John N. Williams

Aspects of bilingual lexical processing

M.W. Lee (1997)

 

This dissertation presents an experimental psycholinguistic study of two issues in bilingual lexical processing: spoken word production in a selected language and organisation of bilingual lexical memory. The first chapter gives a description of the bilinguals on whom the experimental work reported in this dissertation was carried out, and explains how the two bilingual issues investigated in this dissertation relate to ongoing work on lexical processing by monolinguals.

Chapter 2 addresses the question of whether words of the unwanted language are activated at all in the lexical selection process during bilinguals' spoken word production in a selected language. Many lines of work in cognitive psychology and applied linguistics can be seen as having some relevance to the question and they are critically reviewed. An experimental study of picture naming by bilinguals is presented, the results of which suggest that, during word production in a selected language, cross-language lexical competition takes place, but there is also a selected language bias which makes within-language lexical competition stronger than cross-language competition.

Chapter 3 is concerned with the system of lexical mappings that underlies bilinguals' lexical knowledge. A survey is provided of the experimental work that descends from Weinreich's descriptive typology of the co-ordinative, compound and sub-ordinative types of bilingual lexical organisation. Four experiments are reported: The first experiment indicates that the bilingual subjects tested used direct lexical links to perform L2-to-L1 word translating; the remaining experiments capitalise on this finding and produce results which suggest that cross-language direct lexical links should be located at an intermediate level between concept and word-form and which therefore challenge the traditional assumption of only a conceptual and a word-form level of representation in bilingual lexical memory.

The two questions addressed in Chapters 2 and 3 have in the past been looked at under various guises. The originality of the current approach lies in the attempt explicitly to relate questions about bilinguals to recent developments in (monolingual-centred) experimental psycholinguistics. The final chapter further argues for the theoretical significance of work on bilingualism by way of a discussion of the implications that the work reported in this dissertation may have for selected basic issues in cognitive science.

 

 

Mutual exclusivity and bilingual lexical acquisition

Kathleen McClure (1998)

 

According to the mutual exclusivity (ME) assumption, children assume that each object has only one label. It is claimed that ME helps guide children's initial hypotheses about the meanings of new words, thus helping to explain why children acquire words as rapidly as they do. Characterised as a default assumption, ME is a first guess on the probable meaning of a new word, but it can be overridden. Evidence that it has been overridden is the presence of dual labels. Bilingual data are particularly useful because they have the potential to present numerous cross-language equivalents, which ME predicts young bilinguals should avoid for the first 50 to 150 words.

This study tested the ME hypothesis using both naturalistic and experimental data from children at the one-word stage acquiring two languages from birth. The study comprised a case study and an experimental study. The case study investigated the extent that ME was overridden in one child by investigating the number of dual labels found. A child acquiring English and Italian was observed from the time he could produce a few words (1;3.1) to just beyond his second birthday (2;0.19). Data were collected in the form of a parental diary record of the words he understood and produced and weekly videotaped sessions of him interacting with his Italian-speaking mother and English-speaking father. The experimental study tested the hypothesis that children can learn second labels for objects by teaching six infant bilinguals first one word in one of their languages and then the equivalent label in the other language.

 

The findings contribute to the debate on whether ME is available to children under 2 1/2 years old by showing that ME was overridden to such an extent in the case study as to question its usefulness as a word-learning principle. Furthermore, the children in the experimental study, contrary to ME"s prediction, were able to learn second labels for objects. The findings also contribute to the debate on the place of word-learning principles in lexical acquisition by offering an alternative view of how children may learn words.

 

Processing Embedding in Humans and Connectionist Models

Ngoni Chipere (1999)

 

