Fossilization, automatization and second language acquisition research

Excerpts from “The Neurobiology of Learning”
by John H. Schumann, Sheila E. Crowell, Nancy E. Jones, Namhee Lee

“Aphasic syndromes caused by BG (basal ganglia) lesions indicate what roles the BG may play in language functions. According to Fabbro (1999), basal ganglia aphasics develop symptoms such as reduced voice volume, foreign accent syndrome, perseveration (involuntary repetition of words, syllables etc.), and agrammatism. Additionally a polyglot’s more fluent language tends to be more seriously damaged than a less fluent language. The interesting fact that the patients’ second languages are better preserved may imply that their second languages are processed more by the declarative memory system. Although the automatization of a second language through the BD is ongoing, it may not be complete. When a patient suffers a BG lesion, the parts of the second language that have already been proceduralized wil be damaged, but other parts of SL that have not been proceduralized will be preserved. In contrast, a first language may have been almost completely proceduralized without leaving much of a trace in the declarative system. This may be why BG-lesion patients cannot produce their first language in spite of their intact declarative memory system...

Knowledge about the BG functions may have important implications for the area of linguistics in general and SLA in particular.

Learning fixed expressions: Chunking

Some researchers have noticed that second language learners tend to learn frequently co-occurring words and delexicalized chunks (Sinclair 1991; Tannen, 1989). This phenomenon may be explained by the chunking mechanism of the BG. Previously, we discussed how the BG participates in the process of chunking the cortically distributed information into a unitary sequence through convergence, divergence and reconvergence. Whenever a second language speaker uses one of the fixed expression, he or she may simply activate the relevant basal ganglia circuit so that he or she does not need to apply a grammar rule or a phonological rule step by step.

Automatization of Syntax and Phonology: DP and IP

Learning and producing the phonology and grammar of a target language probably involve both the direct pathway and the indirect pathway. Through numerous repetitive inputs of the target language and its production, a second language speaker may slowly build up stronger synapses among participating neurons in the cortex and basal ganglia, which represent the syntactic and phonological rules of the target language. Finaly the learner acquires the ability to execute the rules through the direct pathway of the BG. For example, the choice of word order may be the result of basal ganglia function.

Whenever a second language speaker utters a sentence, perhaps there may be two competing word orders in the speaker’s brain, one probably from his or her first language and another from the target language. When the speaker gets into the target language mode,the target language order may be executed through the direct pathway with the competing order being inhibited by the indirect pathway. Other aspects of grammar may be the same.

Phonology is likely to develop in the same way… As the learner improves his or her fluency in the target language through numerous repetitions in listening and speaking, he or she may acquire the ability to execute this rule through the direct pathway.

(Comment: Pronunciation does improve with language use, regardless of the initial silent period. Even after a long silent period the learner's pronunciation will suffer while he struggles to build sentences. However, unlike the early bird, he will always be able to rely on his good ear for the language. If one insists on speaking from the very start and without paying attention to phonology, automatization results in a lot of fossilized errors and fossilization is unfortunately the strongest and most difficult to eradicate at the phonological level. I’ll leave the silent period hypothesis aside. One also needs to consider the usefulness to effort ratio and the “permanent damage” theory.)

Formation of Rules of the Target Language

The formation of correct rules is often a difficult process. To form a correct rule, a speaker has to frequently execute the correct sentences related to the rule. However, a beginner cannot execute the correct sentence easily, and every time he or she executes an incorrect sentence, the wrong rule will be strengthened in the relevant neuronal circuits. A paradoxical situation is unavoidable here. The more often a beginner utters incorrect sentences, the stronger the neuronal circuits representing them may become. However, advanced second language speakers conform to the rules of the target language to a greater extent than beginners.”

(Comment: certainly a lot of wasted synapses)

“Fossilized language speakers have two important characteristics (Harley and Swain, 1978; Selinker, 1972). One is that they have already acquired a certain level of communicative fluidity. They can generate utterances in the target language without undue cognitive planning and without consciously building structures. They show less hesitation when engaged in conversation. In summary their speech has fluency. Another characteristic of fossilized second language speakers is that their learning has stopped or radically slowed down. Their typical utterance structures and phonology do not improve over time although they may be continuously exposed to the target language environment. They continue to make the same grammatical and phonological errors although they are sometimes aware that they are doing so.

(Comment: In second language acquisition the mighty brain is not always your friend. Message transmitted, message received - the path of least resistance. The native error correction mechanism also contributes negatively since the learner is not encouraged to reformulate his utterance. A possible workaround: a decent silent period, careful production).

