fairly fixed base word order (subject / agent - verb - object / theme: SVO), where the agent precedes the verb and the theme (see 1a-b). Although the second illustration (1b) is a subject relative clause, the agent and theme maintain the base order. We call these ‘base order sentences’. Sentences with derived order of the arguments, such as passives (subject / theme - verb - Prepositional Phrase / agent), object relatives (object / theme - subject / agent - verb) with and without complimentizer are also possible. These we call ‘derived order sentences’. In derived order sentences, the arguments are no longer in their base positions as exemplified in (2a-b). (1) (a) Base order, simple active (agent - theme) Kijana a-na-m-gonga msichana Boy s/he-PRESENT-him/her-hit girl ‘The boy is hitting the girl’ (b) Base order, subject relative (agent - theme) Kijana ambaye a-na-m-gonga msichana Boy who s/he-PRESENT-him/her-hit girl ‘The boy who is hitting the girl’... (2) (a) Derived order, passive (theme - agent) na kijana Msichana a-na-gong-wa s/he-PRESENT-hit-PASSIVE by boy Girl ‘The girl is hit by the boy’... 90 (b) CHAPTER 4. SENTENCE COMPREHENSION Derived order, object relative + relative pronoun (theme - agent) Msichana ambaye kijana a-na-m-gonga Girl who boy s/he-PRESENT-him/her-hit ‘The girl who the boy is hitting’ ... (c) Derived order, object relative - relative pronoun (theme - agent) Msichana kijana a-na-m-gonga Girl boy s/he-PRESENT-him/her-hit ‘The girl the boy is hitting’... In a passive sentence (2a), the theme is in clause-initial position, and the agent is in postverbal position in the prepositional phrase. The final suffix of the verb complex also changes from ‘-a’ to passive marker ‘-wa’. As in English, the phrase ‘na’ (by) is included only if the information that follows is important for clarity purposes to the reader or the listener. In an object relative sentence (2b), the theme is in initial position, the relative pronoun ‘ambaye’ (who) preludes the embedded sentence. However, just like in English, the relative pronoun may be left out, as shown in (2c). Verb Morphology The Swahili verb morphology is distinctly more complex than that of English, consisting of numerous affixes, both inflectional and derivational morphemes, attached to the verb root. These affixes (prefixes and suffixes) must occupy specific positions and they perform specific functions. The general position scheme of the affixes in relation to the verb root is shown in (3a). Some illustrations from Abuom et al. (2011) are given in (3b-d). (3) (a) Pre-prefix (Pp) + Subject prefix (Sp) + Tense marker (T) + Object prefix (Op) + ROOT + derivation (d) + Suffix (s) + Post-suffix (Ps). (b) A + li + m + gong + a Sp + T + Op + ROOT + d “S/he hit him/her” 4.1. INTRODUCTION (c) 91 Ha + tu + ta + m + gong + a Pp + Sp + T + Op + ROOT + d “We will not hit him/her” (d) Tu + na + gong + a + n + a Sp + T + ROOT + d + S + Ps “We are hitting each other” As illustrated in (3b-d), the Swahili verb complex, unlike that of English, can function as a complete sentence. The verbal complex consists of: the subject prefix (subject-verb agreement), the tense marker (includes tense and aspect) and the verb root, which are generally obligatory in every grammatical Swahili sentences. However, the object prefix is generally not obligatory when the object of the sentence is overtly present. The subject and object prefixes must always agree in number with the subject and the object of the sentence respectively. 4.1.2 Two theories on sentence comprehension in agrammatic Brocas’s aphasia Several linguistic theories have been proposed to account for the sentence comprehension deficit in monolingual agrammatic individuals11 . We discuss two theories here that formed the basis of this study. The Trace Deletion Hypothesis (TDH) proposed by Grodzinsky (1984) which has undergone a series of revisions (Grodzinsky, 1986; 1995; Drai & Grodzinsky, 2006) finds its origin in Chomsky’s Universal Grammar (see, e.g. Chomsky, 1995). According to this theory, constituents can ‘move out’ their original position, leaving behind a ‘trace’. The trace and the antecedent are, thus, linked as shown by coindexation in (4a-c). The thematic roles are assigned to the original positions in the sentences by the verb and they are transferred to the ‘moved’ constituent via the trace. 11 For comprehension studies, the term Brocas aphasia is often used. We prefer the term agrammatic aphasia here, to show that our participants were not only suffering from Brocas aphasia, but also spoke in telegraphic speech. 92 CHAPTER 4. SENTENCE COMPREHENSION (4) (a) The boyi is hit i by the girl. (b) The boyi whoi the girl is hitting i ... (c) The boyi the girl is hitting i ... According to Grodzinsky’s TDH, in individuals with agrammatic aphasia, the traces are deleted from the sentence representation due to damage to Broca’s area. As a result, the verb cannot assign a thematic role to the moved constituent. In order to interpret such sentences, the agrammatic individuals apply a non-syntactic ‘default strategy’ and assign an agent role to the first NP (‘the boy’) because usually the first NP is the agent. The second NP also gets an agent role, directly assigned by the verb. As a result, the aphasic individuals are faced with a representation consisting of two agents (the boy and the girl) and they have to guess, leading to chance level performance. In subject relatives (‘the boy who is hitting the girl...’), ‘the boy’ is not assigned a thematic role either, because the NP has been moved from its original position and the trace is deleted. However, the default strategy assigns the agent role to this first NP and the verb assigns the theme role to the second NP, resulting in a correct interpretation of the sentence. It is important to note that the TDH assumes that individuals with agrammatic aphasia suffer from a representational deficit. The sentence representations are damaged because the traces are lost. This means that they always apply the default strategy: they can never assign a thematic role to a moved argument. Therefore, their performance will be at chance for all sentences in which the order of the thematic roles is non-canonical (between 33-67% correct on a binary choice test) and above chance on sentences where the agent is in first position. The Derived Order Problem Hypothesis (DOP-H) has a larger scope than the TDH and it is meant to describe the production and comprehension deficits in agrammatic aphasia. It is based on data from many languages (Dutch, English, Turkish, Italian; see, for example, Bastiaanse et al., 2003; Bastiaanse & Thompson, 2003; Bastiaanse & Van Zonneveld, 2005; 2006; Yarbay Duman et al., 2007; 2008; 2011). Contrary to the TDH, it does not assume a representational deficit, but a processing disorder, meaning that the sentence 4.1. INTRODUCTION 93 representations are intact, but linguistic operations cannot always be performed correctly. The more complex the operations are, the more difficult processing will be. The underlying idea is that each language has a base order in which constituents that naturally belong together are adjacent. In English, for example, the base word order is Subject-Verb-Object (SVO), whereas, in Dutch and German it is Subject-Object-Verb (SOV). All other word orders are derived. Notice that the DOP-H is relatively theory-free: although embedded in the Chomskyan tradition, there is no strong idea about movement, merging, or traces. The DOP-H assumes that in order to produce or comprehend a derived word order sentence, extra grammatical operations have to be performed and this is exactly what is difficult for agrammatic speakers. For production this means that predominantly simple, base order sentences will be produced and for comprehension that semantically-reversible sentences with derived order of the arguments will be poorly understood. The DOP-H assumes, just like the TDH, that agrammatic individuals, when faced with a sentence with derived word order, may resort to a default strategy that the first NP is the agent. The main difference between the TDH and the DOP-H is that the DOP-H does not assume that the agrammatic individuals can never parse (or produce) a sentence with derived order of the arguments correctly; they cannot do this correctly all the time and the more processing is required, the more errors they will make. With regard to performance level this implies that the TDH predicts chance level performance for all sentence types with derived order of the arguments. The DOP-H predicts poorer performance on sentences in which the order of the arguments is derived and that when further complexity is added, performance will further decrease. In a study similar to the current one, Yarbay Duman et al. (2011) studied agrammatic sentence comprehension in Turkish. Turkish has a relatively free word order and a complex case system. Interestingly, some derived word order sentence types require a special case assignment system. For example, in object relatives, the subject has genitive case and the object has nominative case. Yarbay et al. (2011) found an interaction effect of word order and case: when case was not according to its default distribution (subject = nominative, object = accusative), performance diminished. This is compatible with a processing account like the DOP-H: comprehension of sentences with derived order of the arguments is impaired due to a processing disorder; adding complexity to these derived order sentences further diminishes performance. 94 CHAPTER 4. SENTENCE COMPREHENSION In sum, both the TDH and DOP-H predict that agrammatic speakers have difficulty understanding sentences in derived word order. While TDH predicts above chance level performance for base order sentences and chance level performance for derived order sentences, the DOP-H predicts better performance for base order sentences than for derived order sentences and further decline of performance when linguistic complexity is added to the sentence. There are two other accounts that could be relevant to our study. The first has to do with Working Memory (WM). It has been argued that comprehension of derived order sentences can be impaired because of an overload on the aphasic individuals’ working memory capacity which is generally assumed to be impaired (Burgio & Basso, 1997; Ivanova & Hallowell, 2012; King & Just, 1991; Miyake, Carpenter & Just, 1994). The question is, of course, how WM impairments may affect comprehension of some sentence types but not of others, that is, what units and / or processes influence the WM operations? It has been argued that the number of words may play a role, or rather the number of lexical items. The number of lexical items is the same in our sentences, but the number of words varies. Interestingly, the sentences are consistently shorter in Swahili than in English. Thus, if one accepts that more words in a sentence negatively affect agrammatic performance, then the Swahili-English bilingual agrammatic individuals should perform worse in English than in Swahili. An alternative WM account argues that it is the distance between the filler and the gap which is critical, that is, the more words between the filler (the argument that has been extracted from its base position) and the gap, (the resulting ‘empty’ base position), the harder it is to process the sentences, because the time that the filler should be kept in (the impaired) WM is longer (e.g. Frazier & Friederici, 1991; but see Friedmann & Gvion, 2003 and Gvion & Friedmann, 2012 for counter evidence). For checking this, a study on bilingual aphasia is particularly interesting, because, again, this distance is consistently shorter in all Swahili derived word order sentences than in the English counterparts. However, this account makes predictions for within language comparison as well. For example, for English, performance should be better on the object relatives without complementizer, since the distance is one word shorter. Another suggestion with respect to WM is that in long sentences the middle part is lost due to the WM deficit. In (5a-c) the resulting structures are given. 4.1. INTRODUCTION 95 (5) (a) passive the man is rescued by the woman → the man ... the woman (b) subject relative (...) the woman who rescues the man → the woman ... the man (c) object relative (...) the man (who) the woman rescues → the man ... rescues If one assumes something like a default strategy that makes the agrammatic individual assign the agent role to the first NP, then comprehension of passives and object relatives will be impaired on a binary choice test if the distractor depicts the same action as the target picture, but with the thematic roles reversed. They will perform relatively well on the subject relatives. However, when a test with four pictures is used, in which the two extra alternatives depict a different action with the same participants (one with the same agent theme relation as in the target sentence and one with the roles reversed), then the agrammatic individuals will point to one of the lexical distractors when hearing a subject relative or a passive, because the verb is wiped from WM. Although not the major focus of the current paper, we will address this WM account in the Discussion section. The data from Swahili-English bilingual agrammatic speakers provide a valuable site for testing the prediction of these theories, as these languages differ in their morpho-syntax: while English has relatively simple morphology, Swahili has a markedly agglutinative morphology with an extra passive marker on the verb. The question for the present study is whether the passive marker on the verb influences performance in Swahili. 4.1.3 Previous studies on Swahili-English bilingual agrammatic aphasia For the present study, the term ‘bilingual’ has been used to include multilingual persons who speak two or more languages or dialects in their everyday lives (Grosjean, 1994). It has been reported in several studies on bilingual aphasia that most individuals suffer from the same type of aphasia in all their languages acquired before an aphasia-producing incident; and the recovery pattern is often parallel (Fabbro, 1999, 2001; Paradis, 2001; Miozzo, Costa, Hernandez & 96 CHAPTER 4. SENTENCE COMPREHENSION Rapp, 2010; Abutalebi, Cappa, & Perani, 2005), however, non-parallel recovery patterns have also been reported (Albert & Obler, 1978; Fabbro, 1999). The differential recovery patterns are argued to stem from differences in the age of acquisition, frequency of use or proficiency level between the languages acquired before an aphasia producing incident. The aphasic participants in the present study are all balanced and proficient bilinguals, who acquired all their three languages (native language, Swahili and English) early in life. It is generally assumed that such bilinguals have shared language processing brain regions (e.g., Abutalebi et al., 2005; Miozzo et al., 2010). One of the consequences of shared language processing regions has been reflected in the limited number of studies that we have conducted on Swahili-English agrammatic speakers. In our first study, Abuom and Bastiaanse (2012), we analysed the production of verb forms for reference to the past and to the present in the spontaneous speech of early balanced agrammatic individuals. The proportion of verb forms referring to the present was normal in their spontaneous speech, but the proportion of verb forms referring to the past was significantly lower than normal. The pattern of performance in both languages was parallel. In our second study (Abuom & Bastiaanse, 2013), agrammatic speakers were tested on production and comprehension of verb forms referring to the past, present and future, using both sentence completion and sentence-picture matching tasks. Their performance showed a selective deficit of reference to the past on both comprehension and production tasks. Again, the pattern of performance was similar in both languages in spite of the morphological differences. What these studies show is that Swahili-English agrammatic speakers have a selective difficulty producing and comprehending sentences in base word order referring to a past time frame, but not to a present time frame. How they produce or comprehend sentences in derived word order is not known yet. The difficulty with base order sentences in the past time frame, which has been found in several other languages in monolingual agrammatic individuals (Bastiaanse et al., 2011; Bastiaanse, 2013), is reflected similarly in both languages of Swahili-English bilingual speakers. It is, however, unclear whether the difficulty with comprehension of sentences in derived word order, found in monolingual studies of other several languages (Dutch, English, Turkish, and German), is reflected in a similar way in Swahili-English bilingual agrammatic individuals. Furthermore, it is interesting to find out if the difficulty with derived order sentences affects both languages of bilinguals in a similar way, quantitatively 4.1. INTRODUCTION 97 and qualitatively. 4.1.4 The predictions for the current study Based on the studies of spontaneous speech and time reference in bilingual Swahili-English agrammatic speakers, we expect similar performance across the two languages regardless of the morphological differences in their verb inflection systems. With respect to the empirical validity of the two theories (TDH and DOPH) in Swahili-English bilingual agrammatic individuals the following is predicted. The TDH (Grodzinsky, 1986; 1995; Drai & Grodzinsky, 2006) predicts (a) above chance level performance for sentences in which the order of the arguments is canonical: active and subject relative sentences in both languages; (b) chance level performance for sentences in which the theme precedes the agent: passive, object relative sentences with and without relative pronoun. The DOP-H (Bastiaanse & Van Zonneveld, 2005) predicts (a) good performance for sentences with the arguments in base order: active and subject relative sentences; (b) poor performance on sentences with derived order of the arguments: passive, object relatives with and without relative pronoun. Furthermore, the DOP-H assumes that added linguistic complexity will lead to additional problems understanding sentences with derived order of the arguments. It is, thus, expected that the extra operation needed for the embedding will influence parsing. In other words, the DOP-H predicts an interaction effect of derived order and embedding, so that it predicts: simple active < subject relative < passive < object relative relative pronoun. In Swahili, the finite verb is morphologically more complex than in English and the study of Yarbay Duman et al. (2011) showed that case morphology interacts with derived order. However, case is a crucial factor for sentence comprehension in Turkish and depends on the sentence type. This is not so for Swahili verb morphology for most sentence types. However, it may be that the passive marker on the verb in Swahili facilitates or complicates performance. This can be accounted for by the DOP-H, but not by the TDH. A WM account predicts poorer performance on longer sentences or longer filler - gap distances. It predicts typical performance: role reversals in passives and lexical (+ role reversals) in object clefts. A final point that should be mentioned is that it is not a priori clear whether we should expect similar impairments in the two languages of the bilingual 98 CHAPTER 4. SENTENCE COMPREHENSION agrammatic individuals. According to Paradis (1988) and Fabbro (2001) a central underlying deficit in bilingual individuals with agrammatic aphasia may cause different surface manifestations in the languages that differ in their grammatical morphology. Taking into account the results of our previous studies in a Swahili-English bilingual agrammatic population, we expect to find similar patterns in both languages, although the extra verb morpheme in Swahili passives may lead to poorer performance in Swahili than English on this sentence type. 4.2 4.2.1 Methods Participants There were 11 non-fluent aphasic/agrammatic and 11 non-brain-damaged (NBD) participants in the study12 . Each participant spoke a Bantu or Nilotic or Indo-Aryan language natively, but all spoke English and Swahili as second languages, learned from 4 years old at school. The NBDs were, as a group, matched on age, native language, and education (a minimum of high school diploma: over 12 years of uninterrupted exposure to English and Swahili) with the agrammatic individuals. The agrammatic individuals were pre-morbidly highly proficient in the two second languages. All participants were righthanded and without a history of psychiatric or developmental speech or language disorders or any other neurological conditions. The aphasic individuals were assessed and diagnosed as suffering from aphasia by both the neurologist and speech therapist at the speech therapy department of the Aga Khan University hospital. The speech therapist confirmed that all the aphasic participants produced speech that was clinically judged as non-fluent and ‘telegraphic’ (slow, effortful with short and simple utterances consisting of mainly content words)13 . Their comprehension of single words in both languages was relatively good, based on their performance on an adapted version of the sub12 With the exception of 1 agrammatic individual (BM), the agrammatic participants participated in a previous study on time reference (see Abuom & Bastiaanse, 2013). Three of the agrammatic individuals (EA, MM & HJ) had also participated in an earlier study on bilingual agrammatic spontaneous speech (see Abuom & Bastiaanse, 2012). 