Evaluation of hypotheses in concept identification

The aim of this paper is to investigate the nature of hypothesis evaluation in conceptual tasks, especially in the identification of bidimensional concepts. In such tasks, hypothesis testing is seen as being composed of sampling and evaluation stages. With complex problems only one hypothesis seems likely to be sampled on each trial, and it is suggested that this hypothesis is evaluated according to a statistical decision-making process. In Experiment I, Ss were given an initial hypothesis involving one of eight rules and required to test it in an attempt to find a concept. When this given hypothesis was true (i.e., the concept), no difference was found between rules for the number of instances selected to the criterion of solution. Moreover, there was a tendency for Ss to choose instances which were predicted to be positive according to the hypothesis under test. Experiment II examined the role of memory in hypothesis testing. Immediate recall of instances selected revealed no difference between true and false hypotheses. Both primacy and recency effects were evident in recall. The number of instances correctly recalled was more than expected by models of hypothesis sampling and evaluation, and this was attributed to Ss having a low criterion for recall without intrusions. Examination of intrusions suggested that Ss may have retained some, but not all, of the features of the stimuli selected. Some differences were found between rules and between positive and netative instances on recall. These effects were suggested to be due to different amounts of information processing when classifying each type of instance for each rule; the results of Experiment III supported this suggestion.     

Some deterministic models of concept identification

It is shown that deterministic models can compete effectively with stochastic models in summarizing concept identification behavior. Three groups of deterministic models are examined. Examination of individual learners' trial by trial behavior in a concept experiment shows: (1) One person exhibited behavior consistent with a Hypothesis Permutation (HP) model despite being a nonlearner who showed no evidence of improvement over a period of 24 trials. However, when all 50 persons studied in each of two treatment groups were examined, only 22 members of one group and 10 of the other showed no inconsistencies with deterministic local consistency assumptions. (2) Certain deterministic computer programs could find at least one satisfactory order for predicting all responses by 18 of the 22 consistent solvers and 6 of the 10 consistent solvers, respectively, in the two groups just mentioned. For these 24 persons, then, a less restrictive deterministic model is adequate than for the others. (3) Those 38 original members of the first treatment group who met a stringent learning criterion were compared with respect to predictions generated by stochastic and mathematized deterministic models. One deterministic model (RSS-U 9-state) is in some respects the best of the models examined, but this success is a partial reflection of estimating eight parameters from the data.     

Children's performance in simple and successive-reversal concept identification problems

In Experiment I sixth-grade children learned two concept identification (CI) problems, a simple and a successive-reversal problem, either sequentially (Se) or concurrently (Con). Retention tests given during the CI problems indicated no significant differences in stimulus or response retention for any of the conditions. CI performance was poorer in the Con condition than in the Se condition. Contrary to adult data and predictions derived from the Bower and Trabasso (1964) model, there were fewer informed errors in the successive-reversal problem than the simple problem.     

Alternative theories of concept identification among older adults

The purpose of the present experiment was to investigate some predictions of hypothesis testing and S-R association (frequency) theories regarding memory for intratrial events on a conjunctive concept-identification task. They have received extensive study with young adults but not with older subjects. The individual events under investigation were feedback, responses, hypotheses, and stimuli. Hypothesis-testing theory requires subjects to retain information concerning the correct hypothesis from one trial to the next whereas frequency does not. 75 subjects (60-70 yr. old) participated in the study. Subjects had difficulty in recalling the correct hypothesis stated on previous trials. These errors occurred on problems with negative response trials, not with incorrect feedback. The results contradict predictions based on hypothesis-testing models but are consistent with frequency theory. Unlike in the studies based on younger adults, present subjects did not recall the hypothesis very well under the conditions in which hypothesis testing was made part of the primary task.     

The effects of perceptual pretraining on concept identification and preference

The effects of perceptual pretraining on preferences and concept identification performance in five- and eight-year-old Ss were investigated. The results indicated that perceptual pretraining facilitated performance on a concept identification task with a non-preferred relevant cue for the five-year-olds, but not for the eight-year-olds. However, perceptual pretraining did not affect preference in either age group and Ss were as consistent in their preference responding before and after perceptual pretraining as was a control group that had not received perceptual pretraining. Preferences were highly consistent in both age groups over a one week period. Perceptual pretraining was interpreted to increase the usability of the less preferred dimensions for the younger Ss, and to leave the affective preference response unaltered in both groups.     

Three-mode factor analytic reference curves for concept identification

Tucker's three-mode common factor analysis is used to explore the intrinsic individual differences in a figural, disjunctive conceptual behavior task. The results reveal differential shifts in the structure of the individual differences during concept identification. Individual patterns can be described by five factors, each of which specifies the parameter vales for the shape of the individual performance functions. A three-way core further identifies the nature of the individual differences factors. These results indicate that averaged curves for the separate response measures would not have adequately reflected the individual data in this sample. The present findings are consistent with earlier work and suggest that a better approach to the study of complex human behavior is to identify the individual differences intrinsic to the situation before mapping antecedent influences on the dimensionality of the behavior.     

The effects of acoustic and semantic similarity on concept identification

Since both acoustic and semantic similarity exert an influence on memory, the role of memory in concept identification (CI) was investigated by varying the acoustic and semantic similarity of the stimuli used in the CI task. Varying acoustic similarity had no effect on CI, but CI was significantly impaired when the dimensions of a CI task were semantically similar.