“These two characteristics may be explained by BG functions and procedural memory. The first characteristic of fossilized second language speakers, natural fluidity, occurs because they have already acquired the target language procedurally, thus, they have obtained automaticity. By repetitive use of the target language, the speakers may have formed procedural memory of (incorrect) linguistic rules of the target language through the basal ganglia circuits. When one acquires a procedural memory of a motor or a cognitive skill, one can execute it automatically…

The other characteristic of the speakers, rigidity of errors, can also be explained with reference to the BG and procedural memory… Procedural memory is formed more slowly than declarative memory. The other side of the coin is that procedural memory is more robust so that, once formed, it is better preserved, and it is also inflexible, and therefore difficult to change. This is why it is so difficult to correct bad habits… If a fossilized second language speaker has already automatized the linguistic skills through basal ganglia circuits, the automatized skils are naturally resistant to correction and change.

An outstanding question is whether fossilized language can be defossilized… First, defossilization perhaps is possible. It is not too rare to meet fossilized speakers in a language classroom. (No kidding!).

This may be possible for two neurobiological reasons. First the brain is always plastic, although the extent of plasticity varies according to many factors. Because the brain maintains plasticity, it is not impossible to form a new rule or to correct an incorrect rule. Second, the anatomy of the brain shows that the procedural memory of the basal ganglia can be influenced by other components…. Dopamine (DA), which is involved motivational modulation of its targets, is very important in this system, projecting from the ventral…to the ventral striatum, the ventral pallidum and the dorsal striatum.
From experience, we all know that automatizing declarative knowledge or altering a habitual procedure is difficult and time-consuming. It requires practice and motivation to sustain that practice. Animals probably acquire declarative and procedural knowledge together as they experience the world. With humans, the symbolic species capable of language, it becomes possible to acquire declarative and procedural skills more separately. This type of learning requires cognitive work and the motivation to do that work. The task, of course, is facilitated by aptitude. From an evolutionary perspective, it is easy to understand why it may be difficult to alter motor procedures. Procedures are developed to help the organism thrive in the environment by allowing automatic responses to stimuli. If they were easily altered or disrupted, the animal’s survival would be threatened. Therefore, when a language learner develops incorrect grammatical structure, these habit-protecting difficulties are encountered, and considerable effort is required to develop the correct procedures to override the maladaptive fossilization”.

Krashen from neurobiological perspective

“Though Krashen himself did not attempt to make a biology-based argument, there are several possible biological assumptions inherent in his position. These are:

1 The areas of the brain involved in subconscious processes (acquisition) are different from those areas involved in conscious processes (learning). That is to say, declarative and nondecalrative learning are accomplished by different areas of the brain.

2 There are no connections between these two brain regions.

3. The declarative system cannot modulate activity in the nondeclarative system. In other words, practicing an explicitly learned rule over and over again will not help the learner to strengthen connections in areas of the brain responsible for proceduralization.

Currently, SLA theorists are moving away from Krashen’s noninterface position, and are taking the stance that rule acquisition in language is a complex cognitive task that lies on the same power function learning curve as other cognitive skills (DeKeyser, 1997). These researchers suggest that SLA is similar to the acquisition of most skills, which appear to involve interactions between the declarative and the nondeclarative memory systems (Berry, 1994, Ellis, 2000;MacWhinney, 1997). Elis, for example, discussed three likely ways in which implicit and explicit knowledge might be converted into implicit knowledge if the learner is at the right stage of linguistic development. Second, explicit knowledge may lead the learner to listen for a recently learned language structure in the input. Third, explicit knowledge might cause learners to notice differences between their own output and the output of native seakers (Ellis, 2000). These three points are not only borne out by observations of adult language learners, they are also true of the underlying biology. Perhaps only the first of the three points needs some revision based on the research presented in this book. Specifically, we would assert that knowledge that is stored declaratively is not converted into nondeclarative knowledge. Instead, learners acquire and store information in both declarative (hippocampus/cortex) loops and nondeclarative (basal ganglia/cortex) loops…

MacWhinney (1997) “cited a number of sources claiming that second language learning is facilitated by explicit instruction. However, he also suggested that implicit learing may still play an important rule in the acquisition of a second language. Further, MacWhinney suggested that explicit instruction may actually be harmful if the structures that are being taught are too complicated, irregular, or simplified to the point of being incorrect…

It is likely that there are significant individual differences in the number of cycles through the hippocampus that are needed for each student to learn a rule, as well as differences in the number of cycles required for different rules for the same student. While some students may immediately recognize the discrepancy between the two forms and rapidly begin producing target-like utterances, other students ay take weeks or even months to resolve this conflict. One reason for this difference is that each student’s brai has been shaped by idiosyncratice experiences. Furthermore, as was previously discussed, this difference between students may partially result from individual differences in the genes responsible for activating the transcription factors CREB and C/EBP.”