The contribution of experience on one task to performance on a subsequently different task

Teachers have appreciated transfer much before neuro-biologists did. The teachings of the "classics, when practised, were not necessarily for school children to become experts in Homer or Virgil. The idea was that such studies have a general disciplinary value that will be transferred to the intellect at large. This is why educational psychologists were driven to investigate transfer. They had to prove that some educational "methods do have merit. In doing so they have also developed basic research methodologies of tremendous value. Consider, for example, the use of "control groups: it was introduced in research on transfer in schooling (Thorndike and Woodworth 1901a,fc; Coover and Angell 1907). A brief example should suffice to illustrate this type of early studies: a group of schoolgirls in London were instructed to memorize a history text and subsequently tested for their "recall. The girls were then divided into two groups. Group A practised poetry over 2 weeks, whereas group B was engaged in working sums. Both groups were later subjected again to a history test. Although both groups got the same score on the first history test, group A fared much better on the final history test. The conclusion was that improvement gained by practise in memorizing one subject of instruction (poetry) was transferred to memory work in other subjects (history): '.the results do appear. to strengthen the case of those who wish children to learn much poetry; it is not an obstacle, but an aid, to the acquisition of other knowledge' (Winch 1908).

Transfer is said to be 'positive' if the experience on the first task results in improvement on the subsequent task, and 'negative' if that experience on the first task impairs the performance on the subsequent task (e.g. Woltz et al. 2000). Of course, experience on task i could result in positive transfer on task j but negative transfer on task k. As transfer is such a comprehensive term, qualifiers are used to specify the type of transfer under consideration. Only the two major uses of the term will be outlined here:

1. Transfer of training refers to the contribution of training on one "skill to the performance on a different skill. Some authors use the terms 'transfer of training' and 'transfer' interchangeably. However, in the behavioural literature 'transfer of training' conventionally implies intentional training on a particular overt behaviour, whereas 'transfer' extends also to higher cognitive processes such as the formation of concepts and hypotheses (see below). This distinction notwithstanding, the demonstration of transfer of training in even 'simple' "classical or "instrumental conditioning may also involve the formation of hypotheses and concepts (e.g. Schusterman 1962; "learning set). The study of Winch (1908), mentioned above, epitomizes a simple transfer of training experiment. Multiple protocols of transfer of training have been developed, and with them measures of the efficacy of transfer (Gagne et al. 1948; Murdock 1957; Hammerton 1967; Kolers and Roediger 1984). These measures commonly reflect better performance on the initial encounter with the new task, or 'saving' of training time or trials (on 'saving', see also "experimental extinction). Clearly, efficient transfer of training procedures is of great practical importance, as it could save time and money on training for specific jobs, either in the real situation or in simulators (Hammerton 1967; Hammerton and Tickner 1967).

In our discussion so far we have considered transfer of training as an intertask phenomenon. A somewhat different point of view propounds that even in what is formally considered by the experimenter as a single task, "recall is actually a 'retrieval task' that replays the

'"acquisition task'. As a consequence, performance on retrieval could benefit from transfer of experience on acquisition. This is further discussed as 'transfer appropriate processing' (Morris et al. 1977) under "retrieval.

2. Analogical transfer refers to the use of a familiar problem to solve a novel problem of a similar type (Gick and Holyoak 1983; Reeves and Weisberg 1994). The familiar problem is dubbed the 'source problem', 'base analogue', or 'base domain', and the new problem the 'target problem' or 'target domain'. For example, the popular depiction of the Rutherford model of the atom uses an analogy to the solar system (the base domain) to understand atomic structure (the target domain); the nucleus is depicted as the sun and the electrons as the stars (Gentner 1983). Much of our thinking is subserved by noticing similarities and analogies and generating "metaphors (Tversky 1977; Gibbs 1994). It is doubtful whether we could construe the world otherwise, although probably only a minority of scientists would agree with Nietzche that 'truth is a moving army of metaphors' (Nietzche 1873). By the way, 'metaphor' in Greek means 'transfer', in the colloquial meaning of'transfer'.

Over the years, a number of'theories', or "models, of analogical transfer have been advanced (Reeves and Weisberg 1994). They propose multiple stages in analogical transfer. First, there is the "acquisition of information about the base and target domains. This could occur long before the attempt is being made to solve the problem. Next there is the noticing of the base domain and its relevance. This critical stage echoes "insight. Next comes the application of the analogy to the target domain. If hints are provided, the situation is termed 'informed transfer'; otherwise, it is 'spontaneous transfer'. What type of information is used by the brain in doing all this? It helps to realize that problems can be defined at multiple "levels of information. The visible details of the particular problem are elements of the 'surface structure' of the problem, whereas the underlying abstract rules and principles comprise the 'deep structure'. Problems with different surface structures could share a deep structure. One aspect in which models of analogical transfer differ is their view of the relative contribution of 'surface' and 'deep' structures, or of data and "stimuli vs. rules, to the successful analogy (Reeves and Weisberg 1994; for a window to related debates in the cognitive sciences, see Gentner and Medina 1998).

A widely cited example will serve to illustrate the surface and deep structure. In this example, a problem entitled 'the radiation problem' is the target problem, and another, 'the general problem', the base problem (Gick and Holyoak 1983). The radiation problem is about a physician, who is faced with a patient with a malignant, life-threatening stomach tumour. It is impossible to operate on the patient. Radiation can destroy the tumour, but at the desired intensity the ray will destroy the healthy tissue on the way to the tumour. Lower density radiation is harmless to the healthy tissue but will not affect the tumour either. What shall the physician do? Now, here is the general problem: a country is ruled from a strong fortress by a dictator. A rebel army general leads his army to capture the fortress. Many roads lead to the fortress from different directions. All of them are mined. The mines are set to be detonated by the passage of large army units but not by small units. An attack by the entire army at once, which otherwise would capture the fortress, will detonate the mines. The solution: dispatch the soldiers in small units via many roads so that the entire army arrives together at the fortress at the same time. Subjects given the general problem were more likely to solve the radiation problem. Its solution: reduce the intensity of the rays but irradiate from several directions simultaneously (this is actually the correct clinical procedure). The two problems differ in their surface elements (physician, patient, hospital, vs. general, fortress, army, etc.), but share a deep structure (disjoining followed by "coincident convergence).

Transfer protocols are incorporated into a variety of studies on "engrams (e.g. Karni and Bertini 1997; Buckley and Gaffan 1998). They are used among others to measure the "generalization and extent of learning as a function of the activity (or damage) in a particular brain area. The identity of the brain areas involved depends on the paradigm used; for example, somato-sensory "cortex in motor skill. It is not yet established which brain areas are specifically critical for analogical transfer, although the frontal cortex is a safe bet.

A timely issue concerning transfer relates to the interplay between the extent of generalization of knowledge that is required to support transfer on the one hand, and the extensive specialization demanded from modern technological society on the other. To be effective, transfer must involve some degree of similarity, in data, procedures, rules, or cognitive "maps, between the source and the target tasks. Modern technology encourages more and more differentiation and specialization in order to master skills at an expert level. This may reduce substantially the overlap between old and new skills. How will it affect our ability to transfer knowledge from previous to future jobs? The problem surely calls for rethinking and identification of those cognitive skills that could promote efficient transfer of knowledge from what we know today to what we will have to know tomorrow. The impact on our educational and training systems might be profound.

Selected associations: Acquisition, Generalization, Learning set, Priming

0 0

Post a comment