13 Unfortunately, in Kenya there are no tests available to establish the aphasia syndrome. The BDAE cannot be used to classify the aphasia type in Kenya because of cultural bias. 4.2. METHODS 99 task for auditory comprehension of single words (nouns, verbs, colors, shapes, letters, numbers) from the Boston Diagnostic Aphasia Examination (BDAE: Goodglass & Kaplan, 1972). A few pictures of this task were substituted, because some items are unknown in Kenya and, hence, no Swahili word was available (for example, the hammock was changed to a swing). The level of comprehension of sentences in both languages was also found to be comparable based on their performance on a syntactic comprehension test (t (10) = 0.48, p = 0.64), a sub-test of Bilingual Aphasia Test (BAT: Paradis & Mwansau, 1990). The demographic details of the participants and their scores on the BDAE subtest for auditory word comprehension and on the BAT subtest for syntactic comprehension are shown in the Table 4.1. Considering their agrammatic output in both languages and their relatively good comprehension, it is assumed that they suffer from classical Broca’s aphasia. Table 4.1: Demographic details of the agrammatic individuals and their scores (% correct) on the test for auditory comprehension of words (on an adapted version of the BDAE-test) and on the subtest of Bilingual aphasia Test (BAT) in Swahili and English. EA JN SS PN MM JA MW HJ VK HS BM Mean 4.2.2 Age (years) Gender Handedness Education (years) Years post stroke/ head trauma Native language Swahili BDAEsubtest scores English BDAEsubtest scores Swahili BATsubtest scores English BATsubtest scores 42 50 30 36 47 46 50 45 25 64 53 44.36 M F F F F M F F F M F R R R R R R R R R R R 16 13 12 14 16 16 16 14 16 16 12 14.64 17 10 17 1 10 1 1.5 10 2 1 0.5 6.45 Nilotic Bantu Indo-Aryan Bantu Nilotic Bantu Bantu Nilotic Bantu Indo-Aryan Bantu 100 100 100 100 100 100 98.6 100 100 99 100 99.8 99 100 99 100 100 100 100 100 98.6 98.6 100 99.6 59 74 64 60 83 67 66 64 86 64 66 68.5 60 57 71 70 83 63 67 64 84 60 63 67.5 Materials and procedure An adaptation of the sub-test of the Verb and Sentence Test (Bastiaanse et al., 2002; 2003) for sentence comprehension to Swahili and English was used to test whether word order, embedding and verb complexity influence comprehension differently in the two languages of bilingual agrammatic individuals. The task in each language included 200 semantically reversible sentences (all in the present time frame) distributed equally into five sentence types: 40 active 100 CHAPTER 4. SENTENCE COMPREHENSION sentences; 40 passive sentences; 40 subject relative sentences; 40 object relative sentences with and 40 object relative sentences without relative pronoun. In Table 4.2, example sentences for each condition are given. All sentences had a transitive action verb with NPs in singular form. There were 40 sets of pictures, with each page consisting of a set of four different pictures. Table 4.2: Examples of target sentences in the sentence-picture matching task. Condition Sentence types Target sentence in English Target sentence in Swahili Base Order Active The man is rescuing the woman Point to the man who is rescuing the woman Mwanamme anamwokoa mwanamke Nionyeshe mwanamme ambaye anamwokoa mwanamke Mwanamke anaokolewa na mwanamme Nionyeshe mwanamke ambaye mwanamme anamwokoa Nionyeshe mwanamke mwanamme anamwokoa Subject relative Derived Order Passive Object relative + relative pronoun The woman is rescued by the man Point to the woman who the man is rescuing Object relative - relative pronoun Point to the woman the man is rescuing The tasks involved auditory sentence-picture matching. The participant was shown a set of four pictures on one page and was asked to look at them all. The examiner read a sentence aloud and asked the participant to point to the picture matching the sentence (see 6 and Figure 4.1). One picture matched the sentence and the three other pictures were distractors to help determine error types: reversed role distractor, lexical distractor, and reversed role + lexical distractor. 4.2. METHODS 101 Figure 4.1: An example of a set of pictures used in sentence-picture matching task taken from the VAST (Bastiaanse et al., 2002). (6) Experimenter: Mwanamme a-na-m-wokoa mwanamke Man he/she-Present-him/her-rescue woman ‘The man is rescuing the woman’ The two languages were tested separately on two different days with an interval of two weeks for each of the agrammatic individuals. The order of the tests in each language was varied for each participant: either English first, followed by Swahili or vice versa. To ensure the participants understood the task before starting the test, each session began with ten practice items during which corrections were allowed and feedback given. No further feedback was given once the test began. Each session lasted two hours, with a break for the 102 CHAPTER 4. SENTENCE COMPREHENSION agrammatic speakers. All the sessions were held in a speech therapy room at the Aga Khan University Hospital in Nairobi, Kenya. 4.3 Results The NBDs made no errors on either test. The results of the agrammatic speakers are shown in Table 4.3 and 4.4. Table 4.3: The percentages correct of the agrammatic individuals on the Swahili test. BASE ORDER Subject Actives Relatives EA JN SS PN MM JA MW HJ VK HS BM Mean 100 100 100 100 100 100 98 100 100 100 100 99.8 95 95 100 93 100 100 98 100 100 100 100 98.3 DERIVED ORDER Object Object Passives Relatives+ Relatives63 68 85 60 80 63 55 78 80 68 63 69.4 40 60 55 40 68 58 58 73 63 53 53 56.5 60 53 33 40 68 53 63 70 58 50 50 54.4 In order to test the TDH, it was compared whether or not the agrammatic individuals performed at chance level. Since virtually all errors (>95%) were simple role reversals, chance level was set on 50%, 16,7% correct (33.3%66,7%). It can be seen from the tables that the TDH correctly predicts the performance on most sentence types, but not on the passives in English, on which the agrammatic individuals perform slightly above chance level. However, what the TDH ignores is the different level of performance between the passives and the object relatives. An ANOVA was applied, to test for interaction effects between derived word order and embedding, as predicted by the DOP-H. We performed a repeated measures analysis of variance to investigate main 4.3. RESULTS 103 Table 4.4: The percentages correct of the agrammatic individuals on the English test. BASE ORDER Subject Actives Relatives EA JN SS PN MM JA MW HJ VK HS BM Mean 95 100 100 100 100 100 100 100 100 100 98 99.4 95 95 95 95 100 98 100 100 100 98 95 97.4 DERIVED ORDER Object Object Passives Relatives+ Relatives60 53 83 65 78 55 80 65 50 63 53 64.1 43 43 60 48 63 48 55 75 40 45 48 51.6 38 50 53 38 70 35 65 80 53 40 45 51.5 effects of language (English and Swahili), word order (base and derived) and embedding (declaratives and relatives) on agrammatic performance. There was no statistically significant effect for language (F (1, 10) = 2.882, p =.120: Swahili and English were similarly affected. There was a statistically significant main effect for word order: F (1, 10) =321.723, p =.000). Furthermore, there was a statistically significant main effect for embedding: F (1, 1 0) = 23.056, p =.001). There were no interaction effects for language x word order (F (1, 10) =1.996, p =.188) nor for language x embedding (F(1,10) =0.078, p =.786). However, there was a significant interaction effect for word order x embedding (F (1, 10)=19.207, p =.001). The interaction effect of word order x embedding is graphically shown in Figure 4.2. T-tests were performed to further explore the significant effects. Both in Swahili and in English, there was a significant effect of argument order: performance on sentences with derived order was worse than on sentences with base order of the arguments (Swahili: t (10) =17.86, p <.0001; English: t (10) =14.26, p <.0001). Performance on embedded sentences (subject relatives and both object relatives) was worse than on the declarative sentences (actives and 104 CHAPTER 4. SENTENCE COMPREHENSION 100 100 80 80 60 60 40 40 20 20 0 0 base order Swahili simple derived order Swahili embedded base order English simple derived order English embedded Figure 4.2: Illustration of the interaction effects for order x embedding in Swahili and English. passives) in both languages (Swahili: t (10) =7.054, p =.0002); English: t (10) = 6.982; p =.0002). Although there was no effect for language, we separately analyzed whether the verb morphology interacts with comprehension. In Swahili, the passive marker is part of the finite verb, whereas in English it is a preposition. This morphological difference does not influence performance (t (10) =1.288, p =.228). 4.3.1 Error analysis The distribution of error types of the agrammatic speakers is presented in Table 4.5. The agrammatic speakers pointed to each of the distracters, but not to the same extent. Role reversal errors were the most prevalent in both Swahili (95.7%) and English (96.8%). When agrammatic speakers heard sentences in derived order condition (passive, object relatives with and without relative pronoun), they most often pointed at the distracters in which the roles were reversed. A few lexical errors were also noted in Swahili (4.1%) and in English (2.7%) and a few errors that involved a combination of both role reversal and lexical errors in Swahili (0.2%) and in English (0.5%). In sum, the agrammatic individuals performed more poorly on sentences in derived order condition than on those in base order condition and performance diminished when another complex grammatical operation has to be performed 4.4. DISCUSSION 105 (embedding), as shown by the interaction effect. They make predominantly role reversal errors in both languages when they fail to understand sentences in derived order condition. Table 4.5: Distribution of error types (raw scores) of agrammatic individuals in Swahili and English tasks. SWAHILI Role Reversal Lexical EA JN SS PN MM JA MW HJ VK HS BM Mean 4.3.2 50 45 51 63 33 51 50 32 39 51 52 47.0 7 5 0 3 1 0 2 0 1 1 2 2.0 Other 0 0 0 1 0 0 0 0 0 0 0 0.1 ENGLISH Role Reversal Lexical 64 63 40 61 36 64 54 32 63 59 62 54.4 4 0 3 1 0 2 1 0 0 3 3 1.5 Other 0 1 1 0 0 0 1 0 0 0 0 0.3 Swahili and English compared Although the factor ‘language’ did not influence the performance, we looked at the correlation between the scores on the two languages, to see whether the agrammatic individuals are similarly affected in both languages. This is clearly the case: there is a strong and significant correlation between their scores in both languages: R = 0.92, p <.0001. 4.4 Discussion The present study examined the comprehension of sentences with base and derived order of the arguments with and without embedding by bilingual agrammatic individuals. The results show that sentences in the derived order condition are more difficult to comprehend than those in base order and that there is an interaction effect for embedding. A similar pattern of performance is 106 CHAPTER 4. SENTENCE COMPREHENSION reflected in both languages, Swahili and English, irrespective of their morphological differences. This implies that the morphological differences in the verb inflection system of the two languages do not play a role in comprehending the sentences. We discuss these findings in relation to the theories of sentence comprehension in agrammatism mentioned in the introduction. 4.4.1 The problems with derived order Based on Grodzinsky and colleague’s (1995; 2006) TDH, the following predictions were made for the current study: (a) above chance level performance for sentences in base order condition, actives and subject relative clauses, in both languages; (b) chance level performance for sentences in derived order condition, passives and object relatives with and without relative pronouns. Although our test had 4 pictures, only one of the distractors was purely related to word order and this was the picture chosen in the large majority of the errors. We, therefore, assumed that chance level was 50% 16.67. In both Swahili and English, the agrammatic individuals show good comprehension of both active and subject relative sentences with scores ranging from 93% to 100%. On the object relatives with and without relative pronoun, the scores are at chance level. So far, the scores are in line with the TDH. On passives in Swahili, however, the agrammatic individuals score above chance, which is incompatible with the TDH. The difference with chance level is not very large (69.4%), this alone is not a reason to reject the TDH, because there is overwhelming evidence that agrammatic individuals show around chance level performance on sentences in which the arguments are not in canonical order (see Grodzinsky et al., 1999). However, what the TDH does not predict is the interaction effect between derived order of the arguments and embedding. The DOP-H is a processing account that assumes worse performance when the syntactic operations become more complex, but it is does not necessarily predict performance at chance level. The specific predictions were that sentences with derived order of the arguments are harder to understand and that embedding will further diminish performance, or, in terms of statistics, that there is an interaction effect of word order and embedding. This is exactly what has been found and this effect is largest in the derived order sentences. That does not mean that sentences with embeddings in which the arguments are in base order are processed faultlessly. It may very well be the case that these sentences are also harder to process, but that application of a default 4.4. DISCUSSION 107 strategy results in almost perfect performance. On the basis of the current data, we cannot exclude this possibility (but see Friedmann, 2008, who showed that agrammatic individuals have problems understanding embedded sentences without traces). 4.4.2 A working memory deficit? Depending on how one would like to measure a WM deficit, different predictions were considered in the Introduction, all of which are falsified by the data of the current study. The short Swahili sentences are comprehended equally well / poor as those in English in all conditions. Filler - gap distance does not play a role either: the object relatives with and without complementizer are understood equally well. A WM-theory that assumes that the beginning and end of the sentence are retained in memory, whereas the middle part is wiped can account for the poor performance on the object relatives, but not on the passives. If in the latter structure the middle part of the sentence, the verb, is not available, lexical distractor pictures will be chosen, particularly those with role reversal, since the first NP will be interpreted as the agent. However, on the whole test by all agrammatic individuals, only 4 of these errors are made, 1 in Swahili and 3 in English. Also, lexical errors are expected for the subject relatives that have the verb in the middle, although without role reversals, because the first NP is the agent. In English in 1% and in Swahili in 0.5% of the items, such a distractor picture was chosen. Therefore, we do not think that a WM deficit can account for the data of the current study. 4.4.3 The effect of the morphological difference between Swahili and English In Swahili, the finite verb is very complex, encapsulating information about the subject, the object, negation, time reference and, in passives, the finite verb contains the passive marker. However, the language in which the agrammatic individuals were tested did not influence their behaviour, implying that the verb complexity does not contribute to their level of performance. This is interesting, because another factor, embedding, does affect their comprehension. Hence, the data suggest that performance, on a test where word order is a crucial clue to understand the sentence, is influenced by structural complexity and not by morphological complexity. That does not mean that their comprehension is 108 CHAPTER 4. SENTENCE COMPREHENSION not affected by verb morphology at all. From two previous studies (Abuom & Bastiaanse, 2012; 2013), we know that Swahili-English bilingual agrammatic individuals have problems with verb inflection when it is used for reference to the past (‘he pushed’ is more difficult than ‘he is pushing’ and ‘he will push’) in comprehension and in production in both languages. However, on the current experiment, the morphological complexity of the Swahili verb paradigm did not influence the agrammatic behaviour. As in our earlier studies (Abuom & Bastiaanse, 2012; 2013), there is a strong and significant correlation between the performance in both languages, meaning that when the agrammatic individuals are poor in one construction in one language, they perform poorly in the other language as well. The same was found in the present study, which is contrary to the assumption that a central underlying deficit in bilingual individuals with agrammatic Broca’s aphasia may cause different surface manifestations in the languages that differ in their grammatical morphology (Paradis, 1988, Fabbro, 2001). Given that these agrammatic speakers were early balanced bilinguals who acquired both languages early in life and were equally proficient in both languages at the time of their aphasia producing incident, we attribute their level of performance in both languages to shared language processing brain regions for derived order sentences (e.g., Miozzo et al., 2010; Abutalebi et al., 2005). In sum, in our group of bilingual agrammatic individuals, sentence comprehension is significantly hampered by derived word order. Embedding further diminished their performance. The latter finding is not compatible with the TDH, nor with any other representational theory. The DOP-H, that is very similar in its assumption that derived order of the constituents is hard for agrammatic individuals, is a processing theory. This means that the syntactic representations in agrammatic aphasia are intact, but application of the linguistic operations to extract information from word order is impaired when word order is derived. The present study demonstrates that the DOP-H makes correct prediction and also shows that application of another linguistic operation (embedding) further reduces correct interpretation of the complex sentences. CHAPTER 5 Sentence Production in Swahili-English bilingual agrammatic speakers Abstract Background: It has been argued that agrammatic speakers’ production of sentences in derived order is impaired (The Derived Order Problem Hypothesis, DOP-H), and that the underlying deficit in bilingual individuals with agrammatic Broca’s aphasia may cause different surface manifestations in the languages when they differ in terms of their grammatical morphology. The current study presents results of a study on sentence production in Swahili-English bilingual agrammatic speakers. The two languages, Swahili and English, differ significantly in terms of their verbal morphology. Aims: The current study tested the production of sentences in base and derived orders of arguments in the two languages of Swahili-English bilingual agrammatic speakers. Methods & procedures: Eight agrammatic and eight non-brain damaged This study has been published online as: Abuom, T. & Bastiaanse, R. (in press). Sentence production in Swahili-English bilingual agrammatic speakers. Aphasiology. DOI:10.1080/02687038.2013.810328 109 110 CHAPTER 5. SENTENCE PRODUCTION individuals participated in the study. A sentence elicitation test was used to examine the production of sentences in base and derived word orders in Swahili and English. The base order condition consisted of active and subject cleft sentences, whereas the derived order condition tested passive and object cleft sentences. Outcomes & Results: The non-brain-damaged individuals performed at ceiling in both languages. The agrammatic speakers’ results, however, show sentences in derived order condition were more difficult to produce than those in base order, similarly across the two languages irrespective of their morphological differences. Moreover, the embedded sentences were also more difficult to produce than simple sentences for agrammatic speakers. Conclusions: The current data partially support the DOP-H and provide new insight into sentence production deficit of bilingual individuals with agrammatic Broca’s aphasia. The findings are discussed with respect to the theories of sentence production in agrammatic speakers. 