Bilingual Lexical and Conceptual Memory Representation

University Degree Linguistics, Classics and related subjects

Bilingual Lexical and Conceptual Memory Representation: A Critical Review

Abstract

The major purpose of this review is to critically evaluate the asymmetry model (Kroll & Stewart, 1994) concerning whether word translation is based on a lexical level or is achieved through conceptual mediation. This model assumes that forward translation (from the first language to the second language) is mainly conceptually mediated, whereas backward translation (from the second language to the first language) is comprehended through direct lexical path. Variables that condition conceptual/semantic processing will therefore have a larger effect on forward than backward translation. This paper first describes the main features derived from the asymmetry model with an overview of various evidence-based sources. It then considers the generalizability of the model with a critical appraisal of controversial evidences due to different experimental paradigms and semantic variations. Further, it illustrates semantic variations in word translation process by comparing and contrasting performances of bilinguals in different orthographic systems, namely Chinese and English. Finally, the paper proposes other experimental paradigms for testing Chinese-English bilingual lexical and conceptual representations.

Key words: word translation, lexical activation, conceptual mediation, bilingual memory representation, asymmetry model, Chinese-English bilinguals

The issue on how the bilinguals’ lexicons and the underlying conceptual memory are organised has always been very important in bilingual research. For individuals who acquire a second language relatively late in life, it is generally believed that they access word meaning in their first language (L1) through direct activation of relevant concepts, however, the word meaning in their second language (L2) is comprehended by first being translated into its L1 equivalents and then understood via the activated L1 concept.

Kroll and Stewart (1994) formulated this idea and proposed the asymmetry model as a theoretical framework for the study of the bilingual memory representation. The model assumes two different pathways for the communications between L1 and L2. The L1 lexicon is closely tied to an underlying conceptual memory, whereas the L2 items are mostly associated with their L1 (translation) equivalents. The outcome of this architecture further suggests the differences in translations between L1 and L2. When an item is translated from an L1 to an L2 (forward translation), translation is mediated by the conceptual memory, which means that the underlying concept is first activated by its L1 label and then recoded into its L2 name. However, when an item is translated from an L2 to an L1 (backward translation), the L2 label is directly coded into its corresponding L1 name at a lexical level and little concept activation is involved here. Taken together, the hypothetical route for forward translation regards concept activation as a necessary mediating step for L1-to-L2 communications, whereas the route for backward translation exemplifies direct lexical associations between the two lexicons and requires no mediation through concepts for L2-to-L1 communication. The bilingual memory representation prescribed by this model is, therefore, asymmetrical in the sense that qualitatively different routes are taken for the two types of translation. The basic architecture of the model is shown in Figure 1.

As depicted in Figure 1, L1 is seen as being larger than L2 because it is assumed that bilinguals have a larger vocabulary in their native language than in their second language. The model identifies lexical-conceptual associations at three different levels of strengths. The thick solid arrow that points both ways indicates a very strong link between conceptual memory and the L1 lexicon. Forward translation is accomplished by first going through the strong L1-to-concept route and then the weak concept-to-L2 path represented by the dashed arrow. In principle, forward translation can also be achieved at a lexical level through a weak L1-to-L2 link. The L2-to-L1 thin solid arrow (of less strength) represents the prominent lexical communication in backward translation.

Figure 1: The basic architecture of the asymmetry model of bilingual memory. Adapted from ‘Category Interference in Translation and Picture Naming: Evidence for Asymmetric Connections between Bilingual Memory Representation’ by J. F. Kroll and E. Stewart, 1994, Journal of Memory & Language, 33, pp.158.

Main Features of the Asymmetry Model – An Overview of the Literature Background

One of the main features described by the asymmetry model demonstrates differential processing of L1 and L2 lexical items and the former appear to be more closely linked to the concepts they represent than the latter. Therefore, translation from L1 to L2 should be more likely to engage conceptual processing than translation from L2 to L1 because L1 can more readily activate concepts than L2 can. Likewise, L2, by virtue of its initial lexical-level connections to L1, should allow rapid lexical-level translation from L2 to L1 without recourse meaning. Thus, L2-to-L1 translation should be faster than L1-to-L2 translation and also less sensitive to the effects of semantic factors.

In one study of the impact of semantic variables on the translation processing, Kroll and Stewart (1994) had relatively proficient Dutch-English bilinguals translate words from L1 to L2 and from L2 to L1 in the context of semantically categorized or semantically mixed (i.e., randomized) lists. Words were translated more rapidly and accurately from L2 to L1 than from L1 to L2. Semantically organized word lists were found to have a negative effect on translation latencies only in forward, whereas backward translation was unaffected by the manipulation of semantic categorization. These findings provided initial support for the claim that only forward translation might have involved concept activation. Other translation studies have reported a similar asymmetry (e.g., Sanchez-Casas et al., 1992).

