Holistic View Of The Bilingual Person English Language Essay

The term ‘bilingual’ in the psycholinguistic literature does not only apply to people who speak two languages equally well because they were exposed constantly to two different languages maybe due to their parents two different native languages. However “Bilingualism” refers to “the regular use of two (or more) languages, and bilinguals are those people who need and use two (or more) languages in their everyday lives.” (Grosjean, 1992, pp. 51).

This represents a holistic view of the bilingual person as a competent and complete communicator, on the other hand though a bilingual person is surely not the result of the sum of two monolinguals.

As early as 1968, Macnamara, Krauthammer and Bolgar wrote: “Within certain limits …, all bilinguals manage to keep their languages distinct and can switch from one to the other. It follows that to some extent bilinguals experience their languages as psychologically distinct systems, and that they have some device to control which one is used at any particular time”. However, fluent bilinguals are capable of ‘switching’ between their two languages, when for example a third person who can speak only one of the two languages is pulled in a conversation or when the topic of the discussion strongly asks for the use of both languages at the same time.

To clarify this point, consider this situation described by Judith Kroll “You are sitting at a café or at the airport when you overhear a conversation in English that suddenly switches to another language and then back to English. If you are a monolingual speaker of English, you may notice the mixture of languages without realizing that you have listened to an impressive cognitive accomplishment by the speaker”. This exceptional achievement is instead a rather common feature of bilinguals language use in which words of two languages mix together in a coherent and meaningful conversation .. In this sense, a bilingual changes the linguistic ‘form’, without alterating the substantial’ meaning” byusing a word which may address the sense of a discourse in a ‘better’ or ‘stronger’ way like choosing between synonyms with the same language (Sridhar & Sridhar, 1980). Yet when the same bilinguals speak to a monolingual they rarely use or ‘switch’ to an alternative language in order to prevent the monolingual speaker from not understanding. These different circumstances and a variety of other situation where this capability arises leads to question of how the information to be processed or expressed is bound to the activation or articulation of a corresponding word or phrase in the appropriate language.

On one hand, for a person repeatedly ‘coping’ with language switches within the conversation, these apparently ‘strange’ words come unexpectedly and may perhaps be more difficult to process than their within-language counterparts.

On the other hand for a bilingual who has to ‘choose’ in which language to speak, the process of ‘finding the right word in the right context’, which French and Jacquet (2004) refer to as ‘lexical access’ may result extremely complex , as in addition to the activation of words in one language other than the target, other ‘parole’ (words) in the other language might be active as well.,. Thus, the simultaneous activation of the two lexico-semantic represenations of a bilingual might address different answers paralleled with the specific processing modality, word recognition or production, driven by the context.

In word recognition, language membership is passively conveyed to a person by the orthographical or phonological characteristics of the word (). However, in word production, the speaker actively and intentionally decides which language to use. Therefore, the speaker can exert some ‘control’ on lexical forms and choose the target which best fits the communicative context among a set of activated representations. We do not claim that the mechanisms and neural dynamics recruited for lexical access are necessarily different in recognition and production, but rather that the processes involved in each may be at least partially different.

The aim of this project will be first to trace the effects of a language switch on models of both language production and language comprehension and second to identify the neural correlates of language switching and the impact a switch may have on the cognitive processes which rule lexical access in order to produce or recognize a word.

Bilingualism and language comprehension

Language comprehension has been investigated in bilingual populations mainly through tasks in which bilinguals are substantially asked to respond to written words in one or both of their languages. In such visual word identification tasks, the language switch is driven by the upcoming stimuli in input, while the output is executed by button press driven by a binary decision.

A large number of studies have addressed bilinguals’ performance in comprehension tasks through both within-language and cross-language tasks such as lexical decision (e.g., Dijkstra, Van Jaarsveld & Ten Brinke, 1998; Dijkstra, Grainger, & Van Heuven, 1999; von Studnitz & Green, 2002), language decision, and categorisation tasks (e.g., Dufour & Kroll, 1995; Grainger & Frenck-Mestre, 1998).