5.1 Introduction Agrammatic speech production is generally characterized by problems with free and bound grammatical morphemes (Goodglass, 1968; Caramazza and Berndt, 1985; Marshall, 1986), specifically with verb inflections for tense (Bastiaanse & Jonkers, 1998; Bastiaanse, Hugen, Kos & Van Zonneveld, 2002; Friedmann & Grodzinsky, 1997; Friedmann, 2000) and time reference (Bastiaanse, 2008; Yarbay-Duman & Bastiaanse, 2009; Bastiaanse, Bamyaci, Hsu, Lee & Yarbay Duman (2011). Apart from these problems with the production of grammatical morphemes, syntactic deficits have also been described: verbs with complex argument structures (the Argument Structure Complexity Hypothesis (ASCH); Thompson, 2003) and sentences in derived word order (the Derived Word Order Problem Hypothesis (DOP-H); Bastiaanse & Van Zonneveld, 2005) have been demonstrated to be hard to produce in spontaneous speech (Thompson, Shapiro & Schendel, 1995; Bastiaanse et al., 2002) as well as in sentence production tasks (Bastiaanse, Hugen, Kos & Van Zonneveld, 2002; Bastiaanse, Koekkoek & Van Zonneveld, 2003; Bastiaanse & Thompson, 2003; Burchert, Meiner & De Bleser, 2008; Yarbay Duman, Aygen, zgirgin, & Bastiaanse, 2007; Yarbay Duman, Aygen & Bastiaanse, 2008). To this end, a number of linguistic theories have been formulated to explain the source of agrammatic production 5.1. INTRODUCTION 111 problems (as discussed in the next section). However, all these theories are based on data from monolingual agrammatic speakers. Paradis (1988) argues that the same underlying deficit may cause different surface manifestations in an aphasic bilingual individual when the languages differ largely in terms of their grammatical morphology. The current study was meant to test the Derived Order Problem Hypothesis (DOP-H) on Swahili-English bilingual agrammatic speakers to determine the pattern of impairment across the two morphologically different languages. In the next section, we discuss some of the theories of agrammatic production, followed by a description of Swahili word order. Then our previous studies on Swahili-English bilingual agrammatic speakers will be discussed and we end with the predictions of the current study. 5.1.1 Theories of sentence production According to syntactic theories within the generative tradition (e.g., Chomsky, 1995; Pollock, 1989), when we produce and understand sentences, they are represented as phrase structures known as syntactic trees. The highest phrasal node in the tree is the complementizer phrase (CP), which hosts complementizers, which are embedding elements like “that”, and wh-morphemes such as “where” and “what”. Thus, the construction of embedded sentences or wh-questions, depends on the CP node being intact and accessible. Hagiwara (1995) was one of the first to claim that the top of the syntactic tree was hard to access for agrammatic individuals (but see Ouhalla, 1993). In 1997, Friedmann and Grodzinsky reported the results of a single case study of a Hebrew speaking agrammatic woman who showed dissociation between tense and agreement morphology in her speech production: tense inflection was impaired while agreement inflection was intact. The authors, by assuming that tense and agreement are represented in separate nodes of the syntactic tree where the tense node is located above the agreement node, as proposed by Pollock (1989), proposed the Tree Pruning Hypothesis (TPH: Friedmann and Grodzinsky, 1997) to account for this dissociation. The TPH claims that the syntactic tree of agrammatic speakers is pruned and higher nodes (from Tense node up, including the CP) are inaccessible. The consequence is that agrammatic speakers are unable to project the syntactic tree up to its highest nodes; hence, tense, wh-questions and embedded structures that require the high nodes should be impaired. 112 CHAPTER 5. SENTENCE PRODUCTION However, the TPH has been challenged by data from several languages based on verb inflection (e.g. Wenzlaff & Clahsen, 2004; 2005 and Burchert, Swoboda, & De Bleser, 2005 for German; Stavrakaki & Kouvava, 2003 and Nanousi, Masterson, Druks, & Atkinson, 2006 for Greek). Other studies showed that it is not only the top of the syntactic tree that is inaccessible: operations low in the syntactic tree are impaired as well (e.g. Bastiaanse et al., 2003; Burchert et al., 2008). Other key theories relate the agrammatic production difficulties to the complexity of sentence structures. Kim and Thompson’s (2000) study on English agrammatic speakers’ production of one argument verbs (e.g. ‘to run’) with two-place (e.g. ‘to read’) and three-place (‘to give’) verbs showed that threeplace verbs were more difficult to produce than two-place verb, which are more difficult than the one place verb. A follow up study by Thompson (2003) on agrammatic performance on unergative (e.g. ‘to sleep’) and unaccusative (e.g. ‘to fall’) verbs revealed a significantly better performance on unergative than unaccusative verbs. To account for these data, Thompson (2003) provided the Argument Structure Complexity Hypothesis (ASCH), attributing the sentence production deficits in agrammatic aphasia to the complexity of the argument structure of the verb: both the number of arguments and movement operations taking place at the surface level directly affect the ability to retrieve verbs and to construct sentences. Bastiaanse and Van Zonneveld (2005) tested the production of sentences with verbs with alternating transitivity, that is, verbs that have a transitive and an unaccusative reading (e.g. to break: ‘he breaks the glass’ vs ‘the glass breaks’) in Dutch agrammatic speakers. Agrammatic performance dropped significantly when sentences had to be produced with unaccusative verbs that require derived word order, that is, that the theme should be in subject position. Bastiaanse and Van Zonneveld (2005) interpreted these results in terms of the Derived Order Problem Hypothesis (DOP-H). The DOP-H is, although inspired by generative grammar, relatively theory-free. It assumes that all languages have a base order and that all other word orders are derived. The DOP-H predicts more problems for agrammatic speakers on producing and comprehending sentences when constituents are in derived order than when they are in base order. It has been supported by production studies in several languages. Bastiaanse et al. (2002a) reported that Dutch agrammatic speakers showed greater difficulty with finite verb production in the matrix clause (derived order) than embedded clause (base order). Basti- 5.1. INTRODUCTION 113 aanse and Thompson (2003) reported that English agrammatic speakers had more problems with auxiliaries in yes/no questions (derived order) than auxiliaries in declarative sentences (base order). In recent studies involving object scrambling (the grammatical object changes places with one of its adjacent constituent, resulting in derived order; this is a frequent structure in Dutch, German and Turkish, but it does not exist in English with its strict word order), both Burchert et al. (2008) and Yarbay Duman et al. (2007; 2008) reported that German and Turkish agrammatic speakers have more difficulties producing sentences in which the object is in derived position. Interestingly, Anjarningsih, Haryadi-Soebadi, Gofir, Bastiaanse (2012) found that agrammatic speakers of Standard Indonesian use a substantial number of passives with derived order in their spontaneous speech, contrary to what the DOP-H predicts. Anjarningsih et al. (2012) suggest that this can be explained by the high frequency of passive sentences in Standard Indonesian. Passive sentences are as frequent as, if not more frequent than, active constructions and the passive construction is the polite form to address people. Therefore the passive construction is firmly anchored in the language system. As mentioned, the DOP-H is an overarching theory that makes predictions for both production and comprehension. Bastiaanse & Van Zonneveld (2006) showed that it correctly predicts the performance on active and passive sentences in Dutch. However, there is more at stake in comprehension as two other studies from our group have demonstrated. Yarbay Duman, zgirgin, Altιnok & Bastiaanse (2011) tested comprehension of sentences with base and derived order in Turkish agrammatic individuals. Turkish uses case to denote syntactic roles. In simple declarative sentences, the subject gets nominative case and the object accusative case, but in structurally complex sentences, case assignment is marked. In object relatives, for example, the subject gets genitive case and the object gets nominative case. The agrammatic individuals had problems with derived order, as predicted by the DOP-H, but there was an interaction effect with case assignment. Derived order sentences were harder to comprehend when case assignment was marked. Similarly in our study on Swahili-English sentence comprehension, Abuom, Shah & Bastiaanse, 2013), the agrammatic individuals performed poorly on the sentences with derived order of the arguments but there was an interaction with the factor embedding: subject relatives (base order of the arguments + embedding) were more difficult than simple actives (base order of the arguments, - embedding), and 114 CHAPTER 5. SENTENCE PRODUCTION object relatives (derived order of the arguments, + embedding) were more difficult than passives (derived order of the arguments, - embedding). This was taken as support for a processing disorder that increases when more linguistic operations are required. Processing derived word order is such an operation, as well as embedding. Many theories on agrammatism are primarily developed to describe the problems with comprehension of semantically reversible sentences (e.g. the boy is chased by the girl ), such as the trace deletion hypothesis (Grodzinsky, 1995; 2000) or the mapping theory (see, e.g. Linebarger, Schwartz & Saffran, 1983; Schwartz, Linebarger, Saffran & Pate, 1987). The DOP-H has a larger domain: it predicts agrammatic problems with, for example, verb position in Dutch (Bastiaanse et al., 2002); auxiliaries in English yes-no questions (Bastiaanse & Thompson, 2003), clitics in Italian (Rossi, 2007) and object scrambling in Dutch (Bastiaanse et al., 2003) and German (Burchert et al., 2008). The present study focuses on word order in Swahili-English bilingual agrammatic speakers. The emphasis is on the order of the arguments to enable testing of the ‘overarching’ aspect of the DOP-H. In a previous study (Abuom et al., 2013), comprehension of actives, passives and subject and object relatives has been tested in this population. The current study uses similar sentence types. Before formulating the predictions for this study, we describe some relevant characteristics of Swahili grammar. 5.1.2 Swahili word order Swahili is a Bantu language spoken mainly in Africa, including Kenya where all participants in this study were drawn from. In Kenya, a multilingual society where an average person speaks at least three languages, the two most dominant languages across the population are Swahili and English. Apart from the two dominant languages, an average Kenyan uses one of the 42 native languages that linguists term “ethnic languages” mostly at home with family members. Each of the 42 languages has been classified as Bantu, Nilotic, Indo-Aryan or Cushitic. Swahili and English share the same status as second languages since both are acquired around the age of 4, after native language acquisition by the majority of Kenyans. Whereas Swahili is taught as one of the subjects from kindergarten to university and also functions as the national language and the language of politics, business, and daily interactions of people from different ethnic backgrounds, English functions as the official language of 5.1. INTRODUCTION 115 instruction in all educational institutions from primary school to university, and also as the language of news broadcasts, parliamentary proceedings, and business. Therefore, a Kenyan with over 12 years of uninterrupted education is generally expected to be equally and highly proficient in both languages. Swahili is an agglutinative language, with a fairly fixed base word order (SVO) at the sentence level, where the subject precedes the verb and the object, see (1a) (Ashton, 1982). Usually, the agent is in subject position and the theme in object position, like in English. Similarly, (1b) is a subject cleft sentence, a structure that allows the speaker to emphasize the agent while maintaining the base word order (agent - verb - theme). Other word orders are also possible in Swahili. For example, in the passive sentence (2a), the theme is in subject position in the sentence resulting to a derived order (theme - verb - agent). The final vowel of the verb complex also changes from ‘-a’ to passive marker ‘-wa’. As in English, the phrase ‘na’ (by) is included only if the information about the agent is important for clarity purposes to the reader or the listener.14 In an object cleft sentence (2b) the theme is in a position preceding the agent, resulting in derived order of the arguments (theme - agent - verb). (1) (a) Mama a-na-m-gonga baba Ag-V-The15 Mother s/he-PRESENT-him/her-hit father ‘The mother is hitting the father’ (b) Ni mama ambaye a-na-m-gonga baba Ag-V-Th It is mother who s/he-PRESENT-him/her-hit father ‘It is mother who is hitting the father’ (2) (a) 14 In Mama a-na-gong-wa na kijana Th-V-Ag Mother s/he-PRESENT-hit-PASSIVE by boy ‘The mother is hit by the boy’ both Swahili and English, the passive sentence is not simply an inversion of the thematic roles, like in sentences with object scrambling: the verb morphology changes into [auxiliary+participle], and the Subject NP becomes a PP (by-phrase). Theories on how this derivation takes place differ largely. What is important for the current study is the fact that he thematic roles are not in base order. 15 Ag=agent, V=verb, Th=theme 116 (b) CHAPTER 5. SENTENCE PRODUCTION Ni baba ambaye mama a-na-m-gonga Th-Ag-V It is father who mother s/he-PRESENT-him/her-hit ‘It is the father who the mother is hitting’ In both passive and object cleft sentences, operations have been applied resulting in arguments that are no longer in their base position. It is also clear from the examples (1-2) that the Swahili finite verb form is more complex than that of English. It consists of several affixes which are both inflectional and derivational morphemes, attached to the verb root. These affixes (prefixes and suffixes) must occupy specific positions and perform specific functions. The general position scheme of the affixes in relation to the verb root is shown in (3a). Some illustrations from Abuom, Obler & Bastiaanse (2011) are given in (3b-d). (3) (a) Pre-prefix (Pp) + Subject prefix (Sp) + Tense marker (T) + Object prefix (Op) + ROOT + derivation (d) + Suffix (s) + Post-suffix (Ps). (b) A + li + m + gong + a Sp + T + Op + ROOT + d “S/he hit him/her” (c) Ha + tu + ta + m + gong + a Pp + Sp + T + Op + ROOT + d “We will not hit him/her” (d) Tu + na + gong + a + n + a Sp + T + ROOT + d + S + Ps “We are hitting each other” The Swahili verb, unlike that of English, can function as a complete sentence. The verb paradigm consists of: subject prefix (subject-verb agreement), tense marker (includes tense and aspect) and verb root, which are generally obligatory in every grammatical Swahili sentence. However, the object prefix is usually not obligatory when the object of the sentence is overtly present. The subject prefix and object prefix must always agree in number with the subject and the object of the sentence respectively, as illustrated in (1-2) above. 5.1. INTRODUCTION 5.1.3 117 Previous studies on Swahili-English bilingual agrammatic speakers Among the several theories of agrammatic production mentioned above, only a few have been tested in bilingual speakers of structurally different languages to determine whether the same underlying deficit may cause different surface manifestations in the different languages of a bilingual (Paradis, 1988). Abuom and colleagues showed in several studies that Swahili-English agrammatic speakers have a selective deficit for production and comprehension of verb forms that refer to the past both in experimental settings (Abuom et al., 2011; Abuom & Bastiaanse, 2013) and in spontaneous speech (Abuom & Bastiaanse, 2012). Moreover, this impairment is more prominent in English with its relatively simple verb inflection paradigm than in Swahili with its extensive, but entirely regular paradigm. These findings show that these agrammatic bilinguals demonstrate the same pattern as agrammatic monolinguals (for an overview, see, Bastiaanse et al., 2011 and Bastiaanse, 2013). Having noted that Swahili-English agrammatic speakers, in the severity range tested, have no problems with the present time frame constructions in base word order, Abuom et al. (2013) tested their comprehension of present time constructions in derived word order. We compared the comprehension of sentences in two conditions: base order and derived order, in both languages (Swahili and English). The results demonstrated that Swahili-English agrammatic speakers had greater difficulty comprehending derived order constructions (passive and object relative sentences) compared to base order conditions (simple active and subject relative sentences). Furthermore, the performance on both languages was strikingly similar irrespective of the morphological differences between the two languages. The findings are compatible with the DOP-H. The DOP-H makes the same predictions for agrammatic comprehension and production, but so far, it has not been tested for similar constructions in both modalities. Moreover, the participants of the current production study are the same as those in the earlier comprehension study. This enables us to analyze whether comprehension and production are affected to the same extent, a finding that would point to central deficit. A final unique point of the current study is that the agrammatic speakers are early balanced bilinguals who were tested in both languages. 118 5.1.4 CHAPTER 5. SENTENCE PRODUCTION The current study We tested the production of different sentence structures, in which two factors were varied: word order (base and derived) and sentence type (simple declarative and embedding). Thus, we tested simple sentences with base and derived order of the arguments (actives and passives, respectively) and sentences with embeddings in base and derived order (subject and object cleft sentences, respectively). The DOP-H predicts more problems with the derived order sentences. However, it is likely that, as in comprehension, the problems increase when embedded sentences have to be produced. It is, therefore, predicted that subject clefts are more difficult than simple actives, although in both sentence types the arguments are in base order and that object clefts are more difficult than passives: Active sentences > subject clefts ≥ passive sentences > object clefts. It is unknown yet whether word order and embedding will have a similar complicating effect, making it unsure whether subject clefts will be more difficult than or equally difficult as passives. Studying these questions in a bilingual population enables us to evaluate whether other factors further complicate sentence production and whether word order and embedding are affected in the same way in two languages from two different language families. As described above, the verb system in Swahili is morphologically very extensive but also highly regular compared to English verbs. Such a complex system requires further processing resources, and may result in poorer performance in Swahili. It may also be the case that there will exist qualitative difference in the responses in the two languages. Thus, it may be that one language is more impaired than the other or that different error patterns are observed. 5.2. METHODS 5.2 5.2.1 119 Methods Participants There were 16 participants in this study: 8 bilingual agrammatic speakers of Swahili and English16,17 from the Aga Khan University hospital (Nairobi Kenya) and 8 non-brain damaged bilingual speakers (NBDs). All were aged between 26 and 51 years, with over 12 years of education. In 7 participants, aphasia was caused by a stroke, in 1 by traumatic brain injury. The agrammatic speakers were early balanced bilinguals equally proficient in English and Swahili pre-morbidly, and none had any history of neurological, hearing or vision problems. The non-brain-damaged participants were matched to the agrammatic speakers based on age, years of education, occupation and native language. Although there are no tests available to establish the aphasia syndrome in Kenya, the aphasic speakers were assessed and diagnosed by neurologists as suffering from aphasia. They were, further, judged by a speech therapist as agrammatic in English based on their spontaneous speech production that was perceived as being slow, effortful and ‘telegraphic’ (short phrases consisting of mainly content words), but with relatively spared auditory comprehension of single words on an adapted version of the BDAE-test (Boston Diagnostic Aphasia Examination: Goodglass, Kaplan & Barresi, 2001). Their demographic details and percentage scores on the test for auditory comprehension of words (on an adapted version of the BDAE-test) in Swahili and English are given in table 5.1. 