Using a transfer paradigm to capture the relationship between picture naming and translation, Sholl, Sankaranarayanan, & Kroll, (1995) provided most compelling evidence supporting the prediction that forward translation is conceptually mediated but backward translation is not. If only translation from L1 to L2 is conceptually mediated, then only forward translation should benefit from prior study during which concepts are names as picture (picture-naming was believed to be conceptually mediated by researches such as (Potter, So, Voneckardt, & Feldman, 1984). This is precisely the result reported by Sholl et al. They found that picture naming facilitated forward but not backward translation when concepts had been named previously as pictures in L2 or L1. Therefore it was assumed that transfer between picture naming and word translation can only take place at a conceptual level, suggesting that minimal concept activation is required in backward translation.

Another main feature derived from the model has to do with the fact that the translation asymmetry is dependent on the bilinguals’ second language proficiency (Kroll, 1993). In early stages of second language learning, less-fluent individuals show a large asymmetry between forward and backward translation. Specifically, forward translation is especially delayed in latency and prone to errors. However, the more fluent bilinguals are faster and more accurate in forward translation. According to the model, as bilinguals become more proficient in their second language, the connection from L2 words to concepts will strengthen, resulting in an increase in the ability to use conceptual activation to direct retrieval of words from the L2 lexicon. A number of studies have examined this developmental prediction of the asymmetry model.

Using a translation recognition task (Degroot, 1992), where a pair of words is presented and subjects must decide whether two words are translations to each other, Talamas, Kroll and Dufour (1995) provided strong evidence that more-fluent bilinguals are able to conceptually mediate the L2. The conditions of interest in the study were the trials on which the two words were not translation equivalents. These ‘no’ trials included cases in which the incorrect translation was similar in form to the correct translation (e.g., man-hambre [hunger] instead of man-hombre [man]), similar in meaning (e.g., man-mujer [woman] instead of man-hombre [man]), or unrelated. Talamas et al. found evidence for interference (i.e., longer response latencies) for both form-related and meaning-related trials relative to unrelated controls. However, the magnitude of the form and meaning interference was a function of fluency in L2. Relatively more fluent English-Spanish bilinguals produced a large meaning interference effect but a small form interference effect. The opposite was true for less-fluent bilinguals, whose performance was more disrupted by similarity in form than in meaning. This pattern is consistent with the claim that less-fluent bilinguals rely initially on lexical form and only later are able to access to meaning with increasing proficiency in the L2. This notion is also supported by the findings of some other studies that also take a developmental perspective (see, e.g., Kroll & Curely, 1988; Chen & Leung, 1989; Dufour and Kroll, 1995; and Kroll, Michael, Tokowicz, & Dufour, 2002).

In the same vein, Sunderman & Kroll (2006) very recently also used a translation recognition task to place predictions of the asymmetry model in the context to investigate lexical processing in a L2. The critical items in their study were non-translation equivalents that were similar to the correct translation in either form or meaning. For instance, for translation equivalents such as cara-face, critical distracters included (a) a form-related neighbour to the first word of the pair (e.g., cara-card), (b) a form-related neighbour to the second word of the pair (e.g., cara-fact), or (c) a meaning-related word (cara-head). It was found that all the L2 learners, regardless of proficiency, experienced interference for meaning-related pairs. However, only the less-proficient learners also showed effects of form-related pairs on translation recognition performance.

Take together, two main features of the asymmetry model assume that L1-to-L2 (forward) translation is mediated by a conceptual memory, whereas L2-to-L1 (backward) translation takes a direct lexical path, therefore forward translation is slower than backward translation. It also suggests that more-fluent bilinguals can effectively access lexical and conceptual connections between their two languages. Less-fluent bilinguals, however, appear to rely heavily on lexical associations from L2-to-L1.

Is the Asymmetry Model Generalizable? – A Critical Appraisal

To evaluate whether the asymmetry model is universally applicable, we need to firstly look at the languages that have been ...

This is a preview of the whole essay

Is the Asymmetry Model Generalizable? – A Critical Appraisal

To evaluate whether the asymmetry model is universally applicable, we need to firstly look at the languages that have been used for the initial development of the model. Most previous studies in the area have examined languages belonging to the same family, namely the Indo-European language family. Specifically, in the studies mentioned above, Kroll and Curley (1988) used native English speakers who had learned German; Kroll and Stewart (1994) looked at English and Dutch, Dufour and Kroll (1995) employed native English speakers who varied in their proficiency in French, whereas Sholl et al. (1995) and Sunderman & Kroll (1999) recruited English-Spanish bilinguals. Some researchers such as de Groot and Hoeks (1995) even employed Dutch-English-French trilingual individuals. These languages are all closely related in terms of some lexical level similarities. Hence, one cannot apply the conclusions generated in these studies to bilinguals speaking two very distinctive languages, such as Chinese and English. It has been documented that the Chinese language has a number of unique characteristics that clearly distinguish itself from English in terms of cognitive processing. Experimentally, for instance, it has been demonstrated that lexical decision is generally faster than naming in Chinese, whereas the reverse is true in English (e.g., Chen, 1992). Later, Chen (1996) suggested the striking differences between the two orthographies to account for such discrepancy.

A wide range of studies on bilinguals who speak Indo-European languages, however, stress the similarity rather than the difference between the two translation directions and demonstrate predominant conceptual processing for both, which sits odds at what the asymmetry model predicts.

