Is behaviourism creeping back?
By Mohamad N. Keramati
University of Ershad Damavan University – Tehran
Abstract
Recent models of language learning, like connectionism and competition models with reinforcement learning algorithms, inspired by one of the fundamental law of behaviourism – association between input and response – are now having their advocates. The question which strikes the mind is whether or not behaviourism is making a come back with the new models of language learning like connectionism. This study intends to find the answer to this question. The author first starts with the principles of behaviourism and reviews the strong impacts of behaviourism in education. Reviewing the principles of connectionism and competition model, the author argues that despite similarities between the new models and behaviourism, they are epistemologically different from behaviouristic view of language learning. It is further claimed that behaviourism, despite being discredited now, had never left the ground and their presence in the field has been very strong.
Research Paper
Background
Behaviouristic view on language learning and teaching, which had its roots in structural linguistics of Bloomfield and the general principles of language learning of behaviouristic psychology of Watson, Thorndike and Skinner, was a dominant learning and teaching approach for few decades in the USA during the 1940s and 1960s. Three general principles of language learning have been identified in this framework (Dakin, 1973): the law of exercise which indicates that language learning is promoted if the learner repeats the responses to the stimuli. Here, practice plays a fundamental role. The law of effect which places importance on reinforcing the correct or native like response and on correcting the non-native like responses. Here reinforcement, such as approval of correct responses strengthen the association and is necessary to learning. These two principles had been proposed by Thorndike. And, the principle of shaping which mentions that learning will be rapid and smooth if the complex behaviour are broken into smaller units and are learnt bit by bit. Behind all these principles is the assumption that language learning, like other types of learning, takes the form of habit formation (Ellis, 1994). At the heart of behaviourism, we have the classical conditioning of Pavlov – a habit is an automatic response to a stimulus, reinforcement of Thorndike “stimulus-response bond would be formed whenever a reinforcement followed emission of the response in the presence of stimulus” (Anderson, 1995) and generalization of Watson, a conditioned emotional response will become generalized, and operant conditioning of Skinner, a stimulus defines the situation for a response (Anderson, 1995).
Based on the disciplines which influenced behaviourist language learning, Lado (1957) proposed Contrastive Analysis (CA) and claimed that it is possible to predict and describe patterns that will cause difficulty in learning by comparing L1 and L2 systematically. He proposed that the “pedagogical material should be based on this kind of comparison”. Following principles of structural linguistics, CA proposed a bottom-up priority for analysis, moving from phonology to morphology to syntax. The same was claimed to be priority for language learning. Following principle of behaviourist psychology, the proponent of CA assumed that language acquisition involved habit formation in the process of Stimulus – Response – Reinforcement (S-R-R). In other words, learners imitate and repeat (law of exercise) the language they hear (linguistic input or stimulus), and when they are reinforced for that response (law of effect), learning takes place. That is, a permanent modification takes place in the behaviour. Another assumption of CA is the notion of transfer in learning which means that the prior knowledge or the old habit interferes with the new one. If it facilitates learning of the new item, it is called positive transfer and when it inhibits the learning of the new item, it is called negative transfer. According to Lado (1957) the easiest L2 structures are those which exist in the L1 with the same form, meaning and distribution and the most difficult L2 structures are those which have partial similarity, not equivalence, with those of L1. The CA approach was not adequate for the study of SLA, partly because of its link to the behaviourism whose central claims have been criticized seriously by Chomsky’s (1959) review of Skinner’s verbal behaviour and partly because the claims made by CA were not supported by analysing the learners’ errors. Although behaviourist views in language learning which emphasized only the external behaviours have been criticized, the new-behaviourists’ like Hull, Tolman, and Bloom recognized the internal characteristics of the learners.
New-behaviourism
The new-behaviourist, Hall , was the first behaviourist who recognized the importance of learner’s internal characteristics such as motivation. Anderson (1995) mentions that the main goal of “Hull and other theorists was to develop a systematic theory of classical conditioning to explain all behaviour – human and animal” (p. 16). The fundamental issue for Hall was to connect learning and motivation. He believed that learning was insufficient for behaviour and therefore we need to have motivation. Skinner (1953) echoes the same concept when he talks about the learners’ self-control and self monitoring used as reinforcers for learning. What Hall wanted to do – relating internal characteristics to external characteristics – was done by new behaviourist Benjamin Bloom. Bloom and Krathwohl (1956) linked internal and external behaviours in their very influential taxonomy of learning. For them learning is observable behaviours at three levels of cognitive, affective and psychomotor. The importance of cognition has even been recognized by new-behaviourist Edwin Chase Tolman as early as 1948. But, his finding was not recognized then, due to the Skinner’s argument that studying behaviour was more useful. The influence of new behaviourists in current educational practice like, learning outcomes, curriculum planning, assessment and teacher’s role in the classroom (Jordan, A., Carlile, O., and Stack, A., 2008) is both influential and undeniable. Bloom’s taxonomies, particularly in cognitive domain, which deals how internal knowledge can be revealed by external behaviour (i.e., learning a grammatical rule moves from recognising the rule to being able to use it appropriately in the right place at the right time) have been used extensively in curriculum development on how to sequence the instructional objectives. Relating cognitive development and behaviour has had direct implications for classroom teachers. They could construct activities to allow the learners to practise the behaviours and of course check these behaviours when they wanted to assess the learners’ performance (Jordan, A., Carlile, O., and Stack, A., 2008).