5.2.2 Materials & Procedure A sentence production test for Swahili and English, developed using pictures from Verb and Sentence Test (VAST; Bastiaanse et al., 2002; 2003) was used to test whether word order, embedding and size of the verb inflection paradigm (morphological complexity) influence sentence production differently in the two languages of bilingual agrammatic speakers. The test in each language included 36 semantically reversible sentences distributed over two conditions: base order and derived order. The base order consisted of: 9 simple active sentences (e.g. 16 The 8 agrammatic participants had previously participated in a study on sentence comprehension a year earlier (Abuom et al., 2013). 17 The term bilingual has been used in this study to refer to all those people who use two or more languages or dialects in their everyday lives (Grosjean, 1994). 120 CHAPTER 5. SENTENCE PRODUCTION Table 5.1: Details of the agrammatic participants and their scores (% correct) on the test for auditory comprehension of words (on an adapted version of the BDAE-test) in Swahili and English. EA JN SS PN MM MW HJ VK Mean Age (years) Gender Handedness Education (years) Years post stroke/ head trauma Native language Swahili BDAEsubtest scores English BDAEsubtest scores 43 51 31 37 48 51 46 26 41.63 M F F F F F F F R R R R R R R R 16 13 12 14 16 16 14 16 14.63 18 11 18 2 11 2.5 11 3 9.56 Nilotic Bantu Indo-Aryan Bantu Nilotic Bantu Nilotic Bantu 100 100 100 100 100 98.6 100 100 99.8 99 100 99 100 100 100 100 98.6 99.6 the man is rescuing the woman and 9 subject cleft sentences (e.g. it is the man who is rescuing the woman) with embedding, but arguments in base order, that is, the agent precedes the theme. For derived order, there were 9 passive sentences (e.g. the woman is rescued by the man) with order of the arguments derived, that is, the theme precedes the agent and is in subject position (and the agent in the by-phrase), and 9 object cleft sentences (e.g. it is the woman who the man is rescuing) with derived order of the arguments, the theme precedes the agent, and an embedding. The influence of size of the verb inflection paradigm was controlled by the two languages that the agrammatic individuals speak: Swahili has an extensive verb inflection paradigm that consists of the subject, object, tense, aspect and passive markers (for passive sentences). In English the verb is only inflected for tense and agreement There were 36 pairs of pictures, a pair on each page, depicting the same action but the agent in the first picture was the theme in the second and vice versa. The infinitive of the verb was printed on top (see Table 5.2 for the sentence types and Figure 5.1 for an example picture). The pictures were shown to the participant who was asked to look at them both. The experimenter named the people and animals in the pictures to ensure proper recognition. He then constructed a sentence using the printed verb to 5.2. METHODS 121 describe the first picture and asked the participant to construct a sentence of a similar structure to describe the second picture. For example, for Figure 5.1: RESCUE Figure 5.1: An example of a pair of pictures used in sentence production task (VAST: Bastiaanse et al., 2002). Experimenter: Participant: (pointing to the picture on the left): for this picture, you can say “The woman is rescuing the man” (pointing to the second picture on the right): and for this picture, you can say... “The man is rescuing the woman” Four practice trials (one for each sentence type) preceded the experiment. During the practice trials the participant got feedback and when needed, the sentence was corrected and the participant was asked to repeat the correct sentence, to ensure that the task was properly understood. No help was given during the experiment and only neutral feedback was provided. There was no time limit for participant’s response, and repetitions of the prompting sentence were made as many times as requested. Self-corrections were allowed and only the last answer was scored. 5.2.3 Scoring All answers were transcribed in orthographic script and analyzed both quantitatively (correct / incorrect) and qualitatively (error analysis). 122 CHAPTER 5. SENTENCE PRODUCTION Table 5.2: Examples of target sentences in the sentence production task Sentence type Target sentence in English Target sentence in Swahili Active The man is rescuing the woman It is man who is rescuing the the woman The woman is rescued by the man It is the woman who the man is rescuing Mwanamme anamuokoa mwanamke Ni mwanamme ambaye anamuokoa mwanamke Mwanamke anaokolewa na mwanamme Ni mwanamke ambaye mwanamme anamuokoa Subject cleft Passive Object cleft 5.3 Word order Embedding base - base + der. - der. + Results There were no errors made by the NBDs on either test. The results of the agrammatic speakers are shown in Tables 5.3 and 5.4. Table 5.3: Individual agrammatic speakers’ correct raw scores on Swahili test Agrammatic participant EA JN SS PN MM MW HJ VK mean Actives Subject clefts Passives Object clefts Total derived order max=18 Total simple Total embedded max=9 Total base order max=18 max=9 max=9 max=9 7 9 9 9 9 9 8 9 8.6 2 3 7 6 5 3 5 7 4.8 4 4 6 3 5 3 5 5 4.4 max=18 max=18 0 2 3 2 4 1 5 4 2.6 9 12 16 15 14 12 13 16 13.4 4 6 9 5 9 4 10 9 7 11 13 15 12 14 12 13 14 13.3 2 5 10 8 9 4 10 11 7.4 A repeated measures analysis of variance was performed to investigate main effects of language (English and Swahili), word order (base and derived) and embedding (simple and embedded) on agrammatic performance. There was a significant effect for language: F (1, 7) = 8.430, p = .023, word order: F (1, 7) = 149.075, p < .000, and embedding: F (1, 7) = 62.906, p < .000. There were no interaction effects for language x word order (F (1, 7) =0.023, p =.885), nor for language x embedding (F (1, 7) = 0.955, p =.361) (see Figure 5.2). 5.3. RESULTS 123 Table 5.4: Individual agrammatic speakers’ correct raw scores on English test Agrammatic participant EA JN SS PN MM MW HJ VK mean Actives Subject clefts Passives Object clefts Total derived order max=18 Total simple Total embedded max=9 Total base order max=18 max=9 max=9 max=9 5 9 8 9 8 9 6 9 7.9 2 2 7 3 4 3 5 5 3.9 3 3 5 4 5 5 4 3 4 max=18 max=18 0 1 3 2 0 2 2 2 1.5 7 11 15 12 12 12 11 14 11.8 3 4 8 6 5 7 6 5 5.5 8 12 13 13 13 14 10 12 11.9 2 3 10 5 4 5 7 7 5.4 However, there is a significant interaction effect for word order x embedding (F (1, 7) = 8.188, p = .024). Overall there is no interaction for: language x word order x embedding (F (1, 7) = 0.549, p = .483). We performed t-tests to explore the significant effects further. The agrammatic speakers performed significantly poorer in English than in Swahili (t (7) = 2.903, p = 0.0270). They also showed significant difficulty producing sentences in derived order compared to those in base order condition (Swahili: t(7) = 8.450, p < .0001; English: t(7) = 11.180, p < .0001); They performed poorly on passives compared to active sentences (Swahili: t(7) = 9.379, p < .0001; English: t(7) = 6.675, p < .0001), and worse on object clefts than subject clefts (Swahili: t(7) = 4.123, p = .004; English: t(7) = 5.158, p = .001). Finally, the performance on embedded sentences (subject clefts and object clefts) was worse than on the declarative sentences (actives and passives) in both languages (Swahili: t(7) = 6.666, p <.0001; English: t(7) = 7.021; p < .0001). 5.3.1 Error types The distribution of error types of the agrammatic speakers is presented in Tables 5.5 & 5.6. Errors produced by the agrammatic speakers were distinguished into four categories: word order (when the arguments are not in the correct position: “The man is rescuing the woman” Target: “The woman is rescued by the 124 CHAPTER 5. SENTENCE PRODUCTION 9 9 8 8 7 7 6 6 5 5 4 4 3 3 2 2 1 1 0 0 base order derived order simple base order embedded simple derived order embedded Figure 5.2: Illustration of the interaction effects for word order x embedding in Swahili and English Table 5.5: Individual agrammatic speakers’ distribution of error types (raw scores) on Swahili task. EA JN SS PN MM MW HJ VK mean Word order Embedding Word order & embedding Others: repetition of given verb form 5 5 3 6 4 6 4 4 4.6 7 6 2 3 4 6 4 2 4.3 9 7 6 7 5 8 4 5 6.4 2 0 0 0 0 0 1 0 0.4 man”), embedding (when the relative marker ‘who’ or ‘ambaye’ is omitted’ in the sentence: “the man is hitting the woman” Target: “It is the man who is hitting the woman”) both word order and embedding (when the sentence produced is the reverse of the target sentence and the relative marker is missing: “The man is guarding the woman” Target: “It is the woman who the man 5.3. RESULTS 125 Table 5.6: Individual agrammatic speakers’ distribution of error types (raw scores) on English task. EA JN SS PN MM MW HJ VK mean Word order Embedding Word order & embedding Others: omissions of ‘is’; repetition of given verb form 6 6 4 5 4 4 5 6 5 7 7 2 6 5 6 4 4 5.1 9 8 6 7 9 7 7 7 7.5 4 0 1 0 1 0 3 0 1.1 is guarding”), other (repetition of the given form of the verb: “*Man rescue woman” , omission of ‘is’ for English: “*Man rescuing woman”). Errors of word order and embedding combined occurred equally often in Swahili and English (t (7) = 1.843; p = .108). Other errors, such as repetition of the given verb form in both languages and the omission of ‘is’ in English, hardly occurred. In sum, the results show that Swahili-English agrammatic speakers have more problems producing sentences in derived order - passives and object cleft constructions - than in base order condition. However, within both orders, base and derived, the embedded variant is more difficult than the variant with simple sentence structures. The most common errors are production of sentences with arguments in base position where derived orders was required and of sentences without embedding in the embedded condition. In object clefts both the order and the embedding were often avoided. These findings imply that it is not only the derived order that is a problem to agrammatic speakers, but also embedded constructions. 126 5.3.2 CHAPTER 5. SENTENCE PRODUCTION Swahili and English compared The Swahili version of the test was easier for the agrammatic speakers than the English version. In order to find out whether the grammatical disorder influences word order and embedding in Swahili and English in the same way, the correlation between the scores in the two languages was calculated. The results show that a common underlying impairment is very likely: there is a strong and significant correlation between the scores in the two languages (R = .885, p <.0001). Notice that performance is better in Swahili than in English, despite the larger verb inflection paradigm in Swahili. We will come back to this in the Discussion. 5.3.3 Production and comprehension compared The DOP-H aims to be an overarching theory and has been supported in several studies with production and comprehension data, however, never in the same population. In order to claim a theory to be overarching, the same structures should be impaired to the same degree in the same individuals. To test this, correlations between comprehension and production were calculated. Each of the 8 agrammatic speakers of the current study participated in the comprehension study as well. There is a strong and significant correlation between the production and comprehension scores, both in Swahili (R =. 759, p < .0001 and in English (R = .692, p < .0001). 5.4 Discussion The primary goal of the current study was to test the DOP-H, which predicts difficulty in production of derived order sentences by the agrammatic speakers. However, an interaction with linguistic complexity other than word order was found. Moreover, by testing bilingual agrammatic speakers of two structurally different languages, we were able to measure whether the size of the verb inflection paradigm played a role. Finally, by comparing the data with those on a similar comprehension test, the overarching aspect of the DOP-H could be tested. 5.4. DISCUSSION 5.4.1 127 The DOP-H and linguistic complexity The data show that Swahili-English bilingual agrammatic speakers’ performance on sentence production is affected by word order. Sentences with base word order are produced better than those with derived word order: relatively few errors were made on active and subject cleft sentences compared to passive and object cleft sentences, respectively. This is exactly what the DOPH predicts. Several studies on production involving monolingual agrammatic speakers have supported this hypothesis as well: Bastiaanse et al., (2002a) for Dutch; Bastiaanse and Thompson (2003) for English; Burchert et al. (2008) for German; and Yarbay Duman et al. (2007; 2008) for Turkish. However, two comprehension studies showed that the DOP-H alone does not sufficiently predict the observed patterns. When statistical comparisons are made it is evident that there is an interaction between derived word order and other levels of linguistic complexity. Yarbay Duman et al. (2011) showed that for Turkish agrammatic individuals derived order sentences are hard to interpret, and within the derived order structures, there is also a hierarchy. Performance on derived order sentences with unmarked case assignment (subject nominative case and object accusative case, as in sentences with object scrambling) is better than when case assignment is marked (subject genitive case and object nominative case, as in Turkish object relative sentences). A similar interaction effect was reported by Abuom et al. (2013) for comprehension in SwahiliEnglish bilingual agrammatic individuals. Derived order sentences were harder to comprehend than base order sentences, but there was an interaction with embedding. For both languages, the order of difficulty was: simple actives > subject relatives > passives > object relatives. The same is observed in the present study on agrammatic production (for the same individuals who were tested on comprehension). Derived order is indeed a problem, but embedding is as well and sentences with both derived word order and embedding are hardest to produce: simple actives > subject clefts > passives > object clefts. The error analysis showed that in both languages, the agrammatic speakers frequently omitted the relative pronoun ‘who’ in English and ‘ambaye’ in Swahili suggesting a general difficulty in producing embedded clauses. This is an aspect that is not included in the DOP-H. The question is why embeddings are difficult. Of course, embedding is a difficult operation that is usually introduced by a complementizer. The complementizer itself, being a function word, may be the problem. However, if it is the production of the function 128 CHAPTER 5. SENTENCE PRODUCTION word, rather than the linguistic operation, that is difficult, then one expects agrammatic speakers to produce embedded sentences in which optional complementizers are omitted (such as ‘I think he will come home’ and ‘the boy reading the newspaper is absent-minded’). However, agrammatic spontaneous speech data show that hardly any embeddings are produced, suggesting that it is embedding that is the problem, rather than the production of the complementizer per se. (see, e.g. Abuom & Bastiaanse, 2012, for data on Swahili-English agrammatic spontaneous speech, Bastiaanse et al., 2002a , for Dutch spontaneous speech and Saffran, Berndt & Schwartz, 1989, for English spontaneous speech). The problem with embedding in agrammatic speech production is a robust phenomenon that has previously been reported in several studies in aphasiology literature. Data from spontaneous speech in various languages show that agrammatic speakers have severe difficulties with the production of embedded sentences which manifest in their avoidance of complex sentences and in errors they make when they try to produce them. This was reported for English (Thompson et al., 1996, 1997; Bates, Friederici, Wulfeck, & Juarez, 1988), for Italian and German (Bates et al., 1988), French (Nespoulous, et al. 1988; 1990), Japanese, (Hagiwara, 1995) and Dutch, Swedish, Polish, and Finnish in Menn and Obler’s (1990) corpora. The theory of Hagiwara (1995) and the TPH Friedmann and Grodzinsky, 1997) attribute this difficulty to a pruned syntactic from the T node up, including the CP, in agrammatic speakers’ syntactic representation. Since the relatives ‘who’ and ‘ambaye’ are in complementizers, the TPH correctly predicts these omissions. When the TPH assumes that the subject of a passive sentence is in the specifier of the complementizer phrase, it can also explain the problems with passive sentences. However, the TPH is not an adequate alternative to explain the current data. First, the TPH does not predict that object clefts are more difficult than subject clefts. Second, the TPH predicts tense problems and the agrammatic speakers in the current study did not make one single tense error. The TPH, therefore, partly explains the problem with embedded structures observed in the current data, but it does not account for the difficulties with derived order sentences. We, thus, conclude that both the DOP-H and the TPH can only account partly for the current data. While the DOP-H explains the difficulty with derived order sentences, the TPH can account for the difficulty with embedded structures. A unifying theory that can combine both problems is still lacking. 5.4. DISCUSSION 5.4.2 129 Production in Swahili and English One of the questions was whether the extensive Swahili verb forms would result in poor performance compared to the relatively simple English verb forms we used. We deliberately used those forms that were most accurately produced in an earlier study (Abuom & Bastiaanse, 2012; 2013) on time reference: present imperfect for Swahili and the gerund for English. In these studies, the same pattern was observed as in the current study: performance in Swahili, with its extensive verb morphology, was significantly better than in English. Hence, we can conclude that the complexity of the verb form does not influence production. However, like in two studies on time reference through verb inflection and in the study of Abuom et al. (2011), performance was better in Swahili. Given the relatively poor but irregular morphological system of English in contrast with the stronger and more regular system in Swahili, the current finding is in line with Menn and Obler (1990). They suggest that in languages with a small paradigmatic system (such as Swahili) fewer errors in grammatical morphology are made than in languages with a broader range of choices within the paradigm (such as English. It cannot be excluded, of course, that the agrammatic speakers are (all) more proficient in English than in Swahili. However, this is quite unlikely, because both languages are used on a daily basis, both before and after the brain damage. An alternative explanation is that embedding and derived word order are more frequent in Swahili than in English. Such frequency counts are not available for Swahili. However, in our spontaneous speech study, both the bilingual non-brain-damaged and the bilingual agrammatic speakers used more embedded sentences in English than in Swahili (Abuom & Bastiaanse, 2012). Also, Bastiaanse, Bouma and Post (2009) showed that the frequency of sentence structures does not influence agrammatic performance. In Dutch, for example, the order Subject - Verb - Object is 50% more frequent than the structure Subject - Object - Verb. However, agrammatic speakers are significantly better in producing the latter order, which is the base order in Dutch. SVO is, despite being more frequent, derived. However, the similar pattern of performance across the two languages irrespective of their morphological differences, even though Swahili appears slightly better preserved than English, suggests one central deficit underlying the general performance patterns. 130 5.4.3 CHAPTER 5. SENTENCE PRODUCTION DOP-H: an overarching processing theory As mentioned in the introduction, DOP-H is an overarching theory predicting more problems for agrammatic speakers on both production and comprehension of sentences whose arguments are in derived position. It has been supported by not only production studies in several languages (Bastiaanse et al., 2002a; Bastiaanse and Thompson, 2003; Burchert et al., 2008; and Yarbay Duman et al., 2007; 2008) but also several comprehension studies (Bastiaanse & Van Zonneveld, 2006); Yarbay Duman et al., 2011). The current data on production as well as the previous finding on comprehension study of bilingual agrammatic speakers (Abuom et al., 2013) further support this theory; and provide new insight into sentence production deficit of bilingual individuals with agrammatic Broca’s aphasia. However, DOP-H does not cover problems with embedding also found in the current study. CHAPTER 6 General Discussion and Conclusion This chapter presents a general discussion and conclusion of the previous chapters of the thesis. In recap, this thesis project investigated the patterns and severity of verb inflection and word-order production and comprehension deficits in Swahili-English bilingual agrammatic speakers in light of several theories of agrammatism. Through a set of behavioral off-line testing techniques -a spontaneous speech elicitation task, a sentence completion task, a sentence-picture matching task, and a sentence production elicitation task– the four studies presented in the previous chapters examined how some of the theories, widely tested among monolinguals, apply to bilingual agrammatic speakers. Furthermore, the universality of these theories in relation to Swahili, a highly agglutinative Bantu language with a complex verb inflection paradigm when compared to most languages studied so far, has also been determined. The results demonstrated significant effects of time reference, word order and embedding (wh-movement) on the performance patterns of bilingual agrammatic speakers across their two languages. Swahili-English bilingual agrammatic speakers were compromised in their verb use in spontaneous speech, and in their production and comprehension of time reference to the past through verb inflection and derived word orders. Although some of these phenomena 131 132 CHAPTER 6. DISCUSSION AND CONCLUSION have been mentioned before, there are several novel findings: (1) these studies have been performed in Swahili, a language in which aphasia has not been studied before; (2) systematic group studies on agrammatic comprehension and production have not been done before in a bilingual population; (3) most interesting, some of the agrammatic speakers participated in both the time reference and the word order and embedding studies on comprehension and production. So far, the problems with time reference, word order and embedding have not been studied in the same individuals before. In this section: first, we highlight the major findings and address major issues arising from the studies in relation to the literature presented in the Introduction. Second, we discuss some of the clinical implications of the study. Finally, we provide some recommendations for future research. 