Initial studies revealed, for example, that when bilinguals were asked to read language-mixed passages, their performance suffered compared to reading single-language passages (Macnamara & Kushnir, 1971). In lexical decision, responses to words where a switch in language occurs were slower than those to a trial nested in a sequence of words from same language s (Thomas & Allport, 1995; Von Studnitz & Green, 1997). Ability to recognize words in one language seems to be influenced by the language memebership of the word immediately preceding (the ”basic language priming” effect) (Grainger & Beauvillain, 1988; Grainger & O’Regan, 1992) even in lists of unrelated words. Fluent bilinguals seem to comfortably manage whichever language they are requested to use, however in all of the contexts mentioned just above a language switch during comprehension hurts their performance. This evidence suggests that even when bilinguals read (e.g., Dijkstra, 2005) or hear (e.g., Marian & Spivey, 2003) one language alone, both languages are still active. Thus, a crucial point here is to establish if and to what extent the other language ‘is still there’ when bilinguals use one language alone. One way of testing this hypotheisis is to isolate ambiguous features of the bilingual’s two languages , meaning to use words that partially overlap or are totally shared in both languages. When two languages share the same alphabet, we may find words called cognates that look or sound the same and mean the same thing as well. For example, In French and Italian, the words balla and balle are almost spelled identically and have the same meaning and. If bilinguals are really capable of shutting down one language and dress as monolinguals, then performance on these special words (cognates) should not differ from that on distinctive and unambiguous words. If the other language results not to be in standby but always ‘on’, then bilinguals should perform differently from monolinguals which in a lexical decision task will need to match the target with only one possible candidate instead of two… A cognate ‘benefit’ on performance has been demonstrated across a variety of tasks (De Groot and Poot, 1997; Van Hell & De Groot, 1998a; Van Hell and Dijkstra, 2002;), providing substantial evidence that cognates are represented or processed differently from non-cognate translation equivalent words in the second language. Cognates and non-cognates also show different priming effects: in one of the earliest explorations of the effects of cognates, De Groot and Nas (1991) found cross-language repetition priming for both cognates and non-cognates, but associative priming only for cognates. Given such evidence they reached three conclusions: (1) the representations of both cognate and non-cognate translations at the lexical level of representation are connected; (2) cognate translations share a representation at the conceptual level while (3) non-cognate translation equivalents are represented in separate concept nodes. De Groot and colleagues’ model of cognate representation has continued to develop, but it remains firmly based on the principle that cognates’ representations in the two languages are shared, or overlapping, more than those of non-cognates. In terms of distributed representations, Van Hell and De Groot (1998) describe the notion of overlap as ‘the patterns of activation for a cognate word and its translation being similar to one another, whereas the patterns of activation for a non-cognate word and its translation may have very little in common. The more features are shared between words, the smaller the “lexical distance” between their corresponding patterns of activation.’

In addition, the cognate effect was found not to be restricted only to conditions where stimuli are presented in written form. Costa, Caramazza, and Sebastián-Gallés (2000), for example, found that bilinguals named pictures with cognate names more quickly than pictures with non-cognate names, while monolinguals showed no difference on the same set of pictures. This confirms that the cognate benefit is not solely due to orthographic overlap in the presented stimuli.

Many studies have took advantage of these special properties of cognate words in order to determine how this linguistic ambiguity impacts on bilinguals’ ability to understand these words in only one of their two languages. Evidence stemming from all these studies strongly supports the idea that the language not in use may be in a sort of ‘sleeping mode’ and anyway exerts an influence on the bilinguals’ lexico-semantic system even when a task ‘tunes’ it to the other language. ‘When cross-language form and meaning converge, bilingual performance is typically facilitated; when cross-language form and meaning conflict, bilingual performance is often hindered, in that it is slower and more likely to be error prone’ (Dijkstra, 2005). These cross-language effects will likely occur especially in the case of a second ‘less dominant’ language given that most of time both languages will never be equally strong.

