Tracy S. Kendler

Reversal and partial reversal shifts with verbal material

A series of four studies was conducted to investigate the influence of mediating representational responses on reversal and partial-reversal shifts involving verbal material. The general assumption tested was that the speed with which a reversal shift is executed depends on the accessibility of appropriate mediating representational responses. In Exp. I Ss were required initially to sort eight items into two categories and then during post-shift training to either switch the sorting response for every item (reversal shift) or for only half of the items in each category (half-reversal shift). Two kinds of verbal items were used: eight words, each one of which was an instance of one of two concepts, and eight unrelated consonant trigrams. With the conceptual words the reversal shift was executed more rapidly but no difference between a reversal and half-reversal was found with the trigrams. It was also discovered that up to 24 post-criterion trials of overlearning had no effect on post-shift behavior with either kind of verbal material. Experiment II found that overlearning up to 40 post-criterion trials had only minimal effects on reversal shifts involving trigrams. In Exp. III four kinds of shifts (one-quarter, one-half, three-quarters, and full-reversal) were compared with both conceptual words and trigrams. The full-reversal with conceptual instances proved to be the easiest. No significant difference existed among the other seven groups. Experiment IV, in which Ss initially sorted the conceptual instances into mixed categories and then shifted into either conceptually consistent categories or a reversal shift, showed the latter to be more difficult, suggesting that the accessibility of the representational response was more important to post-shift behavior than the reversal shift per se. The relevance of these results to mediational S-R theory and the verbal-loop hypothesis was discussed.

The development of discrimination learning: a levels--of--functioning explanation.

Publisher Summary A substantial increase in discrimination-learning efficiency between early childhood and young adulthood is demonstrated in a cross-sectional design that compares five age levels between 4 and 18 years on the optional-shift procedure. These data are consistent with the assumption of a quantifiable, longitudinal ontogeny of learning in which the rate of change decreases gradually as a power function of age. The learning ontogeny was explained by proposing two problem-solving modes. One mode is associative, incremental, and usually relatively slow; the other is cognitive and usually relatively fast. The developmental change is accounted for by assuming that the probability that a given child will respond in the cognitive mode increases gradually over age.

Developmental Processes in Discrimination Learning

A review of the discrimination-learning behavior of children indicates that two general theoretical orientations have been mainly responsible for guiding recent research, the mediational-attention con

Optional shifts of children as a function of number of training trials on the initial discrimination.

Abstract When nursery school and third grade children are trained on an initial discrimination for a fixed number of trials (either 16 or 36) before being presented with an optional shift, third graders are more likely to make optional reversals than nursery schoolers. When 16 and 36 acquisition trials are compared, no significantly different effect on optional reversals is observed. The results accord with a mediational explanation of the relation-ship between the reversal behavior of children and their age and learning speed.

The effect of training and stimulus variables on the reversal-shift ontogeny

Abstract Linear functions fitted to the proportion of optional reversal-shifts as a function of log CA were used to assess the effect of stimulus and training variables on the reversal-shift ontogeny. The results were: (1) when the dimensional salience of the two sets of cues in the stimulus compound were similar, there was no significantly different effect of size, color, or form cues on the rate parameter, suggesting that the rates of mediational development for these cues are similar. Reversal-shifts were, however, more likely with the color-form compound than with the size-color compounds throughout the tested age span, indicating that some stimulus conditions are more likely than others to produce mediated processing at all age levels. (2) The facilitative effects of labelling on reversal-shifts were greatest at the younger ages and decreased systematically over age. (3) Overtraining increased reversal-shifts at most ages but whether the facilitative effect interacts with age remains moot. (4) Neither labelling nor overtraining eliminated ontogenetic differences.

World views and the concept of development: A reply to Lerner and Kauffman.

Abstract I reply here to the argument of R. M. Lerner and M. B. Kauffman (1985, Developmental Review , 5 , 309–333) that an adequate concept of human development is incompatible with a mechanist “world view” but rests instead on a principled integration of contextualist and organicist “world views.” I review how each of these metatheoretical positions is described by the philosopher who proposed them and conclude that the version of contextualism and organicism presented by Lerner and Kauffman is so diluted as to lose the essence of their original meaning. In consequence, the concept of development they propose, which includes the notion of integrative levels, causal variables that interact differently at different times in the course of ontogeny, and probabilistic outcomes is more compatible with the mechanistic metatheory they eschew than with the contextualist and organismic ones they ostensibly espouse.

Cross-Sectional Research, Longitudinal Theory, and a Discriminative Transfer Ontogeny

This review argues that an hypothesis about longitudinal development may be derived from, and tested exclusively by, cross-sectional data, when the potential sources of interpretive errors are dealt w

Lower level encoding: Holistic or nonselective?☆

Abstract Two experimental procedures, discriminative learning and classification, reveal a monotonic increase between early childhood and young adulthood in the disposition to encode visual input selectively. This paper compares the capacity of two theories to explain this ontogeny. One theory cites perceptual differentiation as a major mechanism; the other cites transfer of control from a lower, analytic, nonselective encoding level to a higher, selective level. This essay supports the levels theory by showing how a model that assumes young children are disposed to encode analytically and nonselectively can (1) explain data heretofore considered as supporting the perceptual differentiation theory and (2) predict the outcome of two new critical experiments.

Dominance as a function of cue similarity: A test of the intrinsic cue dominance model ☆

Abstract A nonselective model postulating intrinsic cue dominance was tested in simultaneous discrimination tasks involving reversal on one dimension. In this procedure two dimensions are relevant throughout training; however, following initial discrimination training the reward contingency is reversed for one dimension but maintained for the other. Cue dominance was assessed following acquisition of reversal by the use of opposed-cues test trials, and was defined as a greater number of choices of the test compound containing the positive cue of the reversed dimension than of the test compound containing the positive cue of the maintained dimension. In Experiment I, brightness cues dominated orientation cues. In Experiment II, which employed two different sets of relevant cues, more disparate brightness cues dominated the orientation cues for one set and orientation cues dominated less disparate brightness cues for the other. From this, it was concluded that dominance is a function of relative cue similarity.