6.1 Major research findings There are five major findings from the studies presented in this thesis: (1) the agrammatic spontaneous speech in Swahili is not fundamentally different from agrammatic speech in English; (2) time reference to the past through verb inflection is impaired in production and comprehension in both languages as predicted by the PADILIH, but more seriously in English than in Swahili; (3) both production and comprehension of sentences with derived word orders are impaired, as predicted by the DOP-H; (4) the comprehension and production is influenced by embedding; (5) the agrammatic performance in both languages is very similar. The individual research questions as formulated in the final section of the Introduction have already been addressed in the separate research chapters (2-5). In this discussion, some major issues that were presented in the Introduction or which arose in the studies will be addressed. 6.1.1 Language dependent variables in spontaneous speech As shown in the Introduction and other sections of the thesis, there are remarkable structural (morphological) differences between Swahili and English. Swahili, being an agglitunative language, conveys more information in fewer words than in English. Initial analysis of NBDs samples showed that information conveyed in 3 words in English is generally conveyed in 2 words or less in Swahili. It is, therefore, logical that analysis based on equally sized lan- 6.1. MAJOR RESEARCH FINDINGS 133 guage samples was not appropriate for cross-linguistic comparisons. Brodshire and Nicholas (1994) point out that a minimum of 300 words is a reliable sample size for analysis for languages such as English, but that in agglutinative languages, fewer words can be used, especially when a comparison with spontaneous speech in a non-agglutinative language has to be made. Based on this argument we analysed 300 words for English and 200 words for Swahili. Our comparisons of Swahili and English agrammatic speech samples did not reveal any fundamental differences between the two languages, especially in variables that were not expected to yield any differences. Of course, the utterances of the Swahili samples, both of the agrammatic and the non-brain-damaged speakers, were about one third shorter than in the English samples. In both languages, a considerable reduction in utterance length, a delay in speech rate, use of verbs and nouns (whose ratio is, surprisingly, the same as that of controls), production of more ungrammatical sentences and production of fewer embedded sentences was observed when the agrammatic samples were compared to those of non-brain-damaged speakers. Interestingly, a comparison of Kenyan English speech samples with those of American English, which we did to determine whether analysis of spontaneous speech in languages for which aphasia tests are not available is sufficient to identify agrammatism, also showed similar manifestation of agrammatism in the two groups (monolinguals and bilinguals), irrespective of the number of languages mastered pre-morbidly. This implies that it is not always the case that agrammatism manifests differently in languages that differ in terms of their grammatical morphology, as argued by Paradis (1988), but it shows that the general characteristics of agrammatism are universal with only minor differences in specific junctures where the two languages differ (Fabbro, 2001), at least in aphasic individuals who are agrammatic in both languages. In our data, minor differences were found upon further analyses of the use of verb inflections in both languages. Grodzinsky (1991) and Paradis (1995) predict that agrammatic individuals will omit grammatical inflections in languages that allow bare forms, but will substitute them in languages where omissions result in non-words. Indeed, there were some differences in patterns of omissions and substitutions of grammatical morphemes not only in support of Fabbro (2001), but also of Grodzinsky (1991) and Paradis (1995). In Swahili, the agrammatic individuals substituted the finite verb forms with non-finite verb forms by using forms with ku- (infinitive marker) and ka- (consecutive marker) more often 134 CHAPTER 6. DISCUSSION AND CONCLUSION than normal. For English, the agrammatic speakers overused the infinitive (by omission of the tense inflection) and the gerund forms of lexical verbs. However, in both languages the agrammatic speakers had a clear preference for verb forms (both tense and aspect) referring to the present rather than to the past. We interpreted the use of fewer verb forms that refer to the past as pointing to an underlying selective deficit for verb forms referring to the past as predicted by the PADILIH. 6.1.2 Representation vs. processing In the Introduction, several linguistic theories of agrammatism, both processing and representational accounts, were highlighted and their relevance to each of the studies was discussed in detail. The TPH (Friedmann & Grodzinsky, 1997) is a representational account; the TUH (Wenzlaff & Clahsen, 2005) and TAUH (Burchert et al., 2005) are both processing accounts. But all three predict agrammatic speakers’ production difficulties with tense in general. The PADILIH (Bastiaanse et al., 2011), a processing account, finds this prediction to be both too narrow and too broad. According to the PADILIH, it is narrow in the sense that the deficit is not restricted to tensed verb forms but also holds for other verb forms expressing time reference through aspect inflection and through periphrastic verb forms, and it is broad because not all tenses are equally affected: present and future tenses are relatively spared. Our data from the spontaneous speech analysis (Chapter 2) and time reference study (Chapter 3) showed an impaired production and comprehension of verb forms for time reference to the past in both languages. This further supports the PADILIH, expanding it to include bilingual agrammatic speakers’ data as well. The PADILIH, in addressing the question of why reference to the past is impaired, bases its argument on Avrutin’s discourse linking theory, also a processing theory. Avrutin (2006) attributes the agrammatic speakers’ difficulty with tense to discourse linking, which he argues is impaired in agrammatic aphasia. In distinguishing two levels of syntax processing, sentence level versus discourse level, Avrutin argues that because of brain damage, agrammatic speakers do not have sufficient processing resources required for syntactic processing at the discourse level. He does not mention that one specific tense is affected. Zagona (2003) clarifies that only past tense is affected since it requires discourse linking, whereas both present and future tenses do not need such linking. She argues that when referring to the past, speech time and event time do 6.1. MAJOR RESEARCH FINDINGS 135 not coincide, hence, a discourse linked relation has to be made between speech time and an earlier event. Bastiaanse and colleagues (2011), in formulating PADILIH, extended Zagona’s theory on discourse linking to include reference to the past through grammatical morphology in general. Another set of theories, the TDH (Grodzinsky, 1986), a representational account, and the DOP-H (Bastiaanse & Van Zonneveld, 2005), a processing account, focus mainly on the order of constituents in a sentence. Whereas the TDH predicts agrammatic speakers’ problems only with comprehension of sentences whose arguments are in non-canonical order, the DOP-H makes similar predictions for both production and comprehension. Our data on sentence comprehension (Chapter 4) supported both theories, while the data on sentence production (Chapter 5) supported the DOP-H. Although these theories have initially been developed on the basis of the performance of monolingual agrammatic speakers, their predictions can now be extended to include bilingual agrammatic speakers in both their languages. Interestingly, these studies on the effect of word order included base and derived order sentences with and without embedding. Although it is wellknown that agrammatic speakers are very limited in their use of embedded sentences (e.g. Bastiaanse et al., 2002), it has not been shown before that embedding is impaired in both production and comprehension and that there is an interaction between embedding and word order in agrammatic aphasia. Apparently, time reference to the past, derived order of arguments and embedding compromise both production and comprehension of agrammatic speakers, irrespective of language or number of languages pre-morbidly mastered. These data do not fully support representational deficit accounts, which attribute the agrammatic speakers’ problems with speech production and comprehension to a complete lack (loss) of linguistic representation. The fact that the agrammatic participants were able to produce and comprehend simple sentences in present tense and in base word order perfectly well, as well as a few sentences in past tense and derived word orders, shows an intact grammatical representation. Processing deficit accounts can fully account for these data. The agrammatic speakers’ abilities to process discourse linked elements (reference to the past) as well as complex structures (derived word orders and embedded sentences) were seriously affected, but not completely absent. 136 6.1.3 CHAPTER 6. DISCUSSION AND CONCLUSION Bilingual aphasia and the bilingual brain As mentioned in the current studies, the agrammatic participants were balanced bilinguals who were premorbidly equally proficient in Swahili and English, having acquired both languages at an early age (4 years old) and another (native) language from birth. Although the two languages in question are morphologically different (rich vs. poor morphology), the performance of agrammatic individuals across the two languages in all the studies is either similar or slightly better in Swahili. The similarity in performance across the two languages, irrespective of the differences in grammatical morphology, implies that the agrammatic individuals suffered from the same type of aphasia in both languages. They were not only agrammatic in both languages, but the degree of their agrammatism was equally severe as shown by the high correlations between performance in Swahili and English, and between production and comprehension tasks. Therefore, it is unlikely that the two grammars, no matter how different they may be, are represented in different areas of the brain (or separate neuroanatomical representations for the two languages, as suggested by Albert and Obler, 1978). It has, however, been assumed that languages that are acquired early are generally represented in shared processing regions of the brain in early balanced and proficient bilinguals (Miozzo et al., 2010; Abutalebi et al., 2005). Therefore, what these results suggest is that the syntactic representations of the two languages may not have been affected due to brain damage (no patient scored a zero on all tests), but that processing is impaired; that is, complex syntactic operations are hard to handle but apparently these operations take place in one and the same area for both languages. This area is probably in the vicinity of Broca’s area because this area is known to be crucial for linguistic processing and is often damaged or not functioning properly in agrammatic Broca’s aphasia. Apparently, the damage to Broca’s area affects both languages equally. 6.2 The clinical implications of the results of this study These research findings have implications for the diagnosis and treatment of Swahili-English bilinguals with agrammatism. One of the key findings is that both languages are similarly affected, irrespective of their morphological differ- 6.2. CLINICAL IMPLICATIONS 137 ences. The question, therefore, is whether it is sufficient to test and treat only one of the languages or all languages known by the patient. This has been a great challenge for speech therapists in Kenya who are mostly foreigners who can only conduct therapy sessions in English. Paradis (1995) argues that it is not ‘ethically accepatable’ for speech therapists to assess bilingual patients by examining only one of their languages. This argument has been supported by Fabbro (1990) who adds that it is wrong for clinicians to ignore the principles of bilingual aphasia when it comes to the diagnosis of and therapy for language disorders. The current study, however, shows that this is not necessarily the case. All of our patients were examined and treated in English, but their Swahili was equally, if not less, impaired. It is conceivable that treatment in English enhanced performance in Swahili, but as records of treatment were not kept and no pre- and post-treatment testing systematically undertaken, one cannot know. Cross-language generalization of treatment has been reported before. Edmonds and Kiran (2006) found cross-language generalization after lexical training when the non-treated language was the speaker’s more dominant language or when the participants were highly proficient in both their languages. Similarly, Goral, Levy and Kastl (2010), who treated a patient at the grammatical level in English (L2), found improvement in speech rate and morphosyntactic skills in the treated language, English (L2), as well as in non-treated French (L3). It is, therefore, evident that not only the treated language but also the non-treated language can benefit from therapy. Thus, it is important for the clinician to decide which language to treat depending on the level of language dominancy or proficiency prior to brain damage. Other specific findings are that the patients in the current study have difficulties producing and comprehending sentences referring to a past time-frame, as well as those whose order of arguments are derived (in both Swahili and English). It is, therefore, recommendable for therapy purposes that bilingual individuals with agrammatism are explicitly instructed using simple sentence structures in order to improve their communicative abilities. With respect to treatment, some studies suggest that the type of syntactic movement, whether NP- or Wh- movement, is an important factor to consider for aphasic individuals with sentence deficits. For instance, Thompson and colleagues provide evidence from a series of studies that show that training sentences with Whmovement, such as object relative clauses, does not influence production or comprehension of sentences with NP-movement, such as passives, though gen- 138 CHAPTER 6. DISCUSSION AND CONCLUSION eralization across sentences with similar movement operations has often been observed (e.g., from object relative structures to object Wh-question forms) (Dickey & Thompson, 2007; Thompson, Choy, Holland & Cole, 2010; Thompson & Shapiro, 2005; Thompson, Shapiro, Kiran & Sobecks, 2003). The tests used in word order experiments are associated with both types of syntactic operations: NP-movement (i.e., passives) and Wh-movement (i.e., object clefts and object relative clauses) in both production and comprehension tasks. In conclusion, like in the case of monolinguals, we can safely assume that verb and word order deficits are prevalent in bilingual individuals with aphasia. Tests for these deficits are needed in order to understand the extent of patients’ language impairment and for treatment planning. However, there are no relevant and appropriate test batteries for bilingual aphasia that can address the aforementioned deficits. The tests used in several chapters of this study provide a comprehensive assessment tool that can conveniently be used by both clinicians and speech therapists for purposes of screening agrammatic speech and therapy interventions. 6.3 Scope for further study/research In future research, it would be interesting to examine Swahili-English bilingual speakers with fluent aphasia, such as Wernicke’s aphasia and anomia. Furthermore, it would be interesting to test some of the native languages of Kenyan bilingual aphasic speakers, to see whether the patterns of impairments in Swahili and English are reflected in their native languages. Finally, while it was beyond the scope of this study to show if there is a relation between time reference and word order deficits, future research may aim at a common underlying deficit. Summary This thesis focused on the manifestation of agrammatism in Swahili-English bilingual speakers with aphasia, by investigating verb inflection and word order deficits in light of several theories of agrammatism. Whether, and to what extent, some of the linguistic theories of agrammatism apply to bilingual agrammatic speakers of Swahili and English has been extensively addressed in the four study chapters of this thesis. The first study, presented in Chapter 2, aimed to identify the features of Swahili agrammatic narrative and spontaneous speech, and to compare the use of verb inflections for tense and time reference (in Swahili and English) of the agrammatic and non-brain-damaged individuals by addressing three main research questions. The first question was to determine whether there are similarities or differences in the English speech production of monolingual American and bilingual Kenyan aphasic speakers. The reason for this question was to establish whether the analysis of the spontaneous speech of aphasic speakers in languages for which there are no standardized tests (for instance, languages spoken in Kenya) is sufficient to identify agrammatism. The results showed a remarkable similarity between agrammatic phenomena by monolingual American and bilingual Kenyan speakers on typical agrammatic features: utterance length and speech rate are reduced, and the proportion of grammatical sentences and complex sentences is lower than normal. The second question sought to establish the features of Swahili agrammatism and describe how comparable they are to features of English agrammatism. Having established that the aphasic participants were agrammatic in English, additional analyses were done in Swahili and comparisons were made with the English samples. As expected, the Swahili agrammatic samples, when compared to those of non- 139 140 SUMMARY brain-damaged individuals showed features that generally quantify agrammatism: reduced speech output; low speech rate; shorter utterances; limited use of embeddings in sentences; and frequent production of ungrammatical sentences. Furthermore, a comparison of Swahili with English samples showed that these agrammatic symptoms were very similar in both languages. The final question focused on whether the production of verb forms referring to a past time frame was impaired in both languages. Time reference has been tested in several languages, but has not been examined in agrammatic spontaneous speech. This question was precipitated by the need to establish whether the predictions of the Past Discourse Linking Hypothesis (PADILIH) extend to both languages of bilinguals, and more specifically to other agglutinative languages such as Swahili. An analysis of the use of verb forms, taking into account both tense and aspect, revealed a preference for verb forms referring to the present rather than to the past by agrammatic (but not by non-brain-damaged) speakers in both languages, suggesting the possibility of an underlying deficit for reference to the past through verb inflection. The findings on the use of verb inflections for time reference provided the basis for the subsequent study. With respect to the second study, presented in Chapter 3, the Test for Assessing Reference of Time (TART), designed to provide a comprehensive assessment of production and comprehension of reference of time, was used to address three research questions. The three questions were as follows: 1) Is production of time reference to the past through verb inflection impaired in both languages of a bilingual? 