First, when presenting two words subsequently to the participants and asking them to decide whether the second word was the correct translation of the first word (i.e., translation-recognition task), de Groot and Comjis (1995) found that the order in which the translations were presented, Dutch word or English word first (i.e., forward and backward translation respectively), had no effect on the translation recognition performance. This indicates that backward translation involves conceptual levels of processing just as much as that in forward translation (also see Potter et al., 1984; Snodgrass, 1993).

Second, La Heij, de Bruyn, et al. (1990) examined the effect of L1 distracter words on backward translation latencies in a Stroop-like task and the evidence reported might be seen as more damaging to the asymmetry model. They replaced the to-be-translated words by Dutch words that were semantically related or unrelated to the correct response words. For example, the word ‘SPOON’ was followed by the incorrect, but semantically related, Dutch word ‘VORK’ (folk) or by the unrelated word ‘GEIT’ (goat). According to the asymmetry model, no Stroop interference should result from a semantically related L1 distracter word if L2-to-L1 translation is lexically mediated. However, La Heiji et al.’s results showed that the semantically related distracter words resulted in a significant increase in word-translation latencies in comparison with the unrelated words in both translation directions. Mazibuko (1991) and La Heij, Hooglander et al. (1996) later also replicated this finding for Spanish-English and Dutch-English translation respectively. They also reported evidence that appears directly contrary to the predictions of the asymmetry model in that either no translation asymmetry was observed, or an asymmetry occurred in the opposite direction, with faster translation from L1-to-L2 than L2-to-L1 (see, also de Groot & Poot, 1997). It is therefore concluded that qualitatively similar conceptual processing are involved in both forward and backward translation.

Third, de Groot, Dannenburg, and van Hell (1994) examined the effects of various word characteristic variables on both forward and backward translation and challenged the view that translation from L2 to L1 (backward translation) is accomplished via a lexical route. Meaning (semantic) variables, such as imageability (i.e., the easiness of generating concrete mental pictures/images when a certain test word is encountered), context availability (i.e., the easiness of coming up with specific contexts in which what the test word denoted tended to appear), and definition accuracy (i.e., how easily and accurately a test word can be verbally defined) associated with both the stimulus and response words, were found to be correlated with both forward and backward translation latencies. They concluded that although semantic variables play a slightly more important role in forward than in backward translation, the strong version of the asymmetry model, which claims that conceptual processing is never involved in backward translation, should be rejected.

It has also been shown in a number of previous studies that the L2 lexicon has prominent, direct associations with the conceptual memory, just as does the L1 lexicon (see, e.g., Chen, 1990; Keatley, Spinks & de Gekder, 1994). The question, hence, is not whether certain translation pathways exist; rather, it becomes under what conditions the activation of a particular set of routes is especially encouraged and in what situation it is not.

As a result, further development in the conceptualization of the asymmetry model has added on to the prototypical version some extra dimensions that actually condition bilingual lexical/conceptual processing. The general approach to the issue of whether word translation is achieved on the basis of concept mediation or on the basis of direct lexical activation is rather straightforward. Under the assumption that the lexicon does not contain semantic information (e.g., Collins and Loftus, 1975; Glaser and Glaser, 1989), the presence of semantic effects in word translation provides evidence in favour of the concept mediation account, whereas the absence of semantic effects supports the direct lexical association.

De Groot and her colleagues (1992, 1994, and 1995) used the word concreteness as an index of conceptual mediation. De Groot (1992) firstly demonstrated that fluent bilinguals translate concrete words more quickly than abstract words and suggested that the difference in translation performance for concrete and abstract words reflects a difference in the degree of conceptual feature overlap, with concrete translation equivalents sharing a larger number of conceptual features than abstract translation equivalents. De Groot and Hoeks (1995) then investigated the concreteness effect in translation in Dutch trilinguals who speak English as a relatively fluent second language and French as a non-fluent third language. They reported that Dutch-English-French trilinguals produced the standard concreteness effect in translation when translating from Dutch to English (L1 to L2), but not when translating from Dutch to French (L1 to L3). This indicates that only when individuals are relatively fluent in both languages, would it be possible to demonstrate conceptual mediation. De Groot (1992, 1993) developed her own approach which focuses on the aspects of words (i.e., word-type) that appear to be associated with lexical or conceptual processing. De Groot and her colleagues (1994, 1995) later examined the effects of a host of variables (e.g., aforementioned imageability, context availability, and concreteness) thought to reflect lexical and conceptual levels of processing in bilingual performance in translation recognition and translation production. De Groot and Comijs (1995) explored the translation recognition task, where each trial presents a word pair, one word in L1, the second in L2 and the participant has to decide whether or not the words within a pair are translations of one another. They compared the bilinguals’ performance with that in the translation production task, where on each trial the participant has to vocally come up with the translation of the presented word. They suggested that translation recognition and translation production generally respond to the same manipulation.

