Table of contents
2. The multilingual lexicon
2.1. Storage of words
2.2. Word recognition
2.3. Research evidence for a language nonselective access
3. Language interconnection
3.1. Crosslinguistic influence (CLI)
3.2. Crosslinguistic interaction (CLIN)
3.3. Codeswitching (CS)
As more than 7000 languages are being spoken across the globe (cf. Simons and Fenning 2017), multilingualism is a very common phenomenon in today’s globalized and networked world. Speakers of various languages are interconnected and in contact with each other on a daily basis, which means that proficiency in more than one language is increasingly becoming standard as a result of “the ever-growing demand for international communication and international mobility” (Stavans and Hoffmann 2015: 228). Nowadays, it is very likely that an individual is not only bilingual but multilingual because of strongly increased opportunities for language contact resulting from globalisation, migration and the development of English as a lingua franca. Today, English is not only learned and used as a means of communication among non-native English speakers worldwide but also, as a global language, enables people from diverse linguistic backgrounds to understand and communicate with each other. Foreign language learning and teaching has become more and more significant and of international importance. However, “[i]n spite of its importance as a global phenomenon, multilingualism has not received much attention on part of researchers in linguistics, psycholinguistics and applied linguistics” (Cenoz et al. 2003: 1). Although bilingualism and second language acquisition (henceforth SLA) have been fundamental issues in psycholinguistics and applied linguistics throughout the last decades, few researchers have stepped beyond bilingualism and focused on speakers with command of more than two languages (cf. ibid.). Despite the fact that bilinguals and multilinguals have a lot in common, it is important to note “the specific processes resulting from the interaction between the languages that may result from the simultaneous presence of more than two languages in the multilingual person’s mind” (ibid.: 2). Whether or not there are interconnections between the various languages in a multilingual´s mind and to what extent these interactions influence the speaker´s language competences constitute central questions in multilingual research. Furthermore, “[t]he potential ways in which learning a second or third language can influence other languages in the mind of the (emergent) multilingual is an increasingly important issue given the international importance of foreign language learning” (Fung and Murphy 2016: 6).
This paper presents an overview of research on the multilingual lexicon as well as the highly complex storage of words in the multilingual mind. Furthermore, it deals with word recognition processes and discusses the issue of a language selective versus language nonselective access. The second part of this paper is concerned with the various ways in which different languages are interconnected and influence each other. Based on this, the phenomenon of codeswitching is illustrated as a widespread example of crosslinguistic interaction in the multilingual lexicon. This paper aims at presenting an overview of the complex characteristics of the multilingual lexicon. It should also give an idea of the immense amount of linguistic knowledge that is stored and processed within the multilingual mind and, most importantly, emphasize the significant linguistic awareness and cognitive skills that multilinguals display.
2. The multilingual lexicon
The mental lexicon constitutes the systematic and highly complex organization of words in the human mind. It is the mental storage of all information related to words and contains a speaker´s extensive linguistic knowledge. While monolinguals are already capable of “selecting/identifying a word within a third of a second from a lexicon of 50,000 words or more” (Dijkstra 2003: 11), the quantity of stored words in a multilingual mind is significantly higher. In case of a multilingual speaker, the mental lexicon includes tens of thousands of additional words. Therefore, the processes of word recognition during reading and listening and word production when speaking or writing are even more complicated as a large number of extra words make it more difficult to retrieve the target word. However, multilinguals do not take significantly more time in processing words and mistakes occur rather rarely.
Dijkstra describes the mental lexicon as the “language processing system” (2007: 251), which “stores a language user´s knowledge with respect to words” (ibid.: 252) and which “raises some unique and fundamental questions” (ibid.: 251). Research on multilingualism is especially interested in the structure and organization of the multilingual lexicon as well as the closely related word production and word recognition processes in a multilingual´s mind. Thus, the question whether the multilingual lexicon represents a shared database which includes words from different languages expresses a major issue of psycholinguistic research. If different languages are not stored separately, how can multilinguals manage to communicate without “suffering from many misperceptions of words or cross-linguistically based speech errors” (Dijkstra 2003: 11)? Furthermore, Dijkstra proposes another central question of research on multilingualism and the multilingual lexicon: “[h]ow can multilinguals retrieve the right words from their mental lexicon so quickly?” (2007: 252).
2.1. Storage of words
As the mental lexicon stores all linguistic data and includes a vast number of words from different languages in the case of a speaker who has a command of three or more languages, it contains an enormous amount of information to specify every word. This immense knowledge is extremely organized and every word is represented with respect to orthography, phonology, semantics, morphology, pragmatics and language membership information of a word (cf. Dijkstra 2007: 252). Language membership information is essential as it reveals to which language a word belongs. This huge database consisting of all linguistic information is saved in “a multidimensional storage space set up along these dimensions” (ibid.), which constitutes the multilingual lexicon. Depending on the modality of word recognition (i.e. after a visual or auditory input), different dimensions are activated at first. For example, when reading a word, the entry dimension is orthographic, whereas during listening, the first activated code in the multilingual lexicon is phonological. When speaking, we activate a definite concept or precise meaning before producing words (cf. ibid.). These dimensions are also essential to specify each word within the immense database of the multilingual lexicon. Although every word has its unique position in this multidimensional storage space, certain word clusters appear. Words of the same language are stored more closely together than words from different languages. In fact, “[t]he greater the differences between two languages in terms of their word characteristics, the greater the language distance” (ibid.).