Furthermore in conditions where a change in language occurs, the cross-linguistic influence of one language on the other will ‘directly’ affect the processing of words in either one of the two languages. However it is a point of some controversy in the literature whether the ‘costs’ associated with switching between languages might be somehow modulated by language specific or ambiguous cues.

The Bilingual Interactive Activation model (BIA) and language switching

Dijkstra and van Heuven (1998) have proposed a model for word recognition in bilinguals (BIA, the Bilingual Interactive Activation model) in which they try to account for the ‘interaction’ between active word candidates in both languages. Novel to the BIA model is the use of language nodes. When the BIA model encounters a string of letters, the specific visual features of each at a particular letter position excite letters in the system with corresponding features while different letters are inhibited . Activation in turn from letters is driven to words in both languages where each letter figures in the determined position, while all other words are inhibited. At the word level, language membership will not affect inhibition as all words inhibit each other. Activation thriving from word nodes in the same language is carried on to the corresponding language nodes which store activation from words with a specific ‘language tag’, and in turn spread, through a feedback mechanism, inhibition to all word nodes in the other language. Furthermore, these language nodes can be pre-activated reflecting a particular task and this device allows the asymmetric inhibition of words in the two languages; word forms in L1, for instance, can be more inhibited than word forms in L2. The effects of language switching can be explained in this framework through a mechanism which allows lexical activation to flow from one trial over to the next. The BIA hypothesizes that activation of a specific language node paralleled with the presentation of a word in that particular language will not completely decay and fall beneath threshold, therefore when the next item comes up in the other language the corresponding word unit will be partially inhibited. According to this model any cost relative to switching will fall close to zero if the input carries orthographic features unique to a language. Only one or a few word units in that particular language will be active and any advantage or disadvantage held by similar cross-linguistic representations (i.e. as in the case of cognates) of the previous trial will fade out. This model shows that language switching may be a function of the task situation, the nature of stimulus material, as well as the expertise of the bilingual.

Figure 1. The Bilingual Interactive Activation (BIA) model for bilingual word recognition. Arrowheads indicate excitatory connections; black filled circles indicate inhibitory connections. (Dijkstra & van Heuven, 1998)

4.4 The effects of context information and the BIA+ Model

Language is a single word, however in its everyday use it implies the use of a set of multiple words to express meaning. It is possible therefore that evidence for cross-language activity stems from the decontextualized nature of word recognition tasks commonly employed to investigate the bilingual’s two languages. In the context of a conversation or while we read a passage in one language rather than the other cues which shift the balance of activity in favour of the intended language should be conveyed to a mechanism which could ‘virtually’ switch off ‘the other language.

This indeed does not seem to be the case as recent evidence from a number of studies suggest that contextual cues per se are not able to turn completely down the activity of the language not in use. On one hand we would have intuitively predicted that the frame provided by a stringent linguistic context should reduce the number of viable language interpretations. On the other hand, these findings justify the ease of language switching and the relatively ‘low cost’ it entails in terms of processing resources (e.g., Moreno, Federmeier, & Kutas, 2002).