Another approach has been taken to determine whether bilinguals conceptually mediating their second language. It asks individuals to perform a task that, by its nature, requires conceptual processing. If they are conceptually mediating L2, it should be possible for them to perform the task. If they are not able to, then performance should be generally poor. However, the way in which an L2 learner or less-fluent bilingual solves the problem of accessing the required information in such a task may also reveal the manner in which the conceptual connections are eventually established. This approach although does not focus on word translation, it brings other dimensions to the evaluation of the asymmetry model.

One way to require conceptual access is to ask participants to categorize exemplars as members of specified semantic categories (e.g., to decide whether peas are a type of vegetables). Semantic categorization has been used extensively to examine the architecture of semantic memory (e.g., Smith et al., 1974; Rosch, 1975; Dufour and Kroll, 1995). Previous researches on semantic categorization in two languages indicates that fluent bilinguals can categorize equally well within and across their two languages (e.g., Caramazza and Brones, 1980; Potter et al., 1984, Sanchez-Case et al., 1992). This evidence has been taken to support the view that fluent bilinguals can conceptually mediate both L1 and L2 lexicons. However, there are a number of methodological problems. One of the most problematic features is the relatively long interval that occurred between the presentation of the category name and target exemplars. For example, Caramazza and Brones (Experiment 1) used a 2-sec interval between the category and target words. During a long inter-stimulus interval, subjects have the opportunity to translate either or both of the target and exemplar words. Thus subjects might demonstrate similar ease in categorizing in both languages, because they may have performed the task in their preferred language via translation. In another experiment in their study, both words were presented simultaneously. The simultaneous presentation of category word and target exemplar may also be problematic, because the simultaneous format may induce decision processes that also take advantage of translation (see Neely, 1990, for analogous arguments about the lexical decision task). Hence, the previous semantic categorization studies are not conclusive about conceptual mediation in bilinguals with different proficiency in their second language.

An additional main tool to condition conceptual mediation in bilinguals is semantic-priming technique, in which the languages of the prime and target are manipulated. Semantic-priming effect has mainly been explored in lexical decisions. It involves presenting bilinguals with one or two letter strings on a computer screen and the bilinguals are then instructed to decide as quickly and as accurately as possible whether or not the letter string composes a real word. This is often done by pressing one key if the letter string is a real word and a second key if it is not a real word (i.e., nonword). The target word, which they need to give lexical decision, is presented in one of their languages (L1 or L2) and the prime is presented in the other (L2 or L1). This allows researches to examine different priming effects for each direction – L1-to-L2 and L2-to-L1. If bilinguals are able to access concepts for L2 words, and, if concepts are shared across their two languages, then semantic priming should be observed within both L1 and L2 (i.e., within-language priming) and also between L1 and L2 languages (i.e., cross-language priming).

The results of a large number of cross-language semantic priming studies generally support the conclusion that fluent or relatively fluent bilinguals are able to conceptually mediate L2 because priming is typically observed both within and between languages (e.g., Chen & Ng, 1989; de Groot & Nas, 1991). However, under a careful control of presentation conditions in the semantic priming paradigm, recent studies have reported some limitations to the extent of cross-language priming, even when bilinguals are highly proficient in L2. Keatley et al. (1994) reported that, in highly proficient bilinguals, there are asymmetries in the magnitude of semantic priming. They found priming only from L1 to L2, but not from L2 to L1. Similar asymmetric priming results have been described by Tzelgov and Eben-Ezra (1992). These observed asymmetries in priming are consistent with the predictions of the asymmetry model due to the fact that L1 words are more likely to activate their respective meanings than L2 words and are thus more effective primes. However, the problem in interpreting the priming results is that, for bilinguals who are dominant in L1 (which is most of the bilingual population), there will be more information available from an L1 prime than from an L2 prime. Take together within the fact that, on average, an L1 target will also be recognized more rapidly than an L2 target because L1 words are functionally more frequent than their L2 translation equivalents. Due to slower recognition for L2 than for L1, it is also more likely that word meaning will have been activated by the time the L2 word is identified. The observed asymmetries in priming, with larger effects from L1 to L2 than from L2 to L1, although consistent with the predictions of the asymmetry model, may also reflect the limitations of the primed lexical decision paradigm for investigating this issue. Furthermore, few of these studies on semantic priming have examined less fluent bilinguals, making it impossible to assess the prediction that less fluent individuals will fail to demonstrate cross-language priming.