2.2. Word recognition
In monolingual word recognition, the Interactive Activation (IA) model is a well-known description of the perception of visual language input. It includes several stages of processing, which all form representations of the target word at different levels of abstraction. According to McClelland & Rumelhart, in visual word recognition, a feature level, a letter level, a word level and “higher levels of processing that provide “top-down” input to the word level” (1981: 377) are processed. However, they assume that this model does not operate sequentially and the several levels of word recognition are not accessed one after another but rather parallel. In their model, “[p]erceptual processing is assumed to occur in a set of interacting levels, each communicating with several others” (ibid.). Also, excitatory or inhibitory connections between so-called neighbors exist either between adjacent levels or within levels but “[t]here are no connections between nonadjacent levels” (ibid.: 379). Neighbors are words that are very similar to a target word and only differ at one letter position. For example, BIND, KIND, WAND, WILD, and WINK are neighbors of the four-letter word WIND (cf. Dijkstra 2003: 13). McClelland & Rumelhart point out that [c]onnections within the word level are mutually inhibitory, since only one word can occur at any one place at any one time. Connections between the word level and letter level may be either inhibitory or excitatory (depending on whether the letter is a part of the word in the appropriate letter position) (1981: 379).
Furthermore, words which are used more frequently are generally recognized faster than words of lower frequency. This also results in an intensified inhibitory impact of high frequency words on neighbors. Therefore, frequency of use has an important effect on word recognition processes, especially in the first stages of word perception as many word candidates are activated simultaneously.
The monolingual IA-model can easily be extended to a Multilingual Interactive Activation (MIA) model, as proposed by Dijkstra (2003: 17-18). This model represents an integrated multilingual lexicon consisting of all known words in different languages. It is important to note that with each additional language system, the number of existing words in the mental lexicon increases and “[w]ith a larger density of words, the competition between words (lateral inhibition) becomes stronger, and the moment in time that the presented word can be identified is delayed” (ibid.: 18). However, Dijkstra also points out that “this is the case irrespective of whether the newly acquired words are from the same language or from another language” (ibid.). No matter if the vocabulary acquisition occurs in a native language or a foreign language, the more words are added to the mental lexicon the more potential word candidates are activated, which leads to an increasing difficulty in retrieving the right word.
The study of word recognition in multilinguals is primarily considered with language selective versus nonselective access (cf. Dijkstra 2003: 14-17; De Bot 2004: 17-20). According to the language selective (or language specific) position, there are separated networks of words for each language in the multilingual lexicon. After a visual or auditory input, different languages are not activated simultaneously but separately. Therefore, during word recognition only the target language system is accessed in the multilingual lexicon and “[o]nly if the lexical representation corresponding to the input is not found in the lexicon of the target language, is contact established with the other lexical system” (Dijkstra 2003: 14). This viewpoint also suggests a specific selection mechanism, which controls the activation of the relevant language and which operates at a very high level of selectivity as it directs all received information (i.e. the visual or auditory input) only to the lexical system which is required to fulfil the task. On the contrary, the language nonselective (or language nonspecific) access suggests that there is a parallel activation of the existing language systems in the mind. Therefore, lexical candidates from different languages are accessed simultaneously after a visual or auditory input. This viewpoint also implies that words are not primarily organized with respect to the language they belong to but other criteria, e.g. frequency of use (cf. De Bot 2004: 18).
Traditionally, the position of a language selective access to the multilingual lexicon was favored and the hypothesis of separated lexical networks in the multilingual mind was strongly supported. Early studies presented evidence for the language selective access (cf. Dijkstra 2003: 14) as, for example, monolinguals and bilinguals showed no significant differences in reaction time in a lexical decision experiment, which suggested that all study participants processed the input in a language specific manner. However, over the past two decades there has been an increasing number of studies supporting the language nonselective hypothesis (cf. ibid.: 14-16). Various experimental research “have demonstrated that there is language nonselective access rather than selective access in many experimental circumstances”, as Dijkstra (2003: 16) points out. Therefore, it is to assume that multilingual word perception operates in a language nonselective manner and demands a simultaneous activation of possible word candidates from different languages.