The aim of the research was to determine whether individual differences in the ability to comprehend certain multiply embedded syntactic structures arise from individual differences in memory capacity or from individual differences in syntactic competence. This question was addressed through a literature review and three experimental studies. Data from the experiments was then used to assess selected Classical and Connectionist models of syntactic embedding. The literature review indicates that there is insufficient evidence to decide between the two explanations of individual differences in comprehension. The experiments addressed the research question by: a) testing subjects other than the customary university students; b) using structures other than the usual relative clause structures and c) testing subjects before and after they had been provided with memory and comprehension training. In Experiment 1, graduate native, graduate non-native and non-graduate native speakers of English were tested on complex NP, Tough Movement and Parasitic gap structures (e.g. in order, ‘Tom knows that the fact that stealing clothes openly is stupid amuses the boy’; ‘The bank manager will be difficult to get the convict to give a loan to’ and ‘The man who Peter saw after overhearing his girlfriend planning to jilt took the 11 o’clock train’). The non-native graduates obtained the highest comprehension scores, followed by the native graduates. This result was explained in terms of effects of education on comprehension, whereby the explicit grammatical instruction given to graduate non-natives while learning English gave them an advantage over the native speakers. In Experiment 2, post GCSE students of high and low academic ability were tested on comprehension and recall of complex NP sentences (example given above). High academic ability students obtained higher scores in comprehension and recall whereas low academic ability students obtained poor recall scores and generally failed to comprehend the sentences. The low academic ability students were then given memory training. Despite achieving high levels of sentence retention, they still could not comprehend the sentences. Comprehension improved only after comprehension training. It was concluded that the initial failure to comprehend the sentences arose from insufficient syntactic competence, rather than insufficient memory capacity. In Experiment 2, a similar low academic ability group was given comprehension training only. Subsequently, they understood and recalled the test sentences as well as the high academic ability students. It was concluded that the initial failure to recall the sentences was a result of poor comprehension. Classical and Connectionist models of sentence processing were then considered. Classical models assume that language users have a generative grammar which is constrained by a limited working memory capacity. Connectionist models assume that language users induce linguistic structure from experience and comprehend novel sentences through a process of constraint satisfaction. The ability of these two types of model to explain the results was then evaluated. The purpose of this evaluation was to produce a theoretically satisfactory explanation of individual differences in the ability to comprehend complex structures. Such an explanation has consequences for the teaching of language to both native and non-natives speakers. In the Conclusion, general implications of the findings for sentence processing models and for native language teaching were considered.

 

 

The involvement of working memory in reading in a foreign language

Catherine Walter (2000)

 

The L2 threshold effect. In the area of second-language (L2) reading, the L2 reading threshold effect is a well-documented phenomenon: L2 learners who are skilled comprehenders in their first language do not transfer their L1 reading comprehension skills to L2 until they reach a certain level of L2 proficiency (Alderson 1984, Clarke 1988, Markham 1985, Bossers 1991, Carrell 1991, Bernhardt and Kamil 1995). Even though this effect has been amply attested, English language course books continue to contain, and teaching methodology manuals continue to encourage, work on reading comprehension skills at lower intermediate levels (Harmer 1991, pp. 184-211; Nunan 1991, pp. 63-82; McDonough and Shaw 1993, pp. 101-125). Language classroom time is spent working explicitly on reading comprehension skills which many or most learners already possess in L1, and which they will transfer to L2 unproblematically once they have crossed the L2 reading threshold.

The present research had two main aims. The first was to examine the L2 reading threshold more closely. Previous studies had looked at learners' reading comprehension performance in L2 without reference to their reading comprehension performance in L1. In the present series of experiments, each task was performed both in L1 (French) and in L2 (English) by two groups of learners: one at a lower-intermediate L2 proficiency level and the other at an upper-intermediate level. Another lacuna in previous studies had to do with the locus of the reading threshold effect: it was not clear what was failing to be transferred from L1. The present research tested the hypothesis that it is the building of mental representations of text that fails to be transferred in the L2 reading threshold effect.

Working memory. The second main aim of the present research was to test the hypothesis that the reading threshold effect is linked to the development of L2 working memory (WM), and to examine the possibility that this might be due in part to inadequate L2 phonological representations. As in the reading comprehension experiments, all WM and phonological short-term memory tasks in the present series were performed both in L1 and in L2. L1-only studies have shown that low-WM individuals have difficulties in comprehension, and that these difficulties are linked to the unreliability of their mental representations (Yuill and Oakhill 1991; Swanson and Berninger 1995; Cornoldi, De Beni and Pazzaglia 1996); and the reading performance of advanced L2 learners has been shown to correlate with WM (Harrington 1992; Harrington and Sawyer 1992). If L2 learners with adequate L1 comprehension skills who do not transfer these skills to L2 also have much lower WM in L2 than in L1, and if this differentiates them from learners who do transfer their reading comprehension skills to L2, this suggests a link between WM and the threshold effect.