One of the goals of education has always been modification of behaviour or learning outcome, i.e., positive feedback of the teacher to young learners or certification at higher levels. Learners’ outcome refers to both external and internal behaviour, with external behaviour being more straightforward and easily observable, and internal behaviour being more complex and not directly observable. Learning outcome is explicit statement of what a learner can do if, for example, completes a course of study. Learning outcome has been one of the major contributions of behaviourism to the present educational practice, particularly in removing the earlier mystique and vague explanation of the same phenomenon (Jordan, A., Carlile, O., and Stack, A., 2008).
The influence of new-behaviourism in language assessment has also been influential. We know that authentic test task is the one which approximate the real life task or Target Language Use (TLU) task (Bachman, L. and Palmer, , 1996) or is “the performance of behaviour stated in learning outcome under the same condition as those under which they were learnt” (Jordan, A., Carlile, O., and Stack, A., 2008). As an example, if we expect the students write a topic sentence, the test task should be in a way that the students actually write a topic sentence rather that being able to explain what the characteristics of a good topic sentence are. The implication of learning outcome in assessment has also led to the development of the criterion referenced testing where the learners are expected to achieve an already predefined learning outcome. Earlier norm referenced assessment compares the performance of the learners with one another and in relation to a norm. That is, the distribution of the scores or performance of the learners should be in a way to form a normal distribution. In other words, there will be some students who are below the mean score. Whereas in criterion referenced testing, it is possible that all learners meet the criterion and be judged as performing at a satisfactory level. The shift to criterion referenced testing is a surely of benefit to the learners. New-behaviourist principles have also been used in formative assessment where the role of feedback, seen as reinforcement, has been emphasized.
Behaviourist view on language learning is now discredited, but various elements of behaviourist thinking are making a come back. Even as recent as 2000, Eric Kandel won the Nobel Prize in Medicine for his discovery of the molecular basis of three related forms of learning: Sensitization, habituation, and classical conditioning (Kendel, et al., 2000 in Poirier, 2006). In addition connectionist models with reinforcement learning algorithms, inspired by Skinnerian reinforcement learning mechanism have now found support among scholars (Poirier, 2006). In the following section, a review of Connectionism and Competition model will be presented. There, the behaviourist features will be pointed out.
Connectionism
Connectionists like Piaget view mind as a module. But, they differ from Chomsky’s modularity in that for them language faculty is not a module. They distance themselves from Chomsky’s innatist approach to language learning too. Like behaviourist, they focus on the strength of association between stimuli and responses. Reviving parallel distributed processing (Rumelhart &McClelland, 1986) of psychology, connectionists believe that brain is like a computer that would consists of neural networks – complex clusters of links between information nodes. These links of connections become activated or weakened through activation or non-activation, respectfully. For them learning is seen as the process of association between the stimulus and response. They believe that human mind has an innate capacity to make these associations. “As learners are exposed to repeated patters of units in input, they extract regularities in the pattern; probabilistic associations are formed and strengthened” (Torike, 2005). So, for them the notion of innateness is not seen as an innate capacity to learn the abstract rules of language. To use the concrete example of Lightbown and Spada (2006) when a child hears a word in the context of specific object, event or person, an association is created in the mind of the child. The association will be stronger if the frequency of input and nature of feedback should be more. The child has an innate capacity to make these associations. Therefore, whenever the child hears that word, it brings to her mind that object and whenever she sees the object, it brings to her mind that word. Eventually any characteristics of that object or event may trigger the retrieval of the associated word. Language acquisition is, therefore, seen as a result of these types of associations rather than the construction of the abstract rules.
The claim that learning is not innate nor rule based is supported by computer simulations. In learning irregular verbs, it is known that children go through three phases: first they produce the correct form of irregular verb, i.e. went. In the second phase, they over-generalize the regular past tense ending to irregular verbs, i.e., goed, known as U-shape curve of learning for irregular verbs, and in the third phase, they produce irregular form correctly. Rumhelhart and Mc Clelland (1986) demonstrated that a computer that is programmed with a “patterned associator net work” can learn to associate English verb bases with their appropriate past tense forms without any a priori rules, and that it does so with much the same learning curve as the one exhibited by children learning their L1. Pinker (1991) argued that irregular verbs are retrieved from an associative memory, like what connectionists have described, but regular verbs are produced by learners as a result of suffixation rule. However, N.C. Ellis and Schmidt (1997) investigating the claim made by Rumelhart and McClelland (1986) using regular and irregular plural morphology, supported the claim made by Rumelhart and McClelland. They concluded that associative mechanisms are all that we needed in order to explain the acquisition of plural morphology. Their finding did not support Pinker’s finding that regular morphology of past tense was rule-governed and the irregular associative.
