verbal working memory and sla

Recently I have become interested in the relationship between working memory and language acquisition. According to the dominant model of working memory proposed by Baddely and Hitch working memory is divided into separate components. In this model a central executive controls two subcomponents. One, the visuo-spatial sketchpad stores visual and spatial information while the other, the phonological loop, is a limited capacity store of phonological material. The phonological loop is further divided into a short term store where phonological material fades in a short amount of time unless it is refreshed by the other component called the articulatory loop.

Individual capacity of the phonological loop is usually tested by digit span, the maximum number of digits an indivudual can recall, or by nonword repetition. Studies of individuals with deficits in the phonological loop showed that patients with severe impairments in this component of working memory were otherwise able to function normally in cognative tasks.(Baddely, Journal of Communication Disorders, vol 36, 2003). This lead to the hypothesis that the function of the phonological loop is the acquisition of phonologically unfamiliar words, or new vocabulary. Several subsequent studies have found strong correlations between phonological loop capacity and vocabulary in children, and experimental studies have shown that variables known to affect the capacity of the phonological loop impair word-nonword leaning but now the ability to form semantic associations between known word pairs such as “table-rabbit”. (Baddely 2003)

An interesting study was done by Service (Service, E. 1992, Phonology, Working Memory and Foreign language-Learning. Quarterly Journal of Experimental Psychology, Vol. 45) In this Study Service examined the relationship between nonword repetition and the acquisition of English of Finnish Children over a several year period. She found that nonword repetition of nonword conforming to English phonology predicted success in English language classes even after variables such as nonverbal intelligence had been controlled.

This study was interesting to me because it may have implications in second language teaching. With communicative teaching methods there seems to be little room for rote rehearsal of words, as well as more of a focus on pronunciation. It would be interesting to see what benefits there may be in rote rehearsal exercises of new words as well as more attention to the sound system of the target language in the classroom. The hypothesis of Baddely is that working memory allows novel phonological forms to be held briefly while more permanent representations are set up in long term memory. One natural way that the phonological loop maintains phonological forms is through subvocal rehearsal which refreshes the phonological store. Rehearsal and more familiarity of the phonology of the target language may facilitate vocabulary acquisition in the classroom.

Contrastive Analysis Hypothesis

The contrastive analysis hypothesis claimed that the difficulty of second language acquisition could be predicted by the degree of difference between the learners first and second language. In chapter nine Brown points out many of the problems with this hypothesis which lead linguists to abandon it. One of the problems with the hypothesis is that it could not predict many of the errors that learners make, and it also predicted interference problems where none would surface.

The most interesting criticism of the contrastive analysis hypothesis for me was the ‘subtle differences” version of the contrastive analysis hypothesis proposed by Oller and Ziachosseiny (Brown p253). They claimed that more interference between the L1 and target language may occur not when there is a large difference between a structure in the two languages but when learners are required to make more subtle distinctions between the languages. In my own experience learning French and Arabic I have found this to be the case. French and English share a lot of vocabulary so there are a lot of cognates as well as false cognates which can cause confusion for learners. Cognates can help in comprehension, but they can cause some problems in production. For example, when I speak French sometimes I may know a French word, but hesitate to use it because I am not sure whether it is the correct French word or whether it is an English word. False cognates can also cause some problems. For example, one mistake that beginning French students often make is to use the French word ‘personnage” to mean ‘person’, but in French this word means ‘character’ as in a character from a story. In Arabic I do not find problems such as these because there are very few cognates between English and Arabic.

Another interesting criticism of the contrastive analysis hypothesis is the case of interlingual errors. These are the types of errors whose source comes from within the target language. For example, overgeneralizing a rule such as the plural ‘s’ and saying ‘mens’ instead of ‘men’. What is interesting to me about this type of error is that it shows how the regularity of the target language is used by the learner. This will lead to some errors, but overall the regularity of the target language will facilitate learning. Most of the time applying the rule of adding an “s’ to form the plural will work. In my experience learning Arabic I have found regular aspects of the language which are quite different than English, but nevertheless make certain aspects of the language easier to learn. For example, Arabic is a language which relies on roots of three or four consonants. These roots have general meanings, and particular words are formed by adding patterns of vowels and certain consonants. The patterns themselves are derivational so also have a meaning. An example of this would be the root “k,t,b” which carries the meaning “write”. Words derived from this root are “kaatib” writer, “maktab” office, “maktaba” library, “kataba” to write. There are several other words with the same root. This pattern of deriving words from cononant roots is very different from English, but is one aspect of the language which makes vocabulary acquisition much easier. It also allows a student to guess the meaning of an unknown word.