One common methodological problem with the above-mentioned translation recognition paradigm and cross-language semantic-priming technique in bilingual research is that the role of processing time is not fully assessed. The stimulus onset asynchrony (SOA) between the presentations of the subsequent words in two languages should be manipulated. One needs to be aware that each individual bilingual vary in their proficiency in L2, it is difficult to make precise assumptions regarding the time course of processing. The concept behind translation recognition and semantic priming is that as a word is presented, automatic access to its meaning results in activation of both that concept and other concepts that are related to it. The use of long interval between the subsequent words is particularly problematic because, in some of these experiments, participants may have been encouraged to translation the first and/or the second word into the same language. A long SOA can also raise problems in a way that the extended time given makes it easier for participants to generate expectancy to a set of related words. Initial research in semantic priming appears to indicate that SOA lengths less than or equal to 300msec were capable of inhibiting any expectancy strategies. However, more recent work conducted by Hutchison et al. (2001) has suggested that strategic priming still occurs at a relatively short SOA of 300msec. Their data indicated that when the SOA length was decreased to 167msec, expectancy mechanism diminished. This implies that if semantic-priming facilitation is going to measure pure automatic priming effects, SOA length may need to be even shorter than was previously thought. Unfortunately, no sufficient studies have yet been conducted to hypothesize the best time interval between the presentation of the first word, either in L1 or L2, and the onset of the second word, either in L2 or L1, in translation recognition. Furthermore, many of the past studies of translation recognition and semantic-priming task can also be criticized on their inclusion of a high proportion of related trials or translation equivalents. This incident may have also encouraged participants to develop expectations for the upcoming experimental trials, which in turn will improve bilinguals’ performance on the tasks. Therefore, in order to obtain a fair estimate of experimental effects, one needs to keep this proportion as low as possible when designing an experiment (Altarriba & Basnight-Brown, 2007).

Semantic-priming technique is hailed as one that provides the clearest evidence regarding the automatic processing of language, as compared with other techniques mentioned above. In fact, it has become one of the most important tools used to determine whether or not a bilingual’s languages are somehow interconnected and the levels at which this interconnectivity occurs (for a detailed reviews, see Altarriba and Basnight-Brown, 2007). This technique was primarily and mainly used to explore the organization of words and concepts in monolinguals (Meyer and his colleagues, 1971, 1975 and Neely, 1977), and the semantic-priming effect has observed and reported in hundreds of monolingual studies and under many different experimental manipulations (for reviews, see McNamara & Holbrook, 2003, Neely, 1991). However, it has been suggested that in order to gain the most accurate measure of how words are represented in memory in BILINGUALS, an experimental paradigm that is constrained and specific to bilinguals should be implemented. In this case, translation paradigm is unique to bilingual research due to the fact that individuals have to acquire two languages and these two lexicons have to be both activated in order to perform translation tasks (more will be said about this issue in the later section). For now, the point of translation paradigm is to demonstrate that only bilinguals are able to perform this task and the activation of both languages during the task will provide a clearer picture of bilinguals’ lexical and conceptual memory representations.

Methodological problems have been a longstanding barrier to the systematic exploration of issues in bilingual research. Up to now, however, bilingual research has broadened beyond traditional observational approaches to include a diverse set of experimental paradigms. From the effects of semantic manipulation in this latter set of studies, conceptual memory appears to be implicated in both L1-to-L2 and L2-to-L1 directions and to the same extent. At present, it is not all clear why the data of the studies already discussed, dealing with conceptually very similar issues, do not converge more closely. Of course, despite being conceptually very similar, the studies differ from each other in the ways in which the theoretical questions have been implemented, in terms of the participants, the stimulus materials, the experimental procedure, or a particular combination of these three. The choices made on each of these three dimensions may interact in complex ways, the details of which are not yet understood. It will be particularly crucial in future research to determine the relations between the conceptual variables that have been examined in these bilingual studies. It may be the case that the semantic organization of a list (the conceptual variable in the Kroll & Stewart, 1994, study), word concreteness (the conceptual variable in many of de Groot and her colleagues’ experiments), and semantic context (the conceptual variable in the La Heij et al., 1996, study and in experiments on bilingual semantic priming), though all legitimate semantic variables, may influence different processing loci during translation.

Access to Bilingual Memory Representation – Research Goals and Proposed Experiments

The evidence reviewed raised a question as to whether the asymmetry in bilingual memory representation proposed by the asymmetry model is universally applicable and whether it can be generalized beyond limited conditions and across various experimental paradigms. It is very possible that differences in methodologies used in many of the above studies are to blame for the varied results. It will remain to be determined which conditions constrain the form and direction of the asymmetries predicted by the asymmetry model for bilinguals at different stages of L2 proficiency. If bilinguals are capable of accessing conceptual memory equally well for both of their languages, then an important focus in this area of research will concern the form of concepts and modeling of activation of meaning for words in different languages. To address this issue, an alternative approach to modeling the representation of words and concepts in bilingual memory should focus on those aspects of words that appear to be associated with lexical or conceptual processing.

De Groot and her colleagues (1992, 1993, and 1994) have developed this approach by examining the effects of a host of variables thought to reflect lexical and conceptual levels of processing in bilingual performance in translation production and translation recognition. Characteristics of words may constrain the form of lexical and conceptual representation across languages. Two central findings in this work are that concrete words and cognates (words that are similar in spelling, pronunciation, and meaning across languages, e.g., English music and Spanish musica) are translated faster than abstract words and noncognates. These effects of concreteness and cognate status are instances of a group of effects known as WORD-TYPE effects. To accommodate these effects, de Groot (1992) proposed a conceptual feature model in which words in each of the bilinguals’ two languages activate conceptual features that are assumed to be distributed, such that particular concepts correspond to sets of activated features.