Competition model (Bates and MacWhinney 1981; MacWhinny 2001) of language learning offers a theory of perfromance in contrast with Chomsky’s theory of competence. Competition model challenges Chomsky’s deterministic model of language learning, innate hypothesis. Besides, they claim that language learning is non-modular and is not domain specific (Jordan, 2004). Relevant to the discussion in this paper is two of the theoretical commitments. The first one is the connectionist model which competition model uses to model the interaction between lexical mappings. There, they reject nativist view and argue that brain relies on a type of computation that emphasises patterns of connectivity and activation. The second one is that of input- driven learning. According to this commitment, learning is explained in terms of input rather than innate principles and parameter. Cue validity is the key construct in this explanation. The basic claims of competition model is that cues such as stress, intonation, rhythm, morphological marking, and word order are available in input and language processing involves competition among these cues. Different types of cues interact dynamically every time children or adults hear a sentence. Word order or first position of nouns is very strong cue for English Speakers.However, strong cue in one language might be weak cue in another. Transfer of L1 cues strengthen to L2 is something which is likely at early stages of SLA when the systems differ. Research in this issue is, however, not conclusive (Torike, 2005).
Conclusion
Connectionists are Piagtian in that they have modularity of mind in common with Piaget. But, unlike Piaget, they are innatist like Chomsky in believing that mind is predisposed to find the regularities, though their notion of innateness is different for that of Chomsky. Moreover, like Behaviourist they believe in strength of association between stimulus and response and they have an empiricist approach to learning. That they are not regarding learning as being innate, and the emphasize they put on input rather than innate distance them from Chomsky and make them closer to behaviourism. However, despite the fact that they reject innateness in the sense that Chomsky uses, I would like to argue that connectionists differ from behaviourists in fundamental ways: they consider causal explanation and try to overcome all theoretical bias.
I would also like to argue that despite the strong criticism on behaviourism, they have had important and powerful influence on education and learning. The two principles of the law of effect and law of exercise of Thorndike have their strong presence in the recent education. The presence of some of the features of behaviourism in new models of learning, like connectionism, is an evidence of the fact that exclusion of behaviourism from education is neither possible nor recommended. Learners and learning are so complicated that probably one model would not be able to accommodate all the complications. This reminds me of Romi’s story of the blind men describing an elephant. I hope the analogy does not offend the creators of the theorists of the models of learning. The problem lies in the size of the elephant which is as big as life itself. We are concerned with a complicated issue of learners and learning and it seems that each of these theories are shedding light to one dimension of these complicated phenomena. Rather than attempting to deny one theory over the other, perhaps, we need to merge these theories with one another to tackle the complex phenomena.
REFERENCES
Anderson, J. R. (1995). Learing and memory. NY: John Willy & son.
Bachman, L. and Palmer, . (1996). Language testing in practice . Oxford: Oxford University Press.
Bates, E., & MacWhinney, B. (1981). Second language acquisition from a functional perspective: Pragmatic, semantic and perceptual strategies. In H. Winitz (Ed.), New York Academy of Sciences.
Bloom, B. and Krathwohl, D. (1956) Taxonomy of Educational Objectives: The Classification of Educational Goals, by a Committee of College and University Examiners.
Handbooks 1 to 3: The Cognitive, Affective and Psychomotor Domain. New York: Longmans Green.
Chomsky, N. (1957). A Review of B. F. Skinner’s verbal Behavior. Language , pp. 26-58.
Dakin, J. (1973). The language laboratory and language learning. London: Longman.
Ellis, R. (1994). The study of second language acquisition. Oxford: OUP.
Ellis, N. C. and Schmidt, R. 1997: Rules or associations in the acquisition of morphology? The frequency by regularity interaction in human and PDP learning of morphosyntax .
Language and Cognitive Processes, pp. 307- 336.
Jordan, G. (2004). Theory construction in second language acquisition. Amesterdam,
John Benjamin
Jordan, A., Carlie, O., & Stack, A. (2008). Approaches to learning: A guide for teachers. New York, Mc Grow Hill.
Kendle, E. R., Schwartz, J. H and Jessel, T. M. (2000). Principles of neural science. New York: McGraw Hill.
Lado, R. (1957). Linguistics across cultures. Ann Arbor MI: University of Michigan Press.
Lightbown, N., Spada, N. (2006). How languages are learned. Oxford: OUP.
McWhinny, B. (2001). The competition model: The input, the context, and the brain.
In P. Robinson (ed.), Cognition and second language instruction. Cambridge: CUP.
Pinker, S. (1991). Rules of language. Science, pp. 530-535
Poirier, P. (2006). Behaviourism: Varieties. In Elsevier Encyclopaedia of language and Linguistics. pp. 714-719.
McClelland, J. L., Rumelhart, D. E. and. The PDP Research Group, (Eds.) (1986) Parallel distributed processing: Explorations in the Microstructure of Cognition. Psychological and Biological Models . Cambridge, MA: MIT Press.
Skinner, B.F. (1953) Science and Human Behavior. New York: Macmillan.
Tolman, E. (1948). Cognitive maps in rats and men. Psychological Review, pp. 189-208.
Troike, M. S. (2005). Introducing second language acquisition. Cambridge: Cambridge University Press.
1 comment