Cross-disciplinary focus. The present research includes elements of both cognitive psychology and second language acquisition. Therefore, rather than begin with an exhaustive literature review of the different fields to be covered, this dissertation is divided into two parts, each part comprising three experiments:

-                Part 1: Working memory and phonological short-term memory in L1 and L2

-                Part 2: The L2 reading threshold and working memory

Each part begins a general background to the field in question, and each experiment carries its own introduction with a review of the relevant literature. As has been indicated, the L2 reading threshold was the starting point for this research; however, the working memory experiments are reported first, so that the working memory implications of each reading comprehension experiment can be explored as the results are reported.

 

 

Bimodal input, word recognition, and memory

Stephen Bird (2000)

 

This dissertation explored some possible effects of ‘same-language’ subtitled film/video as a foreign language learning tool. A number of studies have shown that same-language subtitling can be beneficial to language learners in terms of overall plot comprehension and word meaning. However, critics argue that same-language subtitling distracts students from the auditory channel, creates a dependency on text, and adversely affects spoken word form learning. In the present dissertation, experiments were designed to measure the effects of single modality and bimodal inputs (text with sound) on spoken word recognition and memory. Six experiments are reported. Subjects first performed training tasks that included target words and/or nonwords in various modalities (sound-only, text-only, and sound with text). Subjects were then given implicit and explicit memory tests. The overall results were the following (1) On the implicit tests, the repetition priming effects for reaction times to words were equivalent in single modality and bimodal conditions; (2) Nonword reaction time priming was absent in all conditions for all experiments except one, which revealed priming for the text-only and bimodal conditions when targets were repeated three times prior to test and subjects were required by the training task to convert print to phonological information; (3) On two experiments’ implicit tests, the bimodal condition showed reduced errors for words and nonwords (4) On explicit tests, the bimodal condition showed significantly higher scores than the two other presentation conditions for words and nonwords. Overall, the results suggest that simultaneous bimodal inputs can improve some implicit and explicit aspects of spoken word form learning without any apparent costs. The results are discussed in terms of implications for same-language subtitling, and models of word recognition and memory.

 

 

An enquiry into cross-language differences in lexical-conceptual relationships and their effect on L2 lexical processing

Kerrie Elston-Güttler (2000)

 

This dissertation presents an experimental psycholinguistic study on L2 lexical processing with the aim to incorporate cross-language lexical-conceptual differences (i.e. instances where equivalence across the L1 German and the L2 English is not one to one) into existing processing models of the L2 mental lexicon. All experiments are in English and use native-German advanced learners of English and a control group of English native speakers. In Chapter 1, a taxonomy of cross-language differences is presented, levels or representation needed in modelling L2 lexical processing are discussed, and it is argued that existing models of the L2 or ‘bilingual’ lexicon have neglected cross-language lexical differences.

Chapter 2 addresses the processing of intra- and inter-language lexical ambiguity with homonyms and ‘false friends’ with three semantic priming experiments (Exps 1, 2A and 2B). The experiments support a distinction between L1-L2 orthographic form links and L1-L2 translational links not controlled by L2 language context. The data show that L1 meanings of ‘false friends’ e.g. chef (= German {boss}) are not primed in L2 sentence context but they are primed out of context, while the L2 translations e.g. to clap and to gossip of an L1 homonym klatschen prime one another both in and out of sentence context.

Chapter 3 explores polysemy in the L1 and its effect on L2 lexical processing with an acceptability judgement task (Exp 3) using words such as bag and pocket that both translate as the polysemous German word Tasche. Results showed that German natives are more likely than natives to make errors in distinguishing between the English words in sentence contexts. The error effects are argues to be caused by activation of a non-specific ‘L1’ lexical concept. Chapter 4 addresses subtler L2 distinctions e.g. shade and shadow not lexicalized in German e.g. Schatten with an online reading task (Exp 4) and a true-or-false task (Exp 5). Native German speakers appear to rely on L1-type non-specific representations when an explicit judgement demanding the exact interpretation of e.g. shade is required, but in the online reading task, no evidence of this was obtained.