The asymmetry model (Kroll & Stewart, 1994) and the conceptual feature model (de Groot, 1992), although not mutually exclusive, focus on different aspects of lexical and semantic processing in bilinguals. The former examines the consequences of the developmental sequence in terms of the asymmetries in bilingual memory representation at different levels of L2 proficiency. The latter focuses on the characteristics of words that may constrain the form of lexical and conceptual representation across languages. The critical evidence reviewed in the earlier section, especially those against the asymmetry model, questioned the generalizability of this model. If one examines memory organization by employing bilinguals speaking two very different languages, such as orthographic Chinese and alphabetic English, and by using a wide range of experimental paradigms, he/she will provide a universal picture of lexical and conceptual memories in bilinguals. If the proposed experiments fail to provide evidence for the asymmetry model, the corresponding findings might be accounted for by the conceptual feature model.

The primary goal for future research is to evaluate the generalizability of the asymmetry model with Chinese and English. As orthographic Chinese and alphabetic English have a sharp contrast in phonology (pronunciation) and orthography (spelling), the bilingual findings based on these two languages will provide a universal picture of the asymmetries, if there is any, between bilinguals’ lexical and conceptual representations. Further more, English is known as ‘the language of communication’ because it is recognized as the International language, whereas Chinese is the language of nearly a quarter of the world’s population. The importance and differences of both these languages therefore constitute a major motivation of further studies.

Under this motivation and attempt to look systematically into the Chinese-English bilingual memory, my 3rd year research project, supervised by Dr. Walter van Heuven, at the School of Psychology, The University of Nottingham, evaluated the asymmetry model and further explored the effects of certain item (i.e., WORD-TYPE effect) attributes on bilingual processing in conjunction with the conceptual feature model (Zhang & van Heuven, 2007). In the study, Chinese words were selected with three levels of semantic transparency: transparent words with the same meaning as their component characters (e.g., 睡眠 meaning sleep; 睡 means sleep, 眠 means sleep); modifier words, where only one of the characters contributes to the meaning of the words (e.g., 脸蛋 - face; 脸 - face, 蛋 - egg); and opaque words with a different meaning from their component characters (e.g., 地方 - place; 地 - ground, 方 - square). The asymmetry model assumes that the effect of semantic transparency (a) will be larger in forward translation (from the first to the second language) than in backward translation (from the second to the first language) and (b) hence forward translation will be slower than backward translation. Two translation tasks were employed under the assumption that translation paradigm is unique to bilingual research (see earlier discussion). Firstly, translation recognition task, where participants were visually presented two words from each of their languages and asked to decide whether they are correct translations of each other. The subsequent words were presented either in Chinese first then English, which represents forward translation, or the reverse, which reflects backward translation. And secondly, translation production task, where a separate group of bilinguals were presented with a word in one of their languages and had to vocally produce its correct translation in the other language. They had to translation from Chinese to English (forward translation) or from English to Chinese (backward translation). The results showed that semantic transparency had a significant impact on both forward and backward translation in both these translation tasks, which indicates that both translation directions are conceptually mediated and that these groups of Chinese-English bilinguals are able to conceptually mediate their L2 (English). Furthermore, forward translation was found to be significantly faster than backward translation in the translation recognition task; however, no such translation asymmetry was evident in the translation production task. These results did not support either of the predictions from the asymmetry model; however, they supported the conceptual feature model in terms of the Chinese WORD-TYPE effect (i.e., the manipulation of semantic transparency associated with conceptual processing).

To my knowledge, the above study was the first to evaluate the asymmetry model by comparing two translation paradigms in two distinct languages, namely Chinese and English. One therefore cannot conclude that the asymmetry model fails to account for Chinese-English bilingual memory representation based on this preliminary research. However, these innovative findings on Chinese-English bilinguals have certainly encouraged further research and brought an extra dimension to the evaluation of current models of bilingual memory representation.

In an attempt to fully assess the role of processing time, future research should investigate the effect of the stimuli onset asynchrony (SOA) of a word pair in translation recognition task. As discussed above, a long SOA may encourage participants to translation the first and/or the second word into the same language; and it can also make it easier for participants to generate expectancy to a set of related words. Unfortunately, no sufficient studies have yet been conducted to hypothesize the best time interval between the presentation of the first word, in L1 or L2, and the onset of the second word, in L2 or L1, in translation recognition. De Groot and Comjis (1995) used 238msec SOA in their study and found that the order in which the translations were presented, Dutch word or English word first, had no effect on the translation recognition performance (i.e., no translation asymmetry). However, with 500msec SOA in the study of Chinese-English bilinguals by Zhang and van Heuven (2007), forward translation was found to be faster than backward translation (i.e., opposite translation asymmetry). One possible cause for this discrepancy is this different time interval between the onset of the first and second word of a translation pair. Further study should manipulate the SOA (e.g., 240msec vs. 500msec, vs. 800msec vs. 1sec) and look systematically into its effects on the asymmetry in both backward and forward translation.