While French vocabulary is more similar to English because of the many cognates, sometimes it is more difficult to learn because of the “subtle differences”. Arabic vocabulary shares few cognates with English but the root system of the Arabic language facilitates vocabulary learning even though this type of system does not exist in English.

natural order hypothesis

In his article Krashen’s Monitor and Occam’s Razor Gregg points out some of the problems associated with his natural order hypothesis, as well as his lack of discussion as to why such an order exists. Most of the research this hypothesis is based on are studies of the acquisition of certain morphemes such as the third person, plural, and possessive –s, or the past tense –ed. One problem with these studies which Gregg points out is that these studies are examining formally dissimilar items which may not be comparable. In addition to this criticism the methods used to elicit samples from learners have also been criticized. The method used in most of these studies if I remember correctly is what is called “obligatory occasional analysis”. It is a method of elicitation which requires learners to supply the correct forms in obligatory occasions. One of the problems of this method is that it cannot account for overuse of these morphemes. For example, a learner may produce the correct form on an obligatory occasion but also produce it incorrectly in other occasions. Therefore it is difficult to tell to what degree the learner has actually acquired the morpheme. In addition to this problem these studies have also been criticized on the basis of what has been referred to as the “comparative fallacy”. This is based on the idea that a student’s interlanguage is a distinct from the target language with its own rules and forms. Interpreting the learner’s language in terms of the target norm may actually obscure our understanding of the interlanguage as a distinct system.

Despite these problems with the morpheme studies I became a little interested in the fact that certain morphemes such as the third person –s seem to be acquired rather late despite their frequency in the input. I became more interested in this subject after I read an article last week by Helen Bird et al. (Helen Bird, Mathew A. Lambon Ralph, Mark S Seidenberg, James L McClelland, and Karalyn Patterson, Deficits in Phonology and past-tense morphology: What’s the connection, Jouranal of Memory and Language 48 (2003) 502-56)

This was a study which examined the evidence for what is called a double disassociation between Wernicke’s and Broca’s aphasics in the formation of irregular and irregular past tense verbs. Previous studies had found that Broca’s aphasics are poor at producing regular past tense verbs as in walked and better at producing regular past tense verbs such as bought. Wernicke’s aphasics on the other hand have more difficulty producing past tense irregulars and little difficulty in producing regular forms. This has led some to posit two separate mechanisms for the formation of the past tense, one for regular forms and another for irregular forms. What this study looked at was whether phonological complexity could account for the problems in Broca’s aphasics ability to produce regular forms. By phonological complexity the authors looked at the consonant and vowel structure of the words as well as the consistency of voicing. For example the past tense irregular bought has a CVC structure while moved has a CVCC structure as well as a consistency of voicing on the last two consonants. The authors controlled for these features, phonological complexity and consistency of voicing, and found that the apparent advantage in the production of the irregular past in Brocas aphasics disappeared when these factors were controlled. This lead the authors to argue that there are not two separate systems responsible for forming regular and irregular past tense forms, but one system which draws on phonological and semantic knowledge.

After I read this study I became curious if there was similar work done on the morpheme acquisition in ESL. It would seem reasonable that certain morphemes are more difficult to acquire because of their phonological complexity, such as the –ed forms, or because of the consistency of voicing which make them less salient. For example, when we form a past tense if the final consonant is not voiced the –ed is also not voiced producing two sequential unvoiced consonants as in Walked /walkt/. According to the authors of this study this would make the –ed morpheme less salient phonologically.

I have not had time to do a thorough search of the literature, but I did find one study which examined this question among others. It was a study done by Goldschneider and Dekeyser: (Jennifer M. Goldschneider, Robert M. DeKeyser (2005) Explaining the “Natural Order of L2 Morpheme Acquisition” in English: A Meta-analysis of Multiple Determinants, Language Learning 55, 27-77)

This study examined whether five factors, perceptual saliency, semantic complexity, morphophonemic regularity, syntactic category, and frequency, could account for the order of acquisition of several morphemes. Their results indicated that the combination of these properties could explain to a large extent the acquisition order of the morphemes they included in their study. They suggest that these factors be grouped into a broader conceptualization of “salience” along with L1 transfer (which they did not include in their study) may account for most of the varience in the order of acquisition of the morphemes.

I don’t have a lot of time to go into detail of how they define each of the five factors which make up “salience”, and there may be some limitations with this type of study. They use what they call a meta-analysis which pools data from several other studies. I have a limited knowledge of research methods, so I don’t know how valid this type of research is. I did think it was an interesting study and wanted to share it with anyone who might be interested.

Krashen

The other authors in our readings did a pretty good job of pointing out many of the problems with Krashen’s theory of SLA. What concerns me the most with his ideas is how much he downplays the roles of production and practice. If his hypotheses are correct there is little need for students to ever speak. Students only need to have ‘low affective filters” and receive interesting and meaningful comprehensible input which is just a little above their current level, and perhaps after a “silent period” of a month to several years the student will decide to produce what he has acquired.