Chapter 5 addresses ‘general’ verbs such as make (German machen) that differ across languages in terms of collocational usage. A phrasal acceptability task (Exp 6) suggests that via the L2-L1 translational link, make in *make a party can activate a lexical network surrounding the L1 word machen, causing incorrect judgements of acceptability on the part of Germans. Each respective cross-language difference sheds light on different modelling issues (i.e. levels and types of representation and degree and nature of L1-L2 interaction), so in Chapter 6, the empirical findings are used to explicate aspects of modelling L2 lexical processing that pertain to cross-language lexical differences.

 

 

Working memory and second language comprehension

Peter Möbius (2000)

 

The thesis presents an experimental psycholinguistic investigation of three aspects of verbal working memory (VWM) in second language comprehension that have received little attention in the literature: a) The study addresses the question to what extent the structural similarity or dissimilarity between L1 and L2 determines working memory capacity in L2 (as measured by a variant of Daneman & Carpenter's (1980) reading span test) by comparing groups of English L2 speakers with structurally diverse L1s (Mandarin Chinese and German) and native controls. b) Building on Perfetti's (1985) efficiency theory of reading and Caplan & Waters' (1999) modularity theory of VWM, the study seeks to identify L1-specific processing deficits that account for between-groups differences in L2-VWM and comprehension. c) The thesis assesses whether psycholinguistic measures of the efficiency of specific processes (e.g. low-level word recognition) provide an alternative to reading spans as a diagnostic of L2 reading skill.

Ninety six subjects participated in a battery of psycholinguistic experiments and conventional comprehension tests, which revealed the following: a) L1 contributed significantly to between-group differences in VWM and comprehension when confounding factors were accounted for. b) There was compelling evidence for modularity of native VWM, whereas the non-native groups displayed a pattern of limited modularity/automaticity, mainly due to processing deficits at the "interpretive level" (Caplan & Waters, 1999). Chinese deficits in low-level word recognition and in semantic processing were identified as primary sources of between-group differences in general VWM and comprehension skill. c) Within groups, measures of the efficiency of interpretive processing predicted individual comprehension skill, whereas reading spans did not. The optimal sets of predictor variables were different for the L2 groups.

The limitations of the Daneman & Carpenter-type indices as diagnostics of L2 comprehension skill, the implications of the experimental results for a model of working memory in L2 reading and secondary findings are discussed. It is proposed that psycholinguistic profiles of the efficiency of discernible processing components in L2 reading may offer valuable information for the design of L1-specific second language instruction, and that they can be used to assess the success of such instruction in overcoming L1-specific "processing bottlenecks".

 

The representation of syntactic information in the bilingual lexicon

Angeliki Salamoura (2005)

 

This dissertation reports an experimental psycholinguistic investigation of the representation of syntactic information in the bilingual mental lexicon. It investigates cross-language processing of syntactic properties of words at the phrase level and provides evidence that bears on two interrelated issues: (i) the shared or independent nature of equivalent L1 and L2 syntactic properties of words in the bilingual lexicon, and (ii) the factors (e.g., morphological, constituent order or thematic information) that influence processing and representation of these syntactic properties. It is argued that existing bilingual models have left these issues unexplored.

The study examines syntactic information pertaining both to nouns and verbs. In relation to nouns, the focus is on the processing of grammatical gender in cases where it does not coincide with natural gender. Concerning verbs, the syntactic attribute tested is the argument structure of ditransitive verbs (e.g., give + NP NP vs. NP PP).  To explore the above questions I employ within- and cross-language tasks that tap into syntactic representation and processing, such as picture-word interference (Schriefers, 1993) for investigating effects of gender, and oral sentence completion paradigms for investigating syntactic priming (Branigan and Pickering, 1998). The pair of languages used in the experiments is Greek (L1) - English (L2) in the case of verb argument structure and Greek (L1) – German (L2) in the case of grammatical gender. The participants are Greek advanced learners of English or German respectively.

The results demonstrate cross-language priming both for grammatical gender and verb argument structure and hence suggest a shared representation for equivalent L1 and L2 syntactic information pertaining to words. More specifically, priming of grammatical gender was not affected by morphological information such as the gender transparency or opaqueness of the suffix of the noun stimuli. On the other hand, priming of verb argument structure was influenced by thematic role information and was also lexically activated (i.e., triggered by a single verb). Finally, a model of the representation of syntactic information in the bilingual mental lexicon that integrates these findings is outlined.