A model can only be generalizable if it can account for all the findings based on different experimental paradigms. An additional goal for future research is to examine Chinese-English bilingual memory representation by using a variety of experimental paradigms that have been used on bilingual research, such as aforementioned semantic-categorization task and semantic-priming technique. However, future studies should improve the methodology used in previous studies, use highly constrained conditions and apply any new findings obtained to current models of bilingual memory. The primary improvement should be directed towards a careful control of (1) the SOA, as the length of the SOA could moderate the strength of experimental effect, as well as influencing the degree to which strategic processing occurs; (2) relatedness proportion or translation equivalents proportion, as high proportions will encourage participants to develop expectations for the upcoming experimental trials, which in turn will improve bilinguals’ performance on the tasks, hence, the proportions should be kept as low as possible; and (3) language proficiency, as the difficulty with generalizing across the reported studies lies in the variability of the participants and of their mode of language acquisition and the varying age of acquisition across the participants. The use of actual tests of language proficiency that are aligned with the variables under investigation (e.g., speaking test for vocal production studies such as translation production) should be adopted rather than subjective ratings of language comprehension in a questionnaire. These measures should also be developed to allow for the identification of which language is dominant for a given speaker. In some cases for instance (Basnight-Brown & Altarriba, 2007), although the bilingual’s first language is Spanish, they are more proficient in their second language English; hence the dominant language in this case is the L2.

Conclusions

This paper reviews recent evidence concerning the relations between lexical and conceptual memory when critically evaluate the asymmetry model (Kroll & Stewart, 1994). Although the main features derived from the model are still controversial which in turn questions the its generlizability, future efforts toward the inclusion of languages with different orthographic systems (e.g., Chinese and English), the improvement of the methodology, the use of constrained conditions will provide a productive way to evaluate the model itself and to understand the dynamic changes that undergo with increasing proficiency in the second language.

Word Count: 6119

References:

Altarriba, J. (2000). Language processing and memory retrieval in Spanish-English bilinguals. Spanish Applied Linguistics, 4, 215-245.

Altarriba, J., & Basnight-Brown, D. M. (2007). Methodological considerations in performing semantic- and translation-priming experiments across languages. Behaviour Research Methods, 39(1), 1-18.

Basnight-Brown, D. M., & Altarriba, J. (2007). Differences in semantic and translation priming across languages: The role of language direction and language dominance. Memory and Cognition, 35(5), 953-965

Caramazza, A., & Brones, I. (1980). Semantic classification by bilinguals. Canadian Journal of Psychology, 34, 77-81.

Chen, H. C. (1990). Lexical processing in a nonnative language: Effects of language proficiency and learning strategy. Memory and Cognition, 18, 279-288.

Chen, H. C. (1992). Reading comprehension in Chinese: Some implications from character reading times. In H. C. Chen & O. J. L. Tzeng (Eds.), Language Processing in Chinese (pp. 175-205). Amsterdam: Elsevier.

Chen, H. C. (1996). Chinese reading and comprehension: A cognitive psychology perspective. In M. H. Bond (Ed.), Handbook of Chinese Psychology (pp. 43-62). Hong Kong: Oxford University Press.

Chen, H. C., & Ho, C. (1986). Development of Stroop interference in Chinese-English bilinguals. Journal of Experimental Psychology: Learning, Memory, and Cognition, 12, 397-401.

Chen, H. C., & Leung, Y. S. (1989). Patterns of lexical processing in a nonnative language. Journal of Experimental Psychology-Learning Memory and Cognition, 15(2), 316-325.

Chen, H. C., & Ng, M. L. (1989). Semantic facilitation and translation priming effects in Chinese-English bilinguals. Memory and Cognition, 17, 454-462.

Collins, A. M., & Loftus, E. (1975). A spreading-activation theory of semantic processing. Psychological Review, 82, 407-428.

de Groot, A. M. B. (1992). Determinants of word translation. Journal of Experimental Psychology-Learning Memory and Cognition, 18(5), 1001-1018.

de Groot, A. M. B. (1993). Word-type effects in bilingual processing tasks: Support for a mixed representation system. In R. Shcreuder & B. Weltens (Eds), The Bilingual Lexicon (pp. 389-412). Amsterdam: Elsevier.

de Groot, A. M. B., Comijs, H. (1995). Translation recognition and translation production: Comparing a new and an old tool in the study of bilingualism. Language Learning, 45(3), 467-509.

de Groot, A. M. B., Dannenburg, L., & van Hell, J. G. (1994). Forward and backward translation by bilinguals. Journal of Memory and Language, 33, 600-629.

de Groot, A. M. B., & Hoeks, J. C. J. (1995). The development of bilingual memory: Evidence from word translation by trilinguals. Language Learning, 45(4), 688-724.

de Groot, A. M. B., & Nas, G. L. J. (1991). Lexical representation of cognates and noncognates in compound bilinguals. Journal of Memory and Language, 30, 90-123.

de Groot, A. M. B., & Poot, R. (1997). Word translation at three levels of proficiency in a second language: The ubiquitous involvement of conceptual memory. Language Learning, 47, 215-264.

Dufour, R., & Kroll, J. F. (1995). Matching words to concepts in two languages: A test of the concept mediation model of bilingual representation. Memory and Cognition, 23(2), 166-180.

