BILINGUAL LEXICAL RETRIEVAL IN DORMANT DUTCH BILINGUALS
Description of an ongoing investigation of retrieval difficulties in the speech of Dutch immigrants in Victoria after decades of non-use of their mother tongue
Tom Ammerlaan
Outline
The investigation focuses on psycholinguistic aspects of language attrition (Weltens, de Bot and Van Els, 1986). Language attrition can be defined as the reduction in knowledge and skill in a language, as a result of either brain injury or a decrease in usage. This area of research generally adopts the assumption that if a person cannot retrieve a particular target item in an obligatory context (that is, where fluent people would use it) then that item has been 'forgotten'. Studies of 'tip-of-the-tongue' phenomena (Browman 1978) and of aphasia (Paradis 1978) have shown however, that this 'loss' van be temporary. These studies furthermore suggest that different types of words vary in their susceptibility to 'loss' (cf. Anderson 1982). The characteristics of retrieval are considered to be responsible for these results. In this context an in-depth comparison of bilingual recall and recognition processes for various word types will assist in clarifying issues in language attrition research.
Is information difficult to access as a result of non-use? What has happened to the information? Does it gradually disappear from memory, or is it repressed?
Current theories on memory assert that what happens to information which is currently not needed depends on the type of information. If the information was processed intensively and exercised in a number of contexts for a prolonged period, then it is stored permanently. Other information which was considered less essential and which was only briefly considered seems to disappear from memory.
The same happens with language: when listening to someone, one generally remembers the line of an argument, although the precise wording is lost. Often the form of the message is not essential. The same principle appears to apply to languages as a whole over time: essential elements or well- rehearsed elements remain in the context of non-use. Investigations on post-high school French reveal a similar trend. Often used phrases like "My name is...", "Could you please...", tend to re-emerge after years of non-use, whilst more obscure vocabulary is difficult to remember.
An issue is whether the same applies to a first language which is no longer used? One would expect the same principles to apply, although it is uncertain to what extent, and how this emerges in actual language usage. After all, 'Essential" is a rather vague term: does it mean that without that language element, other language elements cannot be produced? Does it mean that elements in the used language which are similar to elements in the disused language are retained even though they are less essential? Can one furthermore speak of loss of a mother tongue, which has been rehearsed well and learned profusely, and does loss mean non-production only or non-recognition as well? For these reasons I carried out my investigation into the loss of Dutch in Victoria.
I aimed at investigating the role of the similarity between Dutch and English and of the class of words (function versus content) in the context of recognition and production of Dutch by migrants who had not spoken Dutch for a prolonged period.
Similarity was looked at because investigations into bilingual processing of speech has shown that bilingual people are influenced by this word feature: words that are very similar in both languages (cognates) are affected when their equivalents are used, whereas dissimilar words are not. This would suggest that Dutch migrants should remember similar words like "arm", "huis" and even "deur" (= arm, house, door) better than dissimilar words like "bureau" and "vlinder" (= desk, butterfly) because the latter would not be activated when the English equivalents are used. A pilot study in which the strategies used to recall the Dutch words were looked at confirms this: subjects said they attempted to retrieve Dutch via English, and only once they had translated the words they needed did they assess whether or not the result was correct Dutch.
Various degrees of similarity were investigated for words with the same frequency of occurrence, like phonological similarity (do they sound the same?), morphological similarity (do the translation equivalents have the same number of syllables?) are both words compound or not, and within compound words, do the words share meaning components or not. Examples of the categories are given below.
Subjects were asked to recall words by naming their corresponding picture, or if they could not, by identifying the name of the picture from a list of words presented below it.
Word-class was looked at by means of a recognition experiment, during which migrants were asked to indicate rapidly whether a single function or content word was a Dutch word or not. When analyzing Dutch migrants' speech investigators have noticed a mixture of Dutch and English words, commonly called 'Strutch'. It has been noticed that primarily English function words which are similar emerge in a person's Dutch ("the, an, on, but") whereas English content words are less likely slip in. This would suggest that function words in Dutch and English are shared, and that Dutch function words tend to have disappeared. Whether that was the care was investigated in the recognition experiment: it was argued that if words had disappeared they would not be accurately or rapidly recognised.