2) Is comprehension of time reference to the past through verb inflection impaired in both languages of a bilingual? 3) Is reference to the past similarly impaired in both Swahili and English for bilingual speakers? By exploring these questions, we aimed to establish whether the notable differences in the size of the verb inflection paradigm for time reference in both languages can influence the performance of agrammatic speakers. The relatively large number of affixes in the Swahili verb paradigm makes it more complex, although the inflection for time reference is more regular compared to that of English. As predicted, the findings concurred with the predictions of the PADILIH. The agrammatic speakers were impaired in their production and comprehension of time reference to past, but relatively spared in their time reference to both the present and the future. Furthermore, a similar pattern of impairment was manifested in both languages irrespective of the morphological differences, as noted in the previous study. Our explanation of the problem SUMMARY 141 with time reference to the past is based on Zagona’s (2003) assumption that reference to the past is discourse linked, whereas both the present and future are non-past (the future being a variant of the present), hence not discourse linked. Discourse linked elements have to be processed by discourse syntax, according to Avrutin (2000; 2006). This requires extra processing resources which are not sufficiently available for agrammatic speakers due to their brain damage. For the third study, Chapter 4, we tested bilingual agrammatic speakers’ ability to comprehend sentences whose order of arguments is derived. Two theories, the Trace Deletion Hypothesis (TDH) and the Derived Order Problem Hypothesis (DOP-H), formed the basis for this study, both predicting difficulties with derived order sentences for agrammatic speakers. The TDH, a representational account, predicts above chance level performance for sentences in which arguments are in base order (such as actives and subject relatives), and chance level performance when arguments are in a derived order (i.e. non-canonical sentences such as passives and object relatives with and without relative pronouns). The DOP-H, a processing account, also predicts good performance for sentences whose arguments are in base order, and poor performance for sentences with arguments in a derived order. However, since the DOP-H is a processing account, interpretation of the results is not in terms of a comparison to chance level but rather in terms of (significantly) worse performance on sentences with derived order of the arguments. Although these theories have been widely tested among monolingual agrammatic speakers, no previous study has reported on bilinguals; hence the need to examine the patterns of comprehension deficits that are likely to emerge in the structurally different languages of bilinguals. The results revealed that sentences in a derived order are more difficult to comprehend than those in base order in both Swahili and English, corroborating the predictions of both theories to include bilingual agrammatic speakers. Furthermore, there was a significant interaction effect between derived order of arguments and embedding: sentences with an embedding (in this case, the relative clauses) were more difficult to comprehend than simple declaratives. The DOP-H can account for the further difficulty caused by embedding by suggesting that embedding adds linguistic complexity to sentences, requiring an extra operation to parse. The final study, Chapter 5, examined the production of sentences in both 142 SUMMARY base and derived word orders in Swahili and English. Sentences in which the arguments are in base order (actives and subject clefts) and in derived orders (passives and object clefts) were elicited, based on the predictions of DOP-H– an overarching theory that makes similar predictions for both comprehension and production. As predicted by the DOP-H, the results replicated the findings from the previous study on comprehension: sentences in derived order (the passives and object clefts) were more difficult to produce than those in base order (the actives and subject clefts). Again, there was an interaction effect with embedding, making the subject clefts more difficult than the simple actives and the objects cleft more difficult than the passives. In sum, these studies have shown that bilingual Swahili-English agrammatic speakers are not clinically very different from monolingual agrammatic speakers in terms of how their language production and comprehension are affected following brain damage. Their spontaneous speech in both languages is characterized by significantly (1) lower speech rate; (2) shorter utterances; (3) fewer grammatical sentences; and (4) fewer embeddings. Furthermore, their production and comprehension of sentences is seriously affected by time reference through verb inflection, word order and embedding. Sentence structures with verb inflections for time reference to the past, those whose arguments are in a derived (non-canonical) order, as well as those with embeddings, are poorly comprehended and produced. The pattern of impairment is similar in both languages (Swahili and English) irrespective of their differences in morphological structure, suggesting language representation in shared processing regions of the bilingual brain. Nederlandse Samenvatting Dit proefschrift richtte zich op hoe agrammatisme zich in tweetalige sprekers van Swahili en Engels manifesteert door stoornissen in werkwoordsinflectie en in woordvolgorde te onderzoeken aan de hand van verschillende theorien over agrammatisme. Of, en in welke mate, sommige van die taalkundige theorien over agrammatisme betrekking hebben op tweetalige agrammatische sprekers van Swahili en Engels komt uitgebreid aan bod in de vier hoofdstukken van dit proefschrift. De eerste studie, hoofdstuk 2, was gericht op het identificeren van de kenmerken van narratieve en spontane taal in agrammatisch Swahili, en het vergelijken van het gebruik van werkwoordsinflectie voor tempus en tijdsverwijzing (in Swahili en Engels) door agrammatisch afatische individuen en individuen zonder hersenletsel door middel van drie hoofdonderzoeksvragen. De eerste vraag erop gericht om te bepalen of er overeenkomsten of verschillen waren in taalproductie van het Engels van eentalige Amerikaanse, en tweetalige Keniaanse afatische sprekers. Het doel van deze vraag was het vaststellen of spontanetaalanalyse van afatische sprekers van talen waarvoor geen gestandaardiseerde tests beschikbaar zijn (bijvoorbeeld voor talen die in Kenia worden gesproken) genoeg is om agrammatisme te identificeren. De resultaten toonden een opmerkelijke overeenkomsten aan tussen agrammatische fenomenen in beide talen (Keniaans-Engels en Amerikaans-Engels) op typische agrammatische kenmerken: uitingslengte is korter, spreektempo is lager, en de proportie grammaticale zinnen en complexe zinnen is lager dan normaal. De tweede vraag diende ertoe om vast te stellen wat de kenmerken van agrammatisme in Swahili zijn en hoe vergelijkbaar ze zijn met de kenmerken van agrammatisme in Engels. Na te hebben vastgesteld dat de afatische individuen agrammatisch waren in het En- 143 144 NEDERLANDSE SAMENVATTING gels, werden aanvullende analyses uitgevoerd in Swahili en werden vergelijkingen gemaakt met de Engelse samples. Zoals verwacht vertoonden de agrammatische fragmenten van Swahili in vergelijking met de fragmenten van individuen zonder hersenletsel kenmerken die over het algemeen agrammatisme karakteriseren: een verminderde taalproductie, een laag spreektempo, kortere uitingen, beperkt gebruik van ingebedde zinnen, en naar verhouding veel agrammaticale uitingen. Verder liet een vergelijking van fragmenten in Swahili en in Engels zien dat deze agrammatische symptomen zeer vergelijkbaar waren in beide talen. De laatste vraag focuste op of de productie van werkwoordsvormen die naar het verleden verwijzen aangetast was in beide talen. Tijdsverwijzing is in veel verschillende talen getest, maar is nog niet in agrammatische spontane taal onderzocht. Deze onderzoeksvraag is voortgekomen uit de noodzaak om vast te stellen of de voorspellingen van de Past Discourse Linking Hypothesis (PADILIH) ook betrekking hebben op beide talen van tweetalige mensen, en meer specifiek op andere agglutinerende talen zoals Swahili. Een analyse van het gebruik van werkwoordsvormen die zowel met tempus als aspect rekening houdt onthulde een voorkeur voor werkwoordsvormen die naar het heden verwijzen in plaats van naar het verleden voor agrammatische sprekers (maar niet voor sprekers zonder hersenletsel) in beide talen. Dit suggereert de mogelijkheid van een onderliggende stoornis voor tijdsverwijzing naar het verleden door middel van werkwoordsinflectie. De bevindingen op het gebied van werkwoordsinflectie voor tijdsverwijzing vormde de basis voor de volgende studie. Met betrekking tot de tweede studie, te vinden in hoofdstuk 3, werd de Test for Assessing Reference of Time: ‘Test voor het Onderzoeken van Tijdsverwijzing’ (TART) gebruikt, die ontworpen is om de productie en het begrip van tijdsverwijzing te beoordelen om daarmee de onderzoeksvragen aan de orde te stellen. De drie vragen waren als volgt: 1) Is productie van tijdsverwijzing naar het verleden door werkwoordsinflectie aangetast in de beide talen van een tweetalige? 2) Is het begrip van tijdsverwijzing door werkwoordsinflectie aangetast in de beide talen van een tweetalige? 3) Is verwijzing naar het verleden vergelijkbaar aangetast in zowel Swahili als Engels? Door deze vragen aan te kaarten probeerden we vast te stellen of de opvallende verschillen in de grootte van het werkwoordsinflectieparadigma voor tijdsverwijzing in beide talen de prestatie van agrammatische sprekers kan benvloeden. Het relatief grote aantal affixen van het werkwoordsinflectieparadigma van het Swahili maakt het een complexer paradigma, alhoewel de inflectie voor tijdsverwijzing regelmatiger is dan NEDERLANDSE SAMENVATTING 145 die van het Engels. Zoals voorspeld stemden de bevindingen overeen met de voorspellingen van de PADILIH. De productie en het begrip van tijdsverwijzing naar het verleden van agrammatische sprekers was aangetast, maar de tijdsverwijzing naar het heden en de toekomst was relatief gespaard. Verder kwam het aantastingspatroon op dezelfde manier naar boven in beide talen, onafhankelijk van de morfologische verschillen zoals in de voorgaande studie aangegeven. Onze verklaring voor het probleem met tijdsverwijzing naar het verleden is gebaseerd op Zagona’s (2003) aanname dat verwijzing naar het verleden een koppeling in de discourse (vertoog) vereist, terwijl zowel heden als toekomst niet-verleden zijn (met de toekomst als variant van het heden) en daarom niet discourse-gekoppeld. Discourse-gekoppelde elementen moeten verwerkt worden door discourse-syntaxis volgens Avrutin (2000, 2006). Dit vereist extra verwerkingscapaciteit, waarvan niet voldoende aanwezig is bij agrammatische sprekers vanwege hun hersenletsel. Voor de derde studie, hoofdstuk 4, hebben we het vermogen van tweetalige agrammatische sprekers getest om zinnen met een afgeleide argumentsvolgorde te begrijpen. Twee theorien, de Trace Deletion Hypothesis (TDH) en de Derived Order Problem Hypothesis (DOP-H) vormden de basis van deze studie. Beide voorspellen moeilijkheden met zinnen met afgeleide woordvolgorde voor agrammatische sprekers. De TDH, een representatieverklaring, voorspelt prestatie boven kansniveau voor zinnen waarin de argumenten in hun basisvolgorde staan (zoals actieve zinnen en subjectsrelatieve bijzinnen), en prestatie op kansniveau voor het omgekeerde (d.w.z. niet-canonieke zinnen zoals passieve zinnen en objectsrelatieve bijzinnen met of zonder betrekkelijk voornaamwoord). De DOP-H, een verwerkingsverklaring, voorspelt verder een goede prestatie op zinnen waarvan de argumenten in de basisvolgorde staan, en een slechte prestatie op zinnen waarvan de argumenten in afgeleide volgorde staan. Toch, omdat de DOP-H een verwerkingsverklaring is, verloopt de interpretatie van de resultaten anders dan hierboven, beneden kansniveau, namelijk eerder met betrekking tot (significant) slechtere prestatie op zinnen met een afgeleide volgorde van de argumenten. Alhoewel deze theorien uitgebreid getest zijn met eentalige agrammatische sprekers, was er tot nu toe nog geen studie gericht op tweetalige mensen. Derhalve bestaat de behoefte om de patronen van de begripstoornis te onderzoeken die voorkomen in de structureel andere talen van tweetalige mensen. De resultaten tonen aan dat zinnen met afgeleide volgorde van argumenten moeilijker te begrijpen zijn dan die met 146 NEDERLANDSE SAMENVATTING basisvolgorde in zowel Swahili als Engels, wat bevestigt dat de voorspellingen van beide theorien ook betrekking hebben op tweetalige mensen. Daarbij was er een significante interactie tussen afgeleide argumentsvolgorde en inbeddingen: ingebedde bijzinnen (in dit geval betrekkelijke bijzinnen) waren moeilijker te begrijpen dan hoofdzinnen (zonder inbedding). De DOP-H kan de hogere moeilijkheidsgraad veroorzaakt door inbedding verklaren door te stellen dat inbedding een extra lingustische complexiteit toevoegt aan zinnen die tot een extra verwerkingsoperatie leidt. De laatste studie, hoofdstuk 5, onderzocht de productie van zinnen in zowel basisvolgorde als afgeleide volgorde in Swahili en Engels. Zinnen waarvan de argumenten in basisvolgorde staan (actieve zinnen en subjectskloofzinnen) en in afgeleide volgorde (passieve zinnen en objectkloofzinnen werden ontlokt op basis van de voorspellingen van de DOP-H; een overkoepelende theorie die zowel voor begrip als productie vergelijkbare voorspellingen doet. Zoals door de DOP-H voorspeld werd, repliceerden de resultaten de bevindingen van eerdere begripsstudies: zinnen met afgeleide volgorde (passieve zinnen en objectkloofzinnen) waren moeilijker te produceren dan zinnen met basisvolgorde (actieve zinnen en subjectkloofzinnen). Wederom was er een interactie met inbedding, waarbij de subjectkloofzinnen moeilijker bleken te zijn dan de enkel actieve zinnen en de objectkloofzinnen moeilijker bleken dan de passieve zinnen. Samenvattend hebben deze studies aangetoond dat tweetalige sprekers van Swahili en Engels klinisch gezien niet sterk verschillen van eentalige sprekers wat betreft hun stoornis op het gebied van taalproductie en -begrip ten gevolge van hersenletsel. Hun spontane taal in beide talen wordt gekarakteriseerd door (1) een lager spreektempo, (2) kortere uitingen, (3) minder grammaticaal correcte uitingen en (4) minder inbeddingen. Verder is hun productie en begrip ernstig aangetast door tijdsverwijzing door middel van werkwoordsinflectie, door woordvolgorde en door inbedding. Zinstructuren met werkwoordsinflectie voor tijdsverwijzing naar het verleden, met een afgeleide (niet-canonieke) volgorde van de argumenten, en met inbeddingen worden slecht begrepen en geproduceerd. Het patroon van de stoornis is vergelijkbaar in beide talen (Swahili en Engels) onafhankelijk van hun structurele (morfologische) verschillen, wat taalrepresentatie in gedeelde verwerkingsgebieden in het tweetalige brein suggereert. Appendices 147 148 APPENDICES Appendix to Chapter 2 2.A Swahili speech samples comprising the first 10 utterances of agrammatic participants in Chapter 2 EA Mzee okoa msichana bahari... Man save girl sea “The man save the girl” mshipi vuta msichana... belt pull girl “The belt pull the girl” msichana oga bahari... girl bath sea “The girl bath sea” hewa... “air” mafuriko tele... “flood abundant” mwokozi kuja haraka... saviour come quickly “Saviour come quickly” a - ka - okoa msichana... he - CONSEC. - Save girl “He save the girl” mzee beba msichana inje bahari... man carry girl outside sea “The man carry the girl outside the sea” Msichana... Girl... “The girl...” mvulana panda kiti... boy climb seat “The boy climb the seat” 6.3. FURTHER RESEARCH 149 SW Kuna msichana mmoja akona maji... there is girl one has water “There is one girl who has water” na ha-na hiyo... and NEG-with that “and she does not have that...” kuna maji nyingi sana... there is water a lot very “There is a lot of water” sasa kuna mtu mmoja a-na-taka ku-m-saidia... now there is person one s/he-PRESENT-want INFIN-him/her-help “Now, there is one person who wants to help her” kwa sababu hiyo maji iko nyingi sana... of because that water is a lot very “Because that water is a lot” Sasa a-na-jaribu njia venye a-ta-fanya... now s/he-PRESENT-try way how s/he-FUTURE-do “Now, he is trying to find a way out” sababu iko na miti karibu moja... because there is with trees nearby one... “because there is a tree nearby” na hiyo maji ni nyingi sana... and that water is a lot very “And the water is a lot” I-na-anza kuwa nyingi sana... It-PRESENT-start to be a lot very “The water level is rising” sasa now huyo that “The huyo mtu a-na-taka ku-m-saidia that person s/he-PRESENT-want INFINITIVE-her/him-help msichana... girl person wants to help the girl now” 150 APPENDICES HJ Mafuriko i-li-taka ku-beba huyu msichana... floods it-PAST-want INFINITIVE-carry this girl “The floods wanted to carry this girl” halafu huyu a-li-ona yeye... then this s/he-PAST-see her/him “This one saw her” halafu a-ka-okoa huyu... then s/he-CONSEC-save this “He saved this one” Na huyu a-li-po-piga mayowe... and this s/he-PAST-when-make noise “and when this one made noise” halafu huyu a-ka-m-sikia... then this s/he-CONSEC-her/him-hear “Then this one heard her” halafu a-ka-rudi ku-m-toa... then s/he-CONSEC-return INFINITIVE-her/him-remove “He came back to get her out” halafu a-ka-m-lalisha yeye kwa sakafu... then s/he-CONSEC-her/him-lay her/him on floor “Then he lay her on the floor” halafu a-ka-finyilia tumbo yake... then s/he-CONSEC-press stomach hers “Then he pressed her stomach” halafu a-ka-tapika maji... then s/he-CONSEC-vomit water “Then she vomited water” Huyu msichana a-li-po-kunywa maji... this girl s/he-PAST-when-drink water “When this girl drank water” 6.3. FURTHER RESEARCH 151 MM kuna msichana kwa maji There is girl in water “There is a girl in the water” na-fikiri ni mafuriko PRESENT-think is floods “I think it is floods” from water na kijana a-me-kuja ku-m-uokoa and boy s/he-PRESENT-come INFINITIVE-her/him-save msichana kwa maji... girl “The boy has come to save the girl from the water” Watu wa ku-toka kwa .... people of INFINITIVE-come from “People from....” wa-na-enda ku-okoa watu wa shida... they-PRESENT-go INFINITIVE-save people of problems “They are going to save people with problems” kama hiyo mafuriko... like that flood “Like the floods” na sana sana kama sasa wako na mafuriko kwa mara and very very like now they are with floods for times mingi... many “and especially right now when they have lots of floods” lakini mwisho a-me-pata usaidizi... but last s/he-PERFECT-get help “But at last, she has got help” msichana a-me-pata watu wa familia yake... girl s/he-PERFECT-get people of family hers “The girl has found people from her family” a-me-furahi... s/he-PERFECT-happy “She is happy” 152 APPENDICES JK Hapa u-na-weza ku-ona msichana... Here you-PRESENT-can INFINITIVE-see girl “Here you can see a girl” ndiye a-li-kuwa a-na-zama maji... is the (one) s/he-PAST-be s/he-PRESENT-drown water “She is the one who was drowning in the water”. na iko mtu a-na-kuja ku-m-uokoa... and there is someone s/he-PRESENT-come INFINITIVE-her/him-save “And there is someone coming to save her” na ako na shida a-me-ji-shikilia miti... and has with problem s/he-PERFECT-it-hold trees “And she has a problem she has held a stick” na a-me-beba kamba ya ku-okoa msichana and s/he-PERFECT-carry rope for INFINITIVE-save girl “And he has carried a rope to save the girl” hii ni picha ya msichana a-na-okole-wa... this is picture of girl s/he-PRESENT-save-PASSIVE “This is a picture of a girl who is being saved” a-na-okole-wa kuto-ku-zama kwa maji... s/he-PRESENT-save-PASSIVE NEG-INFINITIVE-drown in water “She is being saved not to drown in water” na jamaa wa... yuko na vitu... and person from... has with things “And the person from....has things” a-na-jaribu ku-m-toa kutoka kwa hiyo s/he-PRESENT-try INFINITIVE-her/her-remove out of that maji... water “He is trying to get her out from the water” na i-na-onekana ana shida ya ku-taka and it-PRESENT-seem has problem of INFINITIVE-want ku-zama... INFIN-drown “And it seems she has drowning problem” 6.3. FURTHER RESEARCH 153 LA Kwenye hii picha mtu a-na-kuja kwa maji... in this picture someone s/he-PRESENT-come into water “In this picture, someone is coming into the water” na kijiti a-na-m-patia huyu... and stick s/he-PRESENT-give this one “And he is giving the stick to this one” a-na-m-uliza a-shikilie... s/he-PRESENT-her/him-ask s/he-hold “He is asking her to hold it” huyu mwanamme a-na-m-patia kijiti... this man s/he-PRESENT-her/him-give stick “This man is giving her a stick” a-weze ku-m-vuta ndani... s/he-can INFINITIVE-her/her-pull inside “For him to be able to pull her from the water” lakini huyu mama a-na-taka ku-shikilia... but this woman s/he-PRESENT-want to-hold “But this woman wants to hold it” lakini ha-wezi... but NEG-able “But not able to...” kwa hivyo a-na-onekana ni kama a-ta-*drown... because of that s/he-PRESENT-seem is like s/he-FUTURE-drown “It seems she will drown” hii maji iko kali sana... this water is strong very “This water is very strong” woman kwa hivyo huyu mwanamme a-na-taka and so this man s/he-PRESENT-want ku-saidia huyu mama INFINITIVE-help this “And so this man wants to help this woman” 154 2.B APPENDICES English Speech samples comprising the first 10 utterances of agrammatic participants in Chapter 2 EA Floods rescue.... man rescue her from river or ocean Girl... rain... me know it... girl...is...was swimming ... after swimming around the lake... after floods came with waves... signal rescue... the belt hold her back.. now swim out water. HJ What do we call the rescuers.. This was wanting to help this girl... and then she went ... and rescue the girl... she wanted to rescue the girl... the girl was help... the rescuers... they just went ... and rescued this girl... and she made it... MM This girl is drowning... and I think a lot of rain... and they are being swept by the water... and this man is a... maybe a.... what are these people... maybe the Red Cross people... and they would like to rescue the girl... in the end the girl will be rescue... Once upon a time this girl lived in the Rift-valley... 