However, a point of some controversy remains and namely the relationship between the word identification system and the linguistic context (as a sentence) or the non-linguistic context information determined in an experimental framework by the task demands (i.e. the participants’ expectations determined by the instructions). One option is that after the initial stages of lexical processing, information of both types (linguistic and context) may exert an influence on the activation level of forms in the target and non-target language. For instance, context information could inhibit lexical candidates or lemmas in the irrelevant language (BIA model by Dijkstra et al., 1998; IC model by Green, 1986, 1998) or just modulate of the activation level of lexical candidates in each language (Grosjean, 1997). A second option is that non-linguistic context information does not ‘directly’ influence the activity in the identification system itself, but affects decision criteria only.. The BIA+ model postulates the existence of two distinct systems: a word identification system and a task/decision system. Linguistic information conveyed by a sequence of words in in a sentence context may modulate the word identification system, while non-linguistic context information (e.g., participants’ expectations and strategies) affects parameter settings in the task/decision system.. However, the model clearly states that the task/decision system and sources of non-linguistic information do not affect the lexical activation levels within the word identification system itself. Therefore while performing in a task (such as lexical decision) an early preconscious, automatic level of processing thriving from activity within the word identification system may be followed by an attention-sensitive level in which lexical forms are selected through a task/decision system with reference to different contextual factors and bound to a specific response relevant to the task at hand (cf. Altenberg and Cairns, 1983, p. 187; Dupoux and Mehler, 1992; Balota, Paul and Spieler, 1999). The task schema, which is set up during the practice set or retrieved from memory, designates the “algorithm” which selects the cognitive processing steps necessary to perform the specific task (Green, 1986, 1998; Norman and Shallice, 1986). The decision mechanism is incorporated in the task schema and monitors continuously the activation level of candidates in the identification system by ‘weighting’ the different levels of activation of targets with respect to each other within the identification system in order to arrive at an output in terms of response. The decision relies upon a lexical selection mechanism, which triggers depending on the breaking of an activation threshold for a lexical candidate. In other words, the identification and task/decision systems, though interconnected, may be partially independent. The two systems use their own criteria for action triggering (i.e., lexical selection and response selection/execution). The identification system is assumed to recognize a word and is able to select a single lexical candidate with a good degree of certainty) when the system reaches a fair stability. The task/decision system triggers a response when its own criteria are met, some of which ruled by lexical activation, while others driven by a tendency towards optimization in terms of how activated and selected representations in the identification system are linked to possible responses. For instance, in lexical decision the input letter string conveys activity to orthographic, semantic and phonological codes, all of which could allow a discrimination of word and non-word input. However, when participants are asked to make a language decision in the sense to press one button if a presented item belongs to one language (e.g., English) and another button if it belongs to another language (e.g., Dutch) only those codes which facilitate the retrieval of language membership information (language tags) are able to address a correct response. Thus, different schemas underlie different tasks, although one task may obey to different schemas. The schema might capture and use information from different sources in parallel, but presently available evidence suggests that orthographic representations play a major role (Pexman and Lupker, 2001). A number of recent experiments have addressed the predictions stemming from the BIA+ model by asking whether the parallel activity of the two languages can be reduced or eliminated when language ambiguous words that produce cross-language effects out of context, are placed in sentence context (e.g., Elston-Güttler, Gunter, & Kotz, 2005; Schwartz & Kroll, 2006;Van Hell,1998).

Schwartz, Kroll, and Diaz (2007) showed that when bilinguals are asked to name a cognate like radio in isolation, they are faster relative to controls if there is both orthographical and phonological overlap across the two languages. However, when they read highly constrained sentences the processing advantage for cognates disappeared while in sentences with a lower closure probability, an advantage for cognates remained, suggesting that knowing the language in which you are reading does not switch off the unintended language.

This last assumption leads to the question of whether the decision criteria in a language switching task is affected when cognates are involved considering that the activation threshold for lexical candidates will be broken not as quickly. According to the BIA+ model, the similarity of the input word to the internal lexical representations establishes their activation level. Therefore the larger the overlap between the input string and a representation in the mental lexicon, the more the internal representation is activated. In the case of two languages with alphabetical writing systems, the number of activated orthographic candidates is determined by factors such as the neighbourhood density and frequency of the target word and its within- and between-language neighbours and not by the word’s language membership. However, If the two input codes specific to each language are different (e.g., letter sets), the activated set of neighbours may become much smaller.

Figure 2. The BIA+ model for bilingual word recognition.

Arrows indicate activation flows between representational pools. Inhibitory connections within pools are omitted. Language nodes could instead be attached to lemma representations between word form and meaning representations. Non-linguistic context only affects the task schema level.

(Dijkstra & Van Heuven – The architecture of the bilingual word recognition sysytem, Bilingualism:Language and Cognition, 5, 2002 )