Labeling and memory effects on categorizing and hypothesizing behavior for biconditional and conditional conceptual rules.

Abstract

Differences in difficulty of learning conceptual rules have been explained within Bourne's truth-table strategy theory by the naturalness-unnaturalness of various assignments of instance classes to response categories. However, since this factor cannot explain why conditional rules are learned more easily than biconditional rules, two other factors were tested. Results showed that when 6"s were encouraged by neutral labeling of response categories (a) to learn the rule for both positive and negative categories and thereby (&) to assign to the negative category only one instance class for the conditional rule compared to two for the biconditional, they learned the conditional rule more proficiently. However, when 6"s were constrained by positivenegative labeling (a) to learn the rule for the positive category and thereby (b) to assign to the positive category only two instance classes for the biconditional rule compared to three for the conditional, they learned the biconditional more proficiently.In conceptual rule-learning tasks, biconditional rules have been more difficult to learn than conditional rules (Bourne, 1967;Haygood & Bourne, 1965). Bourne's (1967) theory that 6"s use a truth-table strategy to learn conceptual rules posits a two-stage process in which: (a) Ss learn to reduce the population of instances to four instance classes, i.e., the TT class, in which the relevant values of both attributes linked by the rule to be learned are present; the TF class, in which only the relevant value of the first attribute is present; the FT class, in which only the relevant value of the second attribute is present; and the FF class, in which the relevant values of neither attribute are present; (&) 5s learn the proper assignment of these instance classes to the positive-instance and negative-instance response categories. For the conditional rule, the TT, FT, and FF classes are positive instances, whereas for the biconditional rule the TT and FF classes are positive instances.

Phonological Concept Learning

“…Previous research found that naïve participants in inductive category learning experiments can verbalize rule hypotheses explicitly, either trial by trial (Bruner, Goodnow, & Austin, ; Conant & Trabasso, ; Peters & Denny, ) or in post‐experiment debriefing (Gottwald, ; Shepard et al., ; Smith, Tracy, & Murray, ). For example, participants trained inductively on two different logical rule schemas (e.g., X and Y vs. X or Y ) can identify which schema correctly classifies a new set of stimuli with different features (Haygood & Bourne, ; King, ), and participants trained inductively on a rule‐described pattern show positive transfer to inductive learning on a new stimulus set where the same rule is instantiated by different features (Bourne & O'Banion, ; Haygood & Bourne, ; Shepard et al., ).…”

Effects of directional and neutral category labels in bidimensional rule-learning problems

“…Thus, within each partition, half the Ss in the neutral label problem solve an analog of the primary rule problem and half an analog of the complementary rule problem. Peters and Denny (1971) hypothesized that the effect of neutral labeling was such that S focuses on the category containing the fewest classes of stimuli, whereas with directional labels, S assumes that the assignments to the positive category must be learned. The prediction must therefore be that leN), lI(N), and III(N), each of which incorporate a sorting category to which only one truth table class is assigned, should be equal in difficulty.…”

“…Seggie (1969), Peters and Denny (1971), and Gottwald (1971) have demonstrated that significant differences in performance can be obtained in the solution of selected bidimensional concept problems between conditions in which the labels of the response categories are directional and conditions in which the labels are neutral. Peters and Denny (1971) utilized a rule-learning task in which the directional labels were yes and no and the neutral labels were A and B, despite the fact that A and B have a strong natural ordering. The design also incorporated a procedure by which S was required to write the current hypothesis following each trial, and E corrected S if the hypothesis was not consistent with the relevant values.…”

“…Presumably, this rationale holds for all problems which are perceived as more complex than the conjunction rule problem. Since the higher "level of learning" demands learning all classifications, this would seem to place an upper limit on the difficulty of solving bidimensional problems, resulting in the qualitative distribution illustrated in Peters and Denny (1971), Seggie (1969), and Bourne and Guy (1968). Partitions are ordered by increasing difficulty of primary-rule problems.…”

Logical Complexity in Morphological Learning: effects of structure and null/overt affixation on learning paradigms

“…A few psychological studies (Peters andDenny 1971, Gottwald 1971) have investigated whether the difficulty of learning a pattern is affected by whether the same task is presented as (a) learning a distinction between examples that belong or do not belong to a category, or (b) learning the same pattern a distinction between two complementary categories A and B (similar to learning the distribution of two distinct affixes in a paradigm). Most relevantly for this paper, Gottwald (1971) considered learning of the four categories in table 1 in two conditions: biased labeling, in which instances were labeled as described in (a) above, and neutral labeling as described in (b) above.…”