A comparison of simple correction and functional feedback in schema learning

Abstract

The role of knowledge of results (KR) in a schematic concept formation task was clarified by manipulating the specificity of information provided during feedback. Ss learned to discriminate between pairs of random forms from two schema families. Verbal correction was compared with a functional KR procedure in which the schema prototypes were superimposed over test stimuli. In no case did verbal correction improve learning in comparison with functional KR. Differences in acquisition persisted in a retention test.The development of the ability to categorize stimuli without external sources of information or prior familiarization with the relevant schema has been termed schematic concept formation (Evans, 1967). Studies have demonstrated that schematic concept formation can occur without extrinsic feedback or knowledge of results (KR) (Evans & Arnoult, 1967;Rosser, 1967). Other studies, including those by Tracy and Evans (1967) and Brown, Walker, and Evans (1968), have indicated further that the ability to learn to categorize visual patterns is not appreciably facilitated by KR in the form of simple "right-wrong" correction.As a plausible explanation for the occurrence of schematic concept formation without extrinsic feedback, Edmonds, Mueller, and Evans (1966) have suggested that information in the stimulus patterns themselves is sufficient to allow selection of relevant attributes in a fashion analogous to cluster analysis. This interpretation appears to be in accord with the Gibsons ' (1955) proposition that Ss can learn to differentiate higher order variables without extrinsic KR.Less easily explained is the striking fact that when correction is administered, little or no facilitation has been observed. While numerous authors (e.g., Miller, 1966; Edmonds & Mueller, 1967;Tracy, 1971) have offered hypotheses to account for the lack of facilitative effects of correction, the explanation initially proposed by Breckenridge, Rankin, and Wright (1969) seems most tenable; that is, on tasks in which there is a moderate level of stimulus redundancy, learning is possible on the ba si s 0 f in trin sic information alone. Intrinsic information, however, is less than perfectly reliable. Extrinsic feedback in the form of correction, in conjunction with the less than perfectly reliable intrinsic information, may confuse S and make it difficult for him to determine accurately why he is correct or incorrect. As a consequence, his performance level is not greater than, and may even be less than, he could achieve on the basis of intrinsic information alone. The learning of appropriate attributes assumed necessary to make perceptual discriminations in a schematic concept formation task may therefore be unimproved by the administration of KR in the form of correction.The purpose of the present study was to clarify the role of KR as it is administered in a schematic concept formation task. It was hypothesized that the administration of KR in conjunction with moderately redundant stimuli would not confuse S if the feedback was made ...

Role of feedback, category size, and stimulus distortion on the acquisition and utilization of ill-defined categories.

“…The results of the present study seem inconsistent with those few studies that have purported to show that verbal corrective feedback is ineffective (Brown et al, 1968;Smallwood & Arnoult, 1974): "Extrinsic feedback in the form of correction, in conjunction with the less than perfectly reliable intrinsic information, may confuse the subject and make it difficult for him to determine accurately why he is correct or incorrect" (Smallwood & Arnoult, 1974,, p. 581). The results of the present study could hardly be less in accord with this view: Regardless of the degree of preexisting structure, feedback was markedly facilitative on learning.…”

“…However, the amount of learning is often slight and unlikely to asymptote at high levels of performance. For example, only 21% of the subjects performed better than chance in the free sorting task used by Evans & Arnoult (1967), and learning increments of less than 10% are often reported (Aiken & Brown, 1971;Smallwood & Arnoult, 1974).…”

“…In addition, feedback controls were often lacking, and the obviousness of categorical structure was not manipulated. For example, only 15-20 presentations were used in the oddity task by Brown et al (1968) and Evans and Edmonds (1966), and 40-60 stimulus presentations were provided in the same-different judgment tasks of Brown and Evans (1969) and Smallwood and Arnoult (1974). 1 1 A striking exception is the study by Bersted et al (1969).…”

“…In addition, even the small amounts of learning often reported may be misleading, because general task difficulty (and not the difficulty occasioned by the unavailability of feedback) may prohibit sizable increments of learning. Failure to manipulate the obviousness of categorical structure (Bersted et al, 1969;Evans & Arnoult, 1967;Smallwood & Arnoult, 1974), such as by varying stimulus complexity or distortion, delimits an assessment of the range of conditions that can be effectively learned in the absence of feedback.…”

“…In spite of these methodological limitations, many researchers have claimed that feedback is either unnecessary (Gibson, 1969) or even potentially harmful (Smallwood & Arnoult, 1974) to category learning. For example, Gibson and Gibson (1955) proposed that perceptual learning occurs by extracting available stimulus information, with external feedback playing little role.…”