2.3. Research evidence for a language nonselective access
Neighborhood effects provide one set of evidence that supports a language nonselective access to the multilingual lexicon. If a word has many neighbors, there is an increased number of word candidates in word recognition, which makes it more difficult to select the target word. As a result, the process of word perception is slowed down. In research on bilingualism and multilingualism, this effect is used to study the problem of language selective/nonselective access since there should be no impact on latencies if multilinguals operate in a language specific manner because neighbors from other languages are not competing word candidates. In an experimental study, van Heuven et al. “found that an increase in the number of Dutch neighbours leads to slower reaction times in an English lexical decision task, and this is supportive for nonselective access “ (De Bot 2004: 18).
Several other studies have shown evidence for a language nonselective access hypothesis by providing data from experimental research with interlingual homographs and interlingual homophones. Interlingual homographs are words in two or more languages that share the same orthographic representation but have different meanings, whereas interlingual homophones sound similar in different languages but also differ in meaning. An example for both is GIFT in English and German (where it means ‘poison’). Similar to studies on neighborhood effects, research on interlingual homographs and homophones argues that these similarities in various languages should have no effect on reaction times if multilinguals operate in a language selective manner. However, in case of a language nonselective access to the multilingual lexicon, the activation of interlingual homophones and homographs will again cause longer reaction times. Although early studies showed no significant differences in latencies between monolinguals and bilinguals with respect to interlingual homographs (cf. Dijkstra 2003: 14), “later studies have revealed that the lexicon of other languages is activated” (ibid.). For example, “[t]his was demonstrated by the finding that in an English lexical decision task performed by Dutch-English bilinguals, the word FIRE was semantically primed by the presentation of the word BRAND (the Dutch word for “FIRE”) on the previous trial” (ibid.: 15). Even though these studies did not discover any differences in latency between interlingual homographs and control words, a strong evidence for the language nonselective hypothesis was provided as they found significant facilitation effects and cognate advantages. Furthermore, lexical processing in a native language was found to be affected by knowledge of foreign languages and vice versa. Dijkstra points out that “[s]everal other studies have shown that language nonselective effects also occur with other types of stimulus materials and different tasks […], [and] that they occur not only from L1 on L2, but also in the opposite direction […]” (2003: 15).
3. Language interconnection
As the multilingual lexicon stores all information about the acquired languages in a highly organized and effective way, research on multilingualism is particularly interested in the consequences of this simultaneous existence of three or more languages in a speaker´s mind. As Cenoz et al. put it, “[t]o find out whether there are interconnections between the various lexicons in the multilingual’s mind is certainly a burning question for research on multilingualism” (2003: 3). Crosslinguistic influence (henceforth CLI) forms an important part of research on bilingualism and multilingualism as it is essential to investigate the connections between different languages in one mind in order to understand the complex processes in language acquisition.
While crosslinguistic influence “is predominantly related with studies of second language acquisition” (Jessner 2003: 45) and mostly discussed in research on bilingualism, it is obviously just as significant in studies of multilingualism and third language acquisition (henceforth TLA). As there are substantial differences in SLA and TLA as well as enhanced linguistic and cognitive skills that multilinguals display, occurring language transfer phenomena in bilingual and multilingual speakers also differ. For example, an individual who has already acquired one (or more) foreign language(s) can use the developed language learning strategies and metalinguistic skills for further language learning and use. Based on these findings, Jessner presents the concept of cross-linguistic interaction (henceforth CLIN) as a ”significant factor representing the nonreducible dynamic aspect of the multilingual system” (2003: 47) in contrast to CLI, which does not cover such dynamic effects in the multilingual mind.
3.1. Crosslinguistic influence (CLI)
The term crosslinguistic influence describes “phenomena such as 'transfer', ‘interference', 'avoidance', 'borrowing', and L2-related aspects of language loss“ (1986: 1), as defined by Kellerman and Sharwood-Smith. It refers to the various ways in which the different languages in one mind can affect each other and is the commonly used term for all interactions between languages in a bilingual or multilingual speaker´s mental lexicon.
CLI has traditionally been recognized as unidirectional and exclusively occurring in a forward direction. There are numerous examples of forward CLI in which an earlier acquired language influences a later learned language and which is often explained as emerging from similarities and differences between the languages (cf. Fung and Murphy 2016: 7). Nevertheless, it is important to consider the bidirectional relationship and high complexity of CLI (cf. Jessner 2003: 45). Numerous studies show that CLI can also occur in a backwards direction (often called reverse CLI), which means a later acquired language exerts an effect on a previously acquired language. Consequently, an L1 does not only affect an L2 but also vice versa. Compared to SLA, in TLA these interconnections of languages appear in an even more complex manner. Whereas in SLA there are only two relationships to consider, there are already six possible ways in which one language could affect another in TLA: L1 à L2, L1 à L3, L2 à L1, L2 à L3, L3 à L1 and L3 à L2. Therefore, it is to assume that the more languages one knows, the more multifaceted the language interactions in one´s mind are.
- Quote paper
- Anonymous, 2017, The multilingual lexicon. An overview of research, Munich, GRIN Verlag, https://www.grin.com/document/379172