6.3. FURTHER RESEARCH JK I can see a girl here who is... who looks like she is agonized ... going through some kind of trouble... I can see this guy is helping... I don’t know... basically she’s holding a piece of wood... what is this?... the belt... he is a belt with him... so far he is not caught the girl... LA The girl drowning... the man trying to help her... this man was walking along the river but he saw girl drowning ... and decided to help her... then he come in with stick give her but still they are not yet out... I think the boy got hurt when he fall... and the mother stopped the dishes to take the boy to hospital... because this thing is very high... the stool... she didn’t because she was busy washing... the mother was a bit... 155 156 APPENDICES Appendix to Chapter 3 3.A Accuracy in production and comprehension of regular (n =13) and irregular (n =7) past tense in English by agrammatic participants in Chapter 3 Production Regular Irregular past-Tense past-Tense (%) (%) SW HJ LA MM JK EA PN JA MW VK HS JN SS mean % 3.B 9 (69.2) 2 (15.4) 11(84.6) 11(84.6) 13 (100.0) 3 (23.1) 1 (7.7) 10 (76.9) 5 (38.5) 6 (46.2 5 (38.5) 2 (15.4) 8 (61.5) 50.9 2 6 4 5 2 4 1 2 3 4 4 3 0 (0) (28.6) (85.7) (57.1) (71.4) (28.6) (57.1) (14.3) (28.6) (42.9) (57.1) (57.1) (42.9) 44.0 9 (45.0) 4 (20.0) 17 (85.0) 15 (75.0) 18 (90.0) 5 (25.0) 5 (25.0) 11(55.0) 7 (35.0) 9 (45.0) 9 (45.0) 6 (30.0) 11(55.0) 48.5 Comprehension Regular Irregular past-Tense past-Tense (%) (%) Compre. past-Tense (%) 13(100.0) 6 (46.2) 10 (76.9) 13(100.0) 13(100.0) 11(84.6) 6 (46.2) 12 (92.3) 9 (69.2) 7 (53.8) 9 (69.2) 8 (61.5) 11(84.6) 75.7 20(100.0) 7 (35.0) 17 (85.0) 20(100.0) 20(100.0) 18 (90.0) 9 (45.0) 16 (80.0) 15 (75.0) 11(55.0) 15 (75.0) 12 (60.0) 14 (70.0) 74.6 7 (100.0) 1(14.3) 7 (100.0) 7 (100.0) 7 (100.0) 7 (100.0) 3 (42.9) 4 (57.1) 6 (85.7) 4 (57.1) 6 (85.7) 4 (57.1) 3 (42.9) 72.5 The English verb pairs used in the Test for Assessing Reference of Time (TART; Bastiaanse et al., 2008) in Chapter 3 Examples to read - to write the letter Test items to to to to to to to Prod. past-Tense (%) paint - to draw a square tear - to glue the paper pour - to drink the milk eat - to peel the orange push - to pull the trolley sharpen - to break the pencil knit - to sew the cloth 6.3. FURTHER RESEARCH 157 to empty - to fill the folder to mop - to sweep the floor to iron - to fold the sweater 3.C The Swahili verb pairs used in the Test for Assessing Reference of Time (TART; Abuom, et al., 2010) in Chapter 3 Examples Kusoma - kuandika barua Test items Kupaka - kuchora umbo Kurarua - kubandika karatasi Kumwaga - kunywa maziwa Kula - kumenya chungwa Kusukuma - kuvuta toroli Kuchonga - kuvunja kalamu Kufuma - kushona nguo Kutoa - kuweka vitabu Kupigadeki - kufagia sakafu Kupigapasi - kukunja sweta 158 APPENDICES References Abuom, T. & Bastiaanse, R (2012). Characteristics of Swahili-English bilingual agrammatic spontaneous speech and the consequences for understanding agrammatic aphasia. Journal of Neurolinguistics, 25, 276-293. Abuom, T. & Bastiaanse, R. (2013). Production and Comprehension of Reference of Time in Swahili-English bilingual agrammatic speakers. Aphasiology, 27, 157-177. Abuom, T. & Bastiaanse, R. (in press). Sentence Production in Swahili-English bilingual agrammatic speakers. Aphasiology. Abuom, T., Sang, H., & Bastiaanse, R. (2010) Test for Assessing Reference of Time (TART): Bilingual Swahili-English version. Groningen: University of Groningen. Abuom, T., Obler, L., & Bastiaanse, R. (2011). Using Swahili and English to test explanations of agrammatism. Aphasiology, 25, 559- 575. Abuom, T., Shah, E., & Bastiaanse, R. (2013). Sentence Comprehension in Swahili-English agrammatic speakers. Journal of Clinical Linguistics & Phonetics, 27, 355-370. Abutalebi, J., Cappa, S. F., & Perani, D. (2005). What can functional imaging tell us about the bilingual brain? Kroll, J. F. & De Groot, A. M. B. (eds.) Handbook of bilingualism: Psycholinguistic approaches. (Ch. 25, pp 497-515). New York: Oxford University Press. Alajouanine, T. (1963). Dostoiewski’s epilepsy. Brain, 86, 209-218. Albert, M. L., & Obler, L. K. (1978). The Bilingual Brain: Neuropsychological and Neurolinguistic Aspects of Bilingualism. New York: Academic Press. Alexiadou, A. & Stavrakaki S. (2006) Clause structure and verb movement in a Greek-English bilingual patient with Broca’s aphasia: Evidence from adverb placement. Brain and Language, 96, 207-220. Anjarningsih, H. Y. (2012). Time reference in Standard Indonesian agrammatic aphasia (Doctoral Dissertation, University of Groningen). Groningen Dissertations in Linguistics (GRODIL),101. Anjarningsih & Bastiaanse. (2011). Verbs and time reference in Indonesian agrammatic speech. Aphasiology, 25, 1562-1578. 159 160 REFERENCES Anjarningsih, H.Y., Haryadi-Soebadi RD, Gofir A, Bastiaanse R. (2012) Characterising agrammatism in Standard Indonesian. Aphasiology, 26, 757-784. Arabatzi, M. & Edwards, S. (2002). Tense and syntactic processes in agrammatic speech. Brain and Language, 80, 314-327. Ashton, E.O. (1982). Swahili grammar. London: Longman, Green and Co. Ltd. Avrutin, S. (2000) Comprehension of discourse-linked and non-discourse-linked questions by children and Broca’s aphasics. In Grodzinksy, Y., Shapiro, L. & Swinney, D. (eds.) Language and brain: Representation and processing. (pp 295-313). San Diego: Academic Press. Avrutin, S. (2006). Weak syntax. In Amunts, K. & Grodzinsky, Y. (eds.) Broca’s region. (pp 49-62). New York: Oxford Press. Ballard, K. J. & Thompson, C. K. (1999). Treatment and generalization of complex sentence production in agrammatism. Journal of Speech, Language, and Hearing Research, 42, 690-707. Basso, A., Razzano, C., Faglioni, P., & Zanobio, M.E. (1990). Confrontation naming, picture description and action naming in aphasic patients. Aphasiology, 4, 185-195. Bastiaanse, R. (2008). Production of verbs in base position by Dutch agrammatic speakers: Inflection versus finiteness. Journal of Neurolinguistics, 21, 104-119. Bastiaanse, R. (2013). Why reference to the past is difficult for agrammatic speakers. Journal of Clinical Linguistics and Phonetics, 27, 244-263.. Bastiaanse, R., Bamyaci, E., Hsu, C., Lee, J., Yarbay Duman, T. & Thompson, C.K. (2011). Time reference in agrammatic aphasia: A cross-linguistic study. Journal of Neurolinguistics, 24, 652-673. Bastiaanse, R., Bouma, G., & Post, W. (2009). Linguistic complexity and frequency in agrammatic speech production. Brain and Language, 109, 18-28. Bastiaanse, R. & Edwards, S. (2004). Word order and finiteness in Dutch and English Broca’s and Wernicke’s aphasia. Brain and Language, 89, 91-107. Bastiaanse, R., Edwards, S., Maas, E. & Rispens, J. (2003) Assessing comprehension and production of verbs and sentences: The Verb and Sentence Test (VAST). Aphasiology, 17, 49-73. Bastiaanse, R., Edwards, S., & Rispens, J. E. (2002b ). The Verb and Sentence Test (VAST). Suffolk: Thames Valley Test Publishers. Bastiaanse, R., Hugen, J., Kos, M., & van Zonneveld, R.(2002a ). Lexical, morphological, and syntactic aspects of verb production in agrammatic speakers. Brain and Language, 80, 142-159. Bastiaanse, R. & Jonkers, R. (1998). Verb retrieval in action naming and REFERENCES 161 spontaneous speech in agrammatic and anomic aphasia. Aphasiology, 12, 951-969. Bastiaanse, R., Jonkers, R., & Thompson, C. K. (2008). Test for assessing reference of time (TART). Groningen: University of Groningen. Bastiaanse, R., Koekkoek, J., & Van Zonneveld, R. (2003). Object scrambling in Dutch Broca’s aphasia. Brain and Language, 86, 287-299. Bastiaanse, R., & Thompson, C.K. (2003). Verb and auxiliary movement in agrammatic Broca’s aphasia. Brain and Language, 84, 286-305. Bastiaanse, R. & Van Zonneveld, R. (2004). Broca’s aphasia, verbs and the mental lexicon. Brain and Language, 90, 198-202. Bastiaanse, R., & Van Zonneveld, R. (2006). Comprehension of passives in Broca’s aphasia. Brain and Language, 96, 135-142. Bastiaanse, R. & Van Zonneveld, R. (2005). Sentence production with verbs of alternating transitivity in Broca’s agrammatic aphasia, Journal of Neurolinguistics, 18, 57-66. Bates, E., Friederici, A., & Wulfeck, B. (1987). Grammatical morphology in aphasia: evidence from three languages. Cortex, 23, 545-574. Bates, E. A., Friederici, A. D., Wulfeck, B. B., & Juarez L. A. (1988). On the preservation of word order in aphasia: Cross-linguistic evidence. Brain and Language, 33, 323-364. Bates, E., Wulfeck, B., & MacWhinney, B. (1991). Crosslinguistic research in aphasia: An overview. Brain and Language, 41, 123-148. Beeke S, Wilkinson R, Maxim J (2003). Exploring aphasic grammar 1: a single case analysis of conversation. Clinical Linguistics and Phonetics, 17, 81 107. Berndt, R.S. & Haendiges, A.N. (2000). Grammatical class in Word and Sentence production: Evidence from and Aphasic patient. Journal of Memory and Language, 43, 249-273. Berndt, R., Mitchun, C. & Haendiges, A. (1996). Comprehension of reversible sentences in ‘agrammatism’: A meta-analysis. Cognition, 58, 289-308. Berndt, R.S., Mitchum, C.S., Haendiges, A.N., & Sandson, J. (1997). Verb retrieval inaphasia. Characterizing single word impairments. Brain and Language, 56, 68-106. Beretta, A. & Munn, A. (1998). Double-agents and trace-deletion in agrammatism. Brain and Language, 65, 404-421. Bresnan, J. & Mchombo, S. (1987). Topic, Pronoun and Agreement in Chichewa. Language, 63, 741-782. Brookshire, R.H. & Nicholas, L.E. (1994). Speech sample size and test-retest stability of connected speech measures for adults with aphasia. Journal of Speech and Hearing Research, 37, 399-407. 162 REFERENCES Burchert, F., Meiner, N. & De Bleser, R. (2008). Production of non-canonical sentences in agrammatic aphasia: Limits in representation or rule application? Brain and Language, 104, 170-179. Burchert, F., Swoboda-Moll, M., & De Bleser, R. (2005). Tense and agreement dissociations in German agrammatic speakers: underspecification versus hierarchy. Brain and Language, 94, 188-199. Burgio, F. & Basso, A (1997). Memory and aphasia. Neuropsychologia, 35, 759-766. Bybee, J. (2007). Frequency of use and the organization of language. Oxford University Press. Caramazza, A., & Berndt, R. S. (1985). A multicomponent deficit view of agrammatic Broca’s aphasia. In M. L. Kean (Ed.), Agrammatism. London: Academic Press. Caramazza, A. & Hillis, A. (1991). Lexical organization of nouns and verbs in the brain. Nature, 349, 788-790. Caramazza, A., & Zurif, E. B. (1976). Dissociation of algorithmic and heuristic processes in language comprehension: Evidence from aphasia. Brain and Language, 3, 572-582. Centeno, J.G. & Anderson, R.T. (2011). A preliminary comparison of verb tense production in Spanish speakers with expressive restrictions. Clinical Linguistics & Phonetics, 25, 864-880. Charlton, M.H. (1964). Aphasia in bilingual and polyglot patients: A neurological and psychological study. Journal of Speech and Hearing Disorders, 29, 307-311. Chomsky, N. (1981). Lectures on government and binding. Dordrecht: Foris Publications. Chomsky, N. (1982). Some Concepts and Consequences of the Theory of Government and Binding. Cambridge. MA: The MIT Press. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT press. Choy, J. J., & Thompson, C. K. (2010). Binding in agrammatic aphasia: processing to comprehension. Aphasiology, 24, 551-579. Clahsen, H. & Ali, M. (2009). Formal features in aphasia: Tense, agreement, and mood in English agrammatism. Journal of Neurolinguistics 22, 436450. Comrie, B. (1976). Aspect: An Introduction to the study of Verbal Apsect and other related problems. Cambridge University Press. Crepaldi, D, Ingignoli, C., Verga, R., Contardi, A., Semenza, C. & Luzzatti, C. (2011). On nouns, verbs, lexemes, and lemmas: Evidence from the spontaneous speech of seven aphasic patients. Aphasiology, 25, 71-92. REFERENCES 163 De Diego-Balaguer, R., Costa, A., Sebastin-Galls, N., Juncadella, M., & Caramazza, A. (2004). Regular and irregular morphology and its relationship with agrammatism: Evidence from two Spanish-Catalan bilinguals. Brain and Language, 91, 212-222. Dimmendaal, G. (2001). Language shift and morphological divergence in the Nilotic area. Sprache und Geschichte in Afrika, 16/17, 83-124. Dickey, M.W., Choy, J.J., & Thompson, C.K. (2007). Real-time comprehension of whmovement in aphasia: Evidence from eyetracking while listening. Brain and Language, 100, 1-22. Dickey, M., Milman, L., & Thompson, C.K. (2005). Preception of functional morphology in agrammatic Broca’s aphasia. Brain and Language, 95, 8283. Dickey, M. W. & Thompson, C. K. (2007). The relation between syntactic and morphological recovery in agrammatic aphasia: A case study. Aphasiology, 21, 604-616. Drai, D., Grodzinsky, Y. 2006. The variability debate: More statistics, more linguistics. Brain & Language, 76.2, 157-170 Dressler, W. 1985. Morphonology: the dynamics of derivation. Ann Arbor: Karoma Press. Edmonds, L., & Kiran, S. (2006). Effect of semantic naming treatment on crosslinguistic generalization in bilingual aphasia. Journal of Speech Language and Hearing Research. 49, 729-748. Fabbro, F. (2001). The bilingual brain: bilingual aphasia. Brain and Language, 79, 201-210. Fabbro, F. (1999). The neurolinguistics of bilingualism. Hove, UK: Psychology Press. Fabbro, F. & Frau, G. (2001). Manifestations of aphasia in Friulian. Journal of Neurolinguistics, 14, 255-279. Faroqi-Shah, Y. & Dickey, M.W. (2009). On-line processing of tense and temporality in agrammatic aphasia. Brain and Language, 108, 97-111. Faroqi-Shah, Y., & Thompson, C. K. (2007). Verb inflections in agrammatic aphasia: Encoding of tense features. Journal of Memory and Language, 56, 129-151. Faroqi-Shah, Y. & Waked, A. (2010). Grammatical category dissociation in multilingual aphasia. Cognitive Neuropsychology, 27, 181-203 Frazier, L., & Friederici, A. (1991). On deriving the properties of agrammatic comprehension. Brain and Language, 40, 51-66. Friedmann, N. (2005). Degrees of severity and recovery in agrammatism: climbing up the syntactic tree. Aphasiology, 19, 1037-1051. Friedmann, N. (1998). Functional categories in agrammatic production: A 164 REFERENCES crosslinguistic study. Unpublished doctoral dissertation. Tel Aviv University. Friedmann, N. (2000). Moving verbs in agrammatic production. In R. Bastiaanse & Y. Grodzinsky (Eds.), Grammatical disorders in aphasia: A neourolinguistic perspective (pp. 152-170). London: Whurr. Friedmann, N. (2002). Question production in agrammatism: the Tree Pruning Hypothesis. Brain and Language, 80, 160-187. Friedmann, N. (2008). Traceless relatives: agrammatic comprehension of relative clauses with resumptive pronouns. Journal of Neurolinguistics, 21, 138-149. Friedmann, N. & Grodzinsky, Y. (1997). Tense and agreement in agrammatic production: Pruning the syntactic tree. Brain and Language, 56, 397-425. Friedmann, N., & Gvion, A. (2003). Sentence comprehension and working memory limitation in aphasia: A dissociation between semantic-syntactic and phonological reactivation. Brain and Language, 86, 23-39. Friedmann, N., Reznick, J., Dolinski-Nuger, D., & Soboleva, K. (2010). Comprehension and production of movement-derived sentences by Russian speakers with agrammatic aphasia. Journal of Neurolinguistics, 23, 44-65 Goodglass, H. (1968). Studies in the grammar of aphasics. In S. Rosenberg, & J. Koplin (Eds.), Developments in applied psycholinguistic research. New York: MacMillan. Goodglass, H. & Kaplan, E. (1972). Boston Diagnostic Aphasia Examination. Philadelphia: Lea & Febiger. Goodglass, H., Kaplan, E. & Barresi, B. (2001) The Boston Diagnostic Aphasia Examination. Philadelphia: Lippincott, Williams & Wilkins. Goral, M. (2011). Treatment of grammatical disorders in bilingual aphasia. Presentation at the Science of Aphasia XII, September 1-5, Barcelona (Spain). Goral, M., Levy, E.S., and Kastl, R. (2010). Cross-language treatment generalization: A case of trilingual aphasia. Aphasiology. 24, 170-187. Grodzinsky, Y. (1984). The syntactic characterization of agrammatism. Cognition,16, 99-120. Grodzinsky, Y. (1986). Language deficits and the theory of syntax. Brain and Language, 27, 135-159. Grodzinsky, Y. (1990). Theoretical Perspectives on Language Deficits. Cambridge, MA: The MIT Press. Grodzinksy, Y. (1991). There is an entity called agrammatic aphasia. Brain and Language, 41, 555-564. Grodzinsky, Y. (1995). A restrictive theory of agrammatic comprehension. Brain and Language, 51, 26-51. REFERENCES 165 Grodzinksy, Y. (1999). The neurology of syntax: Language use without Broca’s area. Behavioral and Brain Sciences, 23, 47-117. Grodzinsky, Y. (2000). Overarching agrammatism. In Grodzinsky, Y., L. Shapiro & D. Swinney, eds., 2000. Language and the brain. San Diego: Academic Press. Grodzinsky, Y., Piango, M. M., Zurif, E., & Drai, D. (1999). The critical role of group studies in neuropsychology: Comprehension regularities in Broca’s aphasia. Brain and Language, 67, 134-147. Grosjean, F. (1994). Individual bilingualism. In R. E. Asher (ed.), The encyclopaedia of language and linguistics. (pp. 1656-1660). Oxford, UK: Pergamon Press. Gvion, A., & Friedmann, N (2012). Aphasia: Acquired language pathology. In G. Khan (Ed.), Encyclopedia of Hebrew Language and Linguistics. Boston, USA: Brill USA. Haarmann, H.J. & Kolk, H.H. (1991). Syntactic priming in Broca’s aphasia: Evidence for slow activation. Aphasiology, 5, 247-263. Hagiwara, H. (1995). The breakdown of functional categories and the economy of derivation. Brain and Language, 50, 92-116. Hartmann, R.R.K. & Stork, F.C. (1972). Dictionary of language and linguistics. London: Applied Science Publishers. Huber W, Poeck K, Weniger D. (2002). Aphasie. In: Hartje W, Poeck K, eds. Klinische Neuropsychologie (pp.93-160). Stuttgart: Thieme. Ivanova, M.V. & Hallowell, B. (2012). Validity of an eye-tracking methods to index working memory in people with and without aphasia. Aphasiology, 26, 556-578. Jacobs, B. J. & Thompson, C. K. (2000). Cross-modal generalization effects of training noncanonical sentence comprehension and production in agrammatic aphasia. Journal of Speech, Language, and Hearing Research, 43, 5-20. Jonkers, R., & Bastiaanse, R. (1998). How selective are selective word class deficits? Two case studies of action and object naming. Aphasiology, 12, 245-256. Jonkers, R. & Bruin, A. de (2009) Tense processing in Broca’s and Wernicke’s aphasia. Aphasiology 23, 1252-1265. Karanth, P., Ahuja, G. K., Nagaraj, D., Pandit, R., & Shivashankar, N. (1991). Cross cultural studies of aphasia. In J. S. Chopra, K. Jagannathan, & I. M. S. Sawhney (Eds.), Modern trends in neurology. New Delhi: Churchill Livingstone. Keach, C. (1995). Subject and Object markers as agreement and pronoun incorporation in Swahili. In A. Akinlabi (Ed), Theoretical approaches to 166 REFERENCES African Linguistics (pp.109-116). Trenton NJ: African World Press. Kim, M., & Thompson, C. K. (2000). Patterns of comprehension and production of nouns and verbs in agrammatism: Implications for lexical organization. Brain and Language, 74, 1-25. King, J. W., & Just, M. A. (1991). Individual differences in syntactic processing: The role of working memory. Journal of Memory and Language, 30, 580-602. Knoph, M. (2011). Language assessment of a Farsi-Norwegian bilingual speaker with aphasia. Clinical Linguistics & Phonetics, 25, 530- 539. Kok, P., van Doorn, A. & Kolk, H. (2007). Inflection and computational load in agrammatic speech. Brain and Language, 102, 273-283. Kolk, H. H. J. (1998). Disorders of syntax in aphasia. Linguistic-descriptive and processing approaches. In B. Stemmer & H. Whitaker (eds) Handbook of Neurolinguistics (pp.249-260). San Diego, CA: Academic Press. Krifka, M. (1995). Swahili. Syntax: Ein Internationales Handbuch Zeitgenossischer Forschung. In . J. Jacobs, A. von Stechow, W. Sternefeld and T. Vennemann (Eds). An International Handbook of Contemporary Research (pp1397-1418). Berlin: Walter de Gruyter. Lee, J., Milman, L, & Thompson, C. K. (2008). Functional category production in English agrammatism. Aphasiology, 22, 893-905. Linebarger, M. C., Schwartz, M. F., & Saffran, E. M. (1983). Sensitivity to grammatical structure in so-called agrammatic aphasics. Cognition, 13, 361-392. Loogman, A. (1965). Swahili Grammar and Syntax, Duquesne studies-African Series Editions. Pittsburg, P.A.: Duquesne Univesrsity Press. Lorch, M.P. (1990). Cross-linguistic study of the agrammatic impairment in verb inflection: Icelandic, Hindi, and Finnish cases’. In J.