This view doesn’t really account for my experience in acquiring second languages, or the experiences of anyone else I know. Comprehension and production seem to me to be somewhat different processes. Most people are able to comprehend much more than they can actually produce. There is also the case of Broca’s aphasics whose comprehension is largely intact, but who have enormous difficulties in production. Of course comprehension most likely helps acquisition, but I think it more likely contributes not by focusing solely on meaning, but by paying attention to both meaning and form and in some way rehearsing constructions in our memory. For example, I don’t really have very much difficulty focusing on the meaning of what is being said as well as its form while I listen to lectures in French. When I hear a construction which is somewhat familiar to me It gets reinforced, and when I understand something produced in a new form (with the aid of extra-linguistic knowledge) I sometimes explicitly focus on the new form and attempt to store it in my memory.

It doesn’t seem reasonable that comprehensible input is “the true and only causative variable in second language acquisition”(Krashen 1983). I have always felt that producing and practicing the target language has allowed me to both practice learned forms and make them more automatic. Speaking also allows me to identify areas in my language proficiency which need more attention and those which I have little difficulty. For example, I may explicitly know how to form a passive sentence in Arabic, and I may also be able to automatically recognize when a passive is not formed correctly. When I speak, however, I may not form the passive correctly. I can immediately identify my own mistake, and realize that I have not completely mastered this construction, and through further practice make its production more automatic. If I never attempted to produce the passive I would not really know the degree to which I had acquired it.

In foreign language contexts students have very little opportunity to practice and test their linguistic skills outside of class. I am not necessarily opposed to allowing a short “silent period” while the students get more comfortable with the classroom routine, but I think production, making mistakes, and learning from them are essential to learning.

Wolf Children

As I was driving home from class I thought more about the significance of feral children to the Critical Period Hypothesis in first language acquisition. I was thinking about the problem backwards. Scientists who develop a hypothesis such as the CPH should not be interested in finding evidence which confirms the hypothesis, but evidence which could possibility falsify it. Because the discovery of feral children is rare, each new case provides an opportunity to test the CPH. It is because of the possibility of falsifying the CPH that every discovery of a feral child should concern anyone one who studies language acquisition, not because they will lead to confirmation of the CPH (confirmation is not the business of science) but because it is a new opportunity to test a hypothesis

Critical Period Hypothesis

I was a little surprised to discover at first just how controversial the idea of a CPH or “sensitive period” is in second language acquisition. After reading the articles, however, there appears to be little disagreement that younger language learners generally outperform adult beginners in the long run, and that it is common for child beginners to acquire native or near native proficiency in an L2 while it is rare or unattested for adult beginners (Marinova-Todd et al. 2000 p11). Most of the disagreement seems to be on what factors explain the general success of child learners and general failure of adult learners to acquire near native levels of proficiency and competence. The CPH and SPH are biological explanations while others such as Flege (1987) point to differences in affective and situational factors as an alternative explanation.

I am a little skeptical that affective and situational factors alone could account for the linear correlation between age of arrival and proficiency of children and the variability and lack of correlation for age of arrival and proficiency shown in studies such as Johnson and Newport study (Brown and Gonzo 75-112). It seems unlikely that children uniformly experience similar affective and motivational characteristics which lead to their success while different affective factors lead to general failure and invariability in adult populations. It would seem likely that there are many adult second language learners with high motivation who are less fearful of taking risks while there are many children who may have low motivation and are more fearful of making errors in production.

Anecdotally I have several friends, mostly Arabic speakers, who began learning English as adults, and who are married to non-Arabic speaking Americans, speak only English at home and work, and who appear to be highly motivated language learners. Despite the fact that they have lived here for a decade or more and are very proficient English speakers, they cannot pass for native speakers.

Whether or not there are biological maturational constraints on second language learning I found that it was difficult to make comparisons between the different articles assigned for class. For example the Flege article examined phonological aspects of L2 learning in both perception and production, the Johnson and Newport study looked at the ultimate attainment of grammatical competence (not performance), and the Wang and Kuhl article looked at the rate of attainment between children of different ages and adults in the perception of Mandan tones after a period of instructed learning. The problem with comparing these different approaches is that they are looking at different aspects of language and language acquisition. It is possible that there is a difference between instructed and naturalistic learning, so that the results of a study such as Wang and Kuhl may not be generalizable to a population which has learned in a more unstructured environment. That there is no difference or a difference which favors adults in more structured environments is not necessarily evidence against the Critical Period Hypothesis. It would just show that adults are more adept in more structured settings. Another problem with the Wang ad Kuhl study is that they are looking at the initial rate of learning and not the ultimate attainment of the learners. According to the Marinova-Todd et Al. article this distinction has been recognized with regard to the acquisition of morphosyntax for some time, citing the Krashen et Al. review (1979) which showed that older learners initially acquire a language faster than child learners, but that the younger learners will eventually attain greater proficiency levels. (Marivona-Todd et Al. P 12) In addition to this study the Flege article cites a study by Snow and Hoefnagel-Hohle which showed that adult accents may be judged more accurate than a childs after a short period of learning, but that the reverse is true after a longer period.