Conjunctive and disjunctive concept learning in humans and squirrel monkeys.

Abstract

Each of 40 humans and 16 squirrel monkeys learned both a conjunctive and a disjunctive concept in a choice procedure in which a positive and a negative instance were presented on each trial. There were 2 relevant and 2 irrelevant dimensions. Humans found disjunctive concepts more difficult, but this trend was slightly reversed for the monkeys. Monkeys were more influenced by the nature of the relevant dimensions than were humans. Vincent learning curves showed improvement prior to criterion in all cases, in agreement with the Trabasso-Bower model for 2-cue concept learning. Results were discussed in terms of hypothesis testing and discrimination learning models of conceptual behaviour.SEVERAL RECENT STUDIES of human concept learning have established that disjunctive concepts (e.g., "an instance with either red or a circle positive") are more difficult to learn than conjunctive concepts (e.g., "all the red circles are positive"). The differential difficulty of these concepts has typically been explained in terms of strategies and hypothesis testing (HT) processes, and there is evidence that subjects are differentially familiar with hypotheses of various kinds (Haygood & Bourne, 1965). Other explanations of these results may, of course, also be tenable.The procedure in which instances are serially presented has often invited comparison between HT explanations and models derived from S-R learning positions, which see concept learning as a complex application of discrimination learning (DL) processes. Both Hunt (1962) and Kendler (1964) conclude that learning models have been quite successful in handling a number of concept learning situations. But they have not generally been applied to the learning of complex concepts-those having more than one relevant dimension. Hunt (1962, p. 57) suggests that an S-R model would predict rather slow learning of both conjunctives and disjunctives. The prediction concerning their comparative difficulty is not clear. The conjunctive requires a subject to discriminate a pattern of two cues from the separate components and from the absence of either component. The disjunctive requires attaching the same response to two

Proficient performance of a conjunctive, recursive task by an African gray parrot (Psittacus erithacus).

“…Tasks that require a subject to associate multiple different, specific signs for each discriminative stimulus are, according to Thomas (1980), misidentified as conjunctive. Subjects in such studies learn to respond, for example, to “A and B, not A and C” (e.g., Rescorla, 1981; Wells & Deffenbacher, 1967). Such specific responses do not demonstrate sensitivity to the abstract (and flexible) logical combinatorial concept basic to conjunction (Thomas, 1980).…”

Conceptual rule learning and chronological age.

“…Indeed at some young enough age, rule differences might vanish altogether. King (1966) and Wells and Deffenbacher (1967) have reported data suggesting this possibility. If young children tend to emphasize one type of stimulus, as implied in the latter two hypotheses given above, which are also essentially associative in nature, one would anticipate rule difference.…”