Glaser, W. R., & Glaser, M. O. (1989). Context effects in Stroop-like word and picture processing. Journal of Experimental Psychology: General, 118, 13-42.

Keatley, C. W., Spinks, J. A., & de Gelder, B. (1994). Asymmetrical cross-language priming effects. Memory and Cognition, 22, 70-84.

Kroll, J. F. (1993). Accessing conceptual representations for words in a second language. In R. Schreuder & B. Wletens (Ed.), The Bilingual Lexicon (pp. 53-81). Philadelphia: John Benjamin.

Kroll, J. F., & Curely, J. (1988). Lexical memory in novice bilinguals: The role of concepts in retrieving second language words. In M. Gruneberg, P. Morris & R. Sykes (Eds.), Practical Aspects of Memory (Vol. 2, pp. 389-395). London: John Wiley & Sons.

Kroll, J. F., Michael, E., Tokowicz, N., & Dufour, R. (2002). The development of lexical fluency in a second language. Second Language Research, 18, 137-171.

Kroll, J. F., & Stewart, E. (1994). Category interference in translation and picture naming - Evidence for asymmetric connections between bilingual memory. Journal of Memory and Language, 33(2), 149-174.

La Heiji, W., de Bruyn, E., Elens, E., Hartsuiker, R., Helaha, D., & van Schelven, L. (1990). Orthographic facilitation and categorical interference in a word-translation variation of the Stroop task. Canadian Journal of Psychology, 44, 76-83.

La Heij, W., Hooglander, A., Kerling, R., & van der Velden, E. (1996). Nonverbal context effects in forward and backward translation: Evidence for concept mediation. Journal of Memory and Language, 35, 648-665.

Mazibuko, T. L. (1991). Interference in translation: Using distracting information to reveal the nature of bilingual language processing. Unpublished masters thesis, Mount Holyoke College, South Hadley, MA.

McNamara, T. P., & Holbrook, J. B. (2003). Semantic memory and priming. In A. F. Healy & R. W. Proctor (Eds.), Handbook of Psychology: Experimental Psychology (Vol. 4, pp. 447-474). New York: Wiley.

Meyer, D. E., & Schvaneveldt, R. W. (1971). Facilitation in recognizing pairs of words: Evidence of a dependence between retrieval operations. Journal of Experimental Psychology, 90, 227-234.

Meyer, D. E., Schvaneveldt, R. W., & Ruddy, M. G. (1975). Loci of context effects on visual word recognition. In P. M. A. Rabbitt & S. Dornic (Eds.), Attention and Performance V (pp. 98-118). New York: Academic Press.

Neely, J. H. (1977). Semantic priming and retrieval from lexical memory: Roles of inhibitionless spreading activation and limited-capacity attention. Journal of Experimental Psychology: General, 106, 226-254.

Neely, J. H. (1991). Semantic priming effects in visual Word recognition: A selective review of current findings and theories. In D. Besner & G. Humphreys (Eds.), Basic Processes in Reading: Visual Word Recognition (pp.246-336). Hillsdale, NJ: Erlbaum.

Potter, M. C., So, K. F., Voneckardt, B., & Feldman, L. B. (1984). Lexical and conceptual representation in beginning and proficient bilinguals. Journal of Verbal Learning and Verbal Behavior, 23(1), 23-38.

Rosch, E. (1975). Cognitive representations of semantic categories. Journal of Experimental Psychology: General, 104, 192-233.

Sanchez-Casas, R. M., Davis, C. W., & Garcia-Albea, J. E. (1992). Bilingual lexical processing: Exploring the cognate/non-cognate distinction. European Journal of Cognitive Psychology, 4, 293-310.

Sholl, A., Sankaranarayanan, A., & Kroll, J. F. (1995). Transfer between picture naming and translation - A test of asymmetries in bilingual memory. Psychological Science, 6(1), 45-49.

Smith, E. E., Shoben, E. J., & Rips, L. J. (1974). Structure and process in semantic memory: A featural model for semantic decisions. Psychological Review, 81, 214-241.

Snodgrass, J. G. (1993). Translating versus picture naming: Similarities and differences. In R. Schreuder & B. Weltens (Eds.), The Bilingual Lexicon (pp. 83-114). Philadelphia: John Benjamin.

Sunderman, G., & Kroll, J. F. (2006). First language activation during second language lexical processing - An investigation of lexical form, meaning, and grammatical class. Studies in Second Language Acquisition, 28(3), 387-422.

Talamas, F., Kroll, J., & Dufour, R. (1995). From form to meaning: Stages in the acquisition of second language vocabulary. Bilingualism: Language and Cognition, 2(1), 45-58.

Tzelgove, J., & Eben-Ezra, S. (1992). Components of the between-language semantic priming effect. European Journal of Cognitive Psychology, 4, 253-272.

Zhang, T., & van Heuven, W. J. B. (2007). Effect of semantic transparency on word translation: Evidence for concept mediation and translation asymmetry in Chinese-English bilinguals. Manuscript in preparation, The University of Nottingham, United Kingdom.