Research on bilingualism, however, also shows that the way words are processed depends on the way they were learned and are currently used. Research on memory shows that this also depends on how well they were learned. Therefore it needed to be determined for how long Dutch had been used prior to migration, and how it was used (if at all) in Australia. There were no records of the level of pre-emigration Dutch, and therefore this needed to be assessed a posteriori. A questionnaire was therefore used to obtain information on how long the migrants had lived in the Netherlands, where they had used Dutch after arrival in Australia, with whom and in what contexts, as well as how proficient they thought they were now. In addition, global tests of their proficiency and fluency in Dutch were conducted to determine to what extent those former native speakers were now able to perform in Dutch.
This information was collected in 1988 with the help of a number of volunteers who kindly donated considerable amount of time. From then onwards the battle began to combine the various types of information to derive a coherent picture of the effect of similarity and word-class on first language attrition.
Some preliminary results.
The background information showed that in the sample most subjects in their thirties and forties only used Dutch in the personal domain, that they rarely attended specific Dutch events like the Kermis or Dutch Church services, and that they rarely had deliberate contact with fellow Dutch migrants. Most were not married to Dutch spouses, and if they were Dutch was used a secret language when the children were not supposed to hear what the parents were saying. Dutch was rarely written, sometimes read and spoken, but most frequently heard. Most participants stated they felt they had lost some of their competence in Dutch.
The tests results corroborated this to some extent: the migrants performed less well than their Dutch controls in the Netherlands. English particularly seemed to interfere with the ability to correctly answer all task items. Which elements of their background seemed to affect their performance was established by means of canonical correlations, regressions and factor analyses. This showed that the age when the migrant departed the Netherlands was critical to how good their Dutch was. The older the subject was on departure, the more Dutch they had picked up, and the more Dutch was used, presumably because for these subjects the native language had become part of their identity.
The picture-naming experiment showed that similarity affected performance: the more similar, the better the Dutch words were named and recognised. Yet the reverse was not the case: in particular, words that were partially similar were difficult to recall and recognise. In particular, morphological similarity disrupted recall and recognition. Disuse appears to have the effect that the membership of a word to one language becomes vague: for dissimilar words it remains clear, for similar words it does not matter as pronouncing one is almost like pronouncing its cognate. But when words are partly similar it is no longer clear which is which. Hoeks (1988) has already shown this in the case of idioms: idioms that were similar became corrupted "to have a blue/black eye" whereas more distinct idioms remained intact.
In comments during the experiment the effect of similarity of form on performance was made clear: subjects recalled "botervlieg" on the basis of "butterfly" instead of "vlinder".
In the lexical decision experiment no difference was found in how accurately and quickly the function and content words were recognised. This suggests that their processing during access from memory was not dissimilar, so that any differential processing must occur at a later stage during language production. From interviews and comments by the migrants it appeared that the migrants produce an English frame for their Dutch sentence, and gradually attempt to fill in the positions with Dutch words. If these words are similar in form and usage to English, they seem more likely candidates than less similar words.
Subjects had also performed a similar task using equivalent English words. When it was investigated how long and short Dutch and English function- and content words were recognised, a remarkable difference appeared: short Dutch words were considered extremely difficult, while in English the long words provided more difficulty. This suggests that the way the words were read was different. Disuse appears to affect how words are read.
Next these effects were compared to the background information. It appeared that subjects who used Dutch as well as English interchangeably tended to be less certain of the distinction between the two languages than subjects who did not use Dutch. It appeared Dutch remain in a 'permastore' that was not affected.
Comparison to the level of proficiency measured in the tasks also showed an interesting result: subjects who were still fairly proficient were less affected by considerations of crosslinguistic similarity than less proficient migrants. This was the same both in recognition as in recall.
In conclusion, it appeared that that processing of Dutch was affected by attrition, and that the effect was most measurable in the production processes. It seems subjects use the tools from English to produce a Dutch sentence, in the belief that the similarity in the grammars of both languages is sufficient to result in successful communication.