-L. Nespoulous and P. Villiard, (eds.) Morphology, phonology and aphasia. Springer Verlag, New York. Lukatela, K., Shankweiler D., & Crain, S. (1995). Syntactic processing in agrammatic aphasia by speakers of a Slavic language. Brain and Language, 49, 50-76. Marshall, J. C. (1986). The description and interpretation of aphasic language disorder. Neuropsychologia, 24, 5-24. Menn, L., & Obler, L. (Eds.). (1990). Agrammatic aphasia: A cross-language narrative sourcebook. Philadelphia: John Benjamin. Menn, L. & Obler, L. (1990). Cross-language data and theories of agrammatism. In Menn, L. & Obler, L.K. (eds.) Agrammatic aphasia: A crosslanguage narrative sourcebook, Vol. 2, 1369-1389. Amsterdam: John Benjamins. REFERENCES 167 Miceli, G., Mazzucchi, A., Menn, L., & Goodglass, H. (1983). Contrasting cases of Italian agrammatic aphasia without comprehension disorder. Brain and Language, 19, 65-97. Miceli, G., Silveri, C., Villa, G., & Caramazza, A. (1984). On the basis for the agrammatic’s difficulty in producing main verbs. Cortex, 20, 207-220. Miozzo, M., Costa, A., Hernndez, M., & Rapp, B. (2010). Lexical processing in the bilingual brain: Evidence from grammatical/morphological deficits. Aphasiology, 24, 262-287. Miyake, A., Carpenter, P. A., & Just, M. A. (1994). A capacity approach to syntactic comprehension disorders: Making normal adults perform like aphasic patients. Cognitive Neuropsychology, 11, 671-717. Mohammed, M.A. (2001). Modern Swahili grammar. Nairobi: East African Educational Publishers LTD. Myachina, E. N. (1981). The Swahili language: a descriptive grammar. London: Routlege & Kegan Paul. Nanousi, V., Masterson, J., Druks, J., & Atkinson, M. (2006). Interpretable vs. uninterpretable features: Evidence from six Greek-speaking agrammatic patients. Journal of Neurolinguistics, 19, 209-238. Nespoulous, J.-L., Dordain, M., Perron, C., Jarema, G., & Chazal, M. (1990). Agrammatism in French: Two case studies. In L. Menn & L. Obler (Eds.), Agrammatic aphasia: A cross-language narrative sourcebook. Philadelphia: John Benjamin. Nespoulous, J.-L., Dordain, M., Perron, C., Ska, B., Bub, D., Caplan, D., Mehler, J., & Lecours, A. R. (1988). Agrammatism in sentence production without comprehension deficits: Reduced availability of syntactic structures or of grammatical morphemes? A case study. Brain and Language, 33, 273-295. Ngonyani, D. (1996). The Morphosyntax of Applicatives. Unpublished Doctoral dissertation, UCLA. Obler, L. K., & Albert, M. L. (1996). Language and communication in aging and dementia. In J. Birren (Ed.), Encyclopedia of gerontology (Vol. 2, pp. 1-6). San Diego, CA: Academic Press. Obler, L. K., & Mahecha, N. (1991). First language loss in bilinguals with brain-damage. In H. Seliger & R. Vago (Eds.), First language attrition: Structural and theoretical perspectives (pp. 53-650). Cambridge, UK: Cambridge University Press. Olness, G.S. (2006). Genre, verb, and coherence in picture-elicited discourse of adults with aphasia. Aphasiology, 20, 175-187. Ouhalla, J. (1993). Functional categories, agrammatism and language acquisition. Linguistische Berichte, 143, 3-36. 168 REFERENCES Ouhalla, J. (1990). Sentential negation, relativized minimality and the aspectual status of auxiliaries. The Linguistic Review, 7, 183-231. Paradis, M. (1995). Introduction: The need for distinctions. In M. Paradis (Ed.), Aspects of bilingual aphasia (1- 9). Oxford, UK: Pergamon Press. Paradis, M. (2001). (ed.). Manifestations of aphasia symptoms in different languages. Oxford: Pergamon Press. Paradis, M. (1988) Recent Developments in studies of agrammatism: Their import for the assessment of bilingual aphasia. Journal of Neurolinguistics, 3, 127-160. Paradis, M., & Goldblum, M.C. (1989). Selected crossed aphasia in a trilingual patient followed by reciprocal antagonism. Brain and Language, 36, 62-75. Paradis, M. & Mwansasu, A. (1990) Bilingual Aphasia Test: English-Swahili version. Hillsdale, NJ: LEA. 12pp. Penke, M. (2001). Controversies about CP: a comparison of language acquisition and language impairment in Broca’s aphasia. Brain and Language, 77, 351-363. Perrot, D. (1969). Teach yourself Swahili. New York: David Mckay Company, Inc. Piñango, M. (2000). Syntactic displacement in Broca’s agrammatic aphasia. In: R. Bastiaanse & Y. Grodzinsky (Eds.), Grammatical disorders in aphasia (pp.75-87). London: Whurr Publishers. Pollock, J. Y. (1989). Verb movement, universal grammar and the structure of IP. Linguistic Inquiry, 20, 365-424. Prins, R. & Bastiaanse, R. (2006). The early history of aphasiology: From the Egyptian surgeons (c. 1700 bc) to Broca (1861). Aphasiology, 20, 762-791. Rossi, E. (2007). Clitic production in Italian agrammatism. Groningen Dissertations in Linguistics (GRODIL). Groningen: University of Groningen. Rossi, E. & Bastiaanse, R. & (2008). Spontaneous speech in Italian agrammatic aphasia: A focus on verb production. Aphasiology, 22, 347-362. Saffran, E. M., Berndt, R. S., Schwartz, M. F. (1989). The quantitative analysis of agrammatic production: Procedure and data. Brain and Language, 37, 440-479. Sanchez, M.E. (1996). Syntactic features in agrammatic production (Unpublished doctoral dissertation). University of British Colombia, Vancouver, Canada. Schwartz, M. F., Linebarger, M. C., Saffran, E. M., & Pate, D. S. (1987). Syntactic transparency and sentence interpretation in aphasia. Language and Cognitive Processes, 2, 85-113. Semenza, C., Luzzatti, C., & Carabelli, S. (1997). Morphological representation of compound nouns: A study on Italian aphasic patients. Journal of REFERENCES 169 Neurolinguistics, 10, 33-43. Simonsen, H. G. & Lind, M. (2002). Past tense expression in a Norwegian man with Broca’s aphasia. In Windsor, F., Kelly, M. L. & Hewkett, N. (eds.) Investigations in clinical linguistics and phonetics (pp 44-56). London: Lawrence Erlbaum. Stavrakaki, S., & Kouvava, S. (2003). Functional categories in agrammatism: Evidence from Greek. Brain and Language, 86, 129- 141. Steere, E. (1976). Handbook of Swahili. London: Society for Promoting Christian Knowledge. Stengel, E. & Zelmanowicz, J. (1933). Uber polyglotte motorische Aphasie. Zeitschrift fur die gesamte Neurologie und Psychiatrie, 149, 292-311. Thompson, C. K. (2003). Unaccusative verb production in agrammatic aphasia: the argument structure complexity hypothesis. Journal of Neurolinguistics, 16, 151-167. Thompson, C. K., & Choy, J. (2009). Pronominal resolution and gap filling in agrammatic aphasia: Evidence from eye movements. Journal of Psycholinguistic Research, 38, 255-283. Thompson, C. K., Choy, J. J., Holland, A., & Cole, R. (2010). Sentactics Computer-Automated Treatment of Underlying Forms. Aphasiology, 24, 1242-1266. Thompson, C. K., Lange, K. L., Schneider, S. L., & Shapiro, L. P. (1997). Agrammatic and non-brain damaged subjects’ verb and verb argument structure production. Aphasiology, 11, 473-490. Thompson, C. K., & Shapiro, L. P. (2005). Treating agrammatic aphasia within a linguistic framework: Treatment of underlying forms. Aphasiology, 19, 1021-1036. Thompson. C. K., Shapiro, L. P., Ballard, K. J., Jacobs, B. J., Schneider, S. S., & Tait, M. E. (1997). Training and generalized production of Whand NP-movement structures in agrammatic aphasia. Journal of Speech, Language, and Hearing Research, 40, 228-244. 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. Thompson, C. K., Shapiro, L. P., Li, L., & Schendel, L. (1995). Analysis of verbs and verb argument structure: A method for quantification of agrammatic language production. Clinical Aphasiology, 23, 121- 140. Thompson, C. K., Shapiro, L. P., Tait, M. E., Jacobs, B. J., & Schneider, S. L. (1996). Training Wh-question production in agrammatic aphasia: Analysis of argument and adjunct movement. Brain and Language, 52, 170 REFERENCES 175-228. Trousseau, A. (1864). ‘De l’aphasie, maladie d’crite recomment sous le nom impropre d’aphmie,’ Gazette des Hpitaux Civils et Militaires 1: 13-14. Tsapkini, K., Jarema, G., & Kehayia, E. (2001). Manifestations of morphological impairments in Greek aphasia: A case study. Journal of Neurolinguistics, 14, 281-296. Varlokosta, S., Valeonti, N., Kakavoulia, M., Lazaridou, M., Economou, A., & Protopapas, A. (2006). The breakdown of functional categories in Greek aphasia: Evidence from agreement, tense, and aspect. Aphasiology, 20, 723-743. Wenzlaff, M. & Clahsen, H. (2005). Finiteness and verb-second in German agrammatism, Brain & Language, 92, 33-44. Wenzlaff, M., & Clahsen, H. (2004). Tense and agreement in German agrammatism. Brain and Language, 89, 57-68. Whitaker, H. A. (1998). Neurolinguistics from the Middle Ages to the premodern area: Historical vignettes. In B. Stemmer & H. A. Whitaker (Eds.), Handbook of neurolinguistics (pp. 27-54). San Diego, CA: Academic Press. Yarbay Duman, T., Aygen, G., & Bastiaanse, R. (2008). The production of Turkish relative clauses in agrammatism: verb inflection and constituent order. Brain and Language, 105, 149-160. Yarbay Duman, T., Aygen, G., Özgirgin, N., & Bastiaanse, R. (2007). Object scrambling and finiteness in Turkish agrammatic production. Journal of Neurolinguistics, 20, 306-331. Yarbay Duman, T. & Bastiaanse, R. (2009). Time reference through verb inflection in Turkish agrammatic aphasia. Brain and Language, 108, 3039. Yarbay Duman, T., Özgirgin, N., Altinok, N., & Bastiaanse, R. (2011). Sentence comprehension in Turkish Broca’s aphasia: An integration problem. Aphasiology, 25, 908-926. Zagona, K. (2003). Tense and anaphora: Is there a tense-specific theory of coreference. In Barrs, A. (ed.) Anaphora: A Reference Guide. (pp 140171). Oxford: Blackwell Publishing. Zagona, K. (in press). Tense, aspect and modality. In Dikken, M. den (ed.) Encyclopedia of Generative Syntax. Cambridge: Cambridge University Press. Groningen Dissertations in Linguistics (Grodil) 1. Henriëtte de Swart (1991). Adverbs of Quantification: A Generalized Quantifier Approach. 2. Eric Hoekstra (1991). Licensing Conditions on Phrase Structure. 3. Dicky Gilbers (1992). Phonological Networks. A Theory of Segment Representation. 4. Helen de Hoop (1992). Case Configuration and Noun Phrase Interpretation. 5. Gosse Bouma (1993). Nonmonotonicity and Categorial Unification Grammar. 6. Peter Blok (1993). The Interpretation of Focus: an epistemic approach to pragmatics. 7. Roelien Bastiaanse (1993). Studies in Aphasia. 8. Bert Bos (1993). Rapid User Interface Development with the Script Language Gist. 9. Wim Kosmeijer (1993). Barriers and Licensing. 10. Jan-Wouter Zwart (1993). Dutch Syntax: A Minimalist Approach. 11. Mark Kas (1993). Essays on Boolean Functions and Negative Polarity. 12. Ton van der Wouden (1994). Negative Contexts. 13. Joop Houtman (1994). Coordination and Constituency: A Study in Categorial Grammar. 14. Petra Hendriks (1995). Comparatives and Categorial Grammar. 15. Maarten de Wind (1995). Inversion in French. 16. Jelly Julia de Jong (1996). The Case of Bound Pronouns in Peripheral Romance. 17. Sjoukje van der Wal (1996). Negative Polarity Items and Negation: Tandem Acquisition. 18. Anastasia Giannakidou (1997). The Landscape of Polarity Items. 19. Karen Lattewitz (1997). Adjacency in Dutch and German. 20. Edith Kaan (1997). Processing Subject-Object Ambiguities in Dutch. 21. Henny Klein (1997). Adverbs of Degree in Dutch. 22. Leonie Bosveld-de Smet (1998). On Mass and Plural Quantification: The Case of French ‘des’/‘du’-NPs. 23. Rita Landeweerd (1998). Discourse Semantics of Perspective and Temporal Structure. 24. Mettina Veenstra (1998). Formalizing the Minimalist Program. 25. Roel Jonkers (1998). Comprehension and Production of Verbs in Aphasic Speakers. 26. Erik F. Tjong Kim Sang (1998). Machine Learning of Phonotactics. 27. Paulien Rijkhoek (1998). On Degree Phrases and Result Clauses. 28. Jan de Jong (1999). Specific Language Impairment in Dutch: Inflectional Morphology and Argument Structure. 171 172 GRODIL 29. Hae-Kyung Wee (1999). Definite Focus. 30. Eun-Hee Lee (2000). Dynamic and Stative Information in Temporal Reasoning: Korean Tense and Aspect in Discourse. 31. Ivilin Stoianov (2001). Connectionist Lexical Processing. 32. Klarien van der Linde (2001). Sonority Substitutions. 33. Monique Lamers (2001). Thematic Information. Sentence Processing: Using Syntactic, Semantic, and 34. Shalom Zuckerman (2001). The Acquisition of “Optional” Movement. 35. Rob Koeling (2001). Dialogue-Based Disambiguation: Using Dialogue Status to Improve Speech Understanding. 36. Esther Ruigendijk (2002). Case Assignment in Agrammatism: a Cross-linguistic Study. 37. Tony Mullen (2002). An Investigation into Compositional Features and Feature Merging for Maximum Entropy-Based Parse Selection. 38. Nanette Bienfait (2002). Grammatica-onderwijs aan allochtone jongeren. 39. Dirk-Bart den Ouden (2002). Phonology in Aphasia: Syllables and Segments in Levelspecific Deficits. 40. Rienk Withaar (2002). Comprehension. The Role of the Phonological Loop in Sentence 41. Kim Sauter (2002). Transfer and Access to Universal Grammar in Adult Second Language Acquisition. 42. Laura Sabourin (2003). Grammatical Gender and Second Language Processing: An ERP Study. 43. Hein van Schie (2003). Visual Semantics. 44. Lilia Schürcks-Grozeva (2003). Binding and Bulgarian. 45. Stasinos Konstantopoulos (2003). Using ILP to Learn Local Linguistic Structures. 46. Wilbert Heeringa (2004). Levenshtein Distance. Measuring Dialect Pronunciation Differences using 47. Wouter Jansen (2004). Laryngeal Contrast and Phonetic Voicing: A Laboratory Phonology Approach to English, Hungarian and Dutch. 48. Judith Rispens (2004). Dyslexia. Syntactic and Phonological Processing in Developmental 49. Danielle Bougaı̈ré (2004). L’approche communicative des campagnes de sensibilisation en santé publique au Burkina Faso: les cas de la planification familiale, du sida et de l’excision. 50. Tanja Gaustad (2004). Linguistic Knowledge and Word Sense Disambiguation. 51. Susanne Schoof (2004). An HPSG Account of Nonfinite Verbal Complements in Latin. 52. M. Begoña Villada Moirón (2005). Data-driven identification of fixed expressions and their modifiability. 53. Robbert Prins (2005). Finite-State Pre-Processing for Natural Language Analysis. 54. Leonoor van der Beek (2005). Topics in Corpus-Based Dutch Syntax. 55. Keiko Yoshioka (2005). Linguistic and gestural introduction and tracking of referents in L1 and L2 discourse. GRODIL 56. Sible Andringa (2005). language proficiency. 173 Form-focused instruction and the development of second 57. Joanneke Prenger (2005). Taal telt! Een onderzoek naar de rol van taalvaardigheid en tekstbegrip in het realistisch wiskundeonderwijs. 58. Neslihan Kansu-Yetkiner (2006). Blood, Shame and Fear: Self-Presentation Strategies of Turkish Women’s Talk about their Health and Sexuality. 59. Mónika Z. Zempléni (2006). Functional imaging of the hemispheric contribution to language processing. 60. Maartje Schreuder (2006). Prosodic Processes in Language and Music. 61. Hidetoshi Shiraishi (2006). Topics in Nivkh Phonology. 62. Tamás Biró (2006). Finding the Right Words: Implementing Optimality Theory with Simulated Annealing. 63. Dieuwke de Goede (2006). Verbs in Spoken Sentence Processing: Unraveling the Activation Pattern of the Matrix Verb. 64. Eleonora Rossi (2007). Clitic production in Italian agrammatism. 65. Holger Hopp (2007). Ultimate Attainment at the Interfaces in Second Language Acquisition: Grammar and Processing. 66. Gerlof Bouma (2008). Starting a Sentence in Dutch: A corpus study of subject- and object-fronting. 67. Julia Klitsch (2008). Open your eyes and listen carefully. Auditory and audiovisual speech perception and the McGurk effect in Dutch speakers with and without aphasia. 68. Janneke ter Beek (2008). Restructuring and Infinitival Complements in Dutch. 69. Jori Mur (2008). Off-line Answer Extraction for Question Answering. 70. Lonneke van der Plas (2008). Automatic Lexico-Semantic Acquisition for Question Answering. 71. Arjen Versloot (2008). Mechanisms of Language Change: Vowel reduction in 15th century West Frisian. 72. Ismail Fahmi (2009). Automatic term and Relation Extraction for Medical Question Answering System. 73. Tuba Yarbay Duman (2009). Reference and Case. Turkish Agrammatic Aphasia: Word Order, Time 74. Maria Trofimova (2009). Case Assignment by Prepositions in Russian Aphasia. 75. Rasmus Steinkrauss (2009). Frequency and Function in WH Question Acquisition. A Usage-Based Case Study of German L1 Acquisition. 76. Marjolein Deunk (2009). Discourse Practices in Preschool. Participation in Everyday Classroom Activities. Young Children’s 77. Sake Jager (2009). Towards ICT-Integrated Language Learning: Developing an Implementation Framework in terms of Pedagogy, Technology and Environment. 78. Francisco Dellatorre Borges (2010). Parse Selection with Support Vector Machines. 79. Geoffrey Andogah (2010). Geographically Constrained Information Retrieval. 80. Jacqueline van Kruiningen (2010). Onderwijsontwerp Probleemoplossing in interprofessioneel overleg. als conversatie. 81. Robert G. Shackleton (2010). Quantitative Assessment of English-American Speech Relationships. 174 GRODIL 82. Tim Van de Cruys (2010). Mining for Meaning: The Extraction of Lexico-semantic Knowledge from Text. 83. Therese Leinonen (2010). An Acoustic Analysis of Vowel Pronunciation in Swedish Dialects. 84. Erik-Jan Smits (2010). Acquiring Quantification. How Children Use Semantics and Pragmatics to Constrain Meaning. 85. Tal Caspi (2010). A Dynamic Perspective on Second Language Development. 86. Teodora Mehotcheva (2010). After the fiesta is over. Foreign language attrition of Spanish in Dutch and German Erasmus Students. 87. Xiaoyan Xu (2010). English language attrition and retention in Chinese and Dutch university students. 88. Jelena Prokić (2010). Families and Resemblances. 89. Radek Šimı́k (2011). Modal existential wh-constructions. 90. Katrien Colman (2011). Behavioral and neuroimaging studies on language processing in Dutch speakers with Parkinson’s disease. 91. Siti Mina Tamah (2011). A Study on Student Interaction in the Implementation of the Jigsaw Technique in Language Teaching. 92. Aletta Kwant (2011). Geraakt door prentenboeken. Effecten van het gebruik van prentenboeken op de sociaal-emotionele ontwikkeling van kleuters. 93. Marlies Kluck (2011). Sentence amalgamation. 94. Anja Schüppert (2011). Origin of asymmetry: Mutual intelligibility of spoken Danish and Swedish. 95. Peter Nabende (2011). Applying Dynamic Bayesian Networks in Transliteration Detection and Generation. 96. Barbara Plank (2011). Domain Adaptation for Parsing. 97. Çağrı Çöltekin (2011). Catching Words in a Stream of Speech: Computational simulations of segmenting transcribed child-directed speech. 98. Dörte Hessler (2011). Audiovisual Processing in Aphasic and Non-Brain-Damaged Listeners: The Whole is More than the Sum of its Parts. 99. Herman Heringa (2012). Appositional constructions. 100. Diana Dimitrova (2012). Neural Correlates of Prosody and Information Structure. 101. Harwintha Anjarningsih (2012). Time Reference in Standard Indonesian Agrammatic Aphasia. 102. Myrte Gosen (2012). Tracing learning in interaction. An analysis of shared reading of picture books at kindergarten. 103. Martijn Wieling (2012). A Quantitative Approach to Social and Geographical Dialect Variation. 104. Gisi Cannizzaro (2012). Early word order and animacy. 105. Kostadin Cholakov (2012). Lexical Acquisition for Computational Grammars. A Unified Model. 106. Karin Beijering (2012). Expressions of epistemic modality in Mainland Scandinavian. A study into the lexicalization-grammaticalization-pragmaticalization interface. 107. Veerle Baaijen (2012). The development of understanding through writing. GRODIL 175 108. Jacolien van Rij (2012). Pronoun processing: Computational, behavioral, and psychophysiological studies in children and adults. 109. Ankelien Schippers (2012). dependencies. Variation and change in Germanic long-distance 110. Hanneke Loerts (2012). Uncommon gender: Eyes and brains, native and second language learners, & grammatical gender. 111. Marjoleine Sloos (2013). Frequency and phonological grammar: approach. Evidence from German, Indonesian, and Japanese. An integrated 112. Aysa Arylova (2013). Possession in the Russian clause. Towards dynamicity in syntax. 113. Daniël de Kok (2013). Reversible Stochastic Attribute-Value Grammars. 114. Gideon Kotzé (2013). Complementary approaches to tree alignment: Combining statistical and rule-based methods. 115. Fridah Katushemererwe (2013). Computational Morphology and Bantu Language Learning: an Implementation for Runyakitara. 116. Ryan C. Taylor (2013). Pronoun Resolution. Tracking Referents: Markedness, World Knowledge and 117. Hana Smiskova-Gustafsson (2013). Perspective. Chunks in L2 Development: A Usage-based 118. Milada Walková (2013). The aspectual function of particles in phrasal verbs. 119. Tom O. Abuom (2013). Verb and word order deficits in Swahili-English bilingual agrammatic speakers. Grodil Secretary of the Department of General Linguistics Postbus 716 9700 AS Groningen The Netherlands
© Copyright 2024 Paperzz