Effects of composition of the positive category on concept learning.

Abstract

2 experiments were performed to clarify the reason that positive instanges-are superior to negatws'festances in conjunctive concept learning. In conjunction, all positive instances embody both relevant attributes (e.g., red and square). In nonconjunctive concepts, the positive-category includes other kinds of instances (e.g., inclusive disjunction: red or square or both). Using nonconjunctive rules, increasing the proportion of instances embodying both, relevant attributes improved performance with the proportion of positive instances held constant. Changing the proportion of positive instances had little effect when the composition of the positive category was held constant, confirming that it is -instances with both relevant attributes, rather than positive instances as such, that influence S's performance.

An algebra of human concept learning

“…Similarly, the parity effect is reasonable in and of itself, and indeed had been observed decades previously (Haygood & Devine, 1967;Hovland & Weiss, 1953;Nahinsky & Slaymaker, 1970). But it represents an additional assumption, because the Up parity preference does not in any way derive from a simplicity bias (again, complexity and parity are orthogonal); so it would be more parsimonious to do without it if possible.…”

The role of the experiment in science education

“…Moreover, even the findings of the many studies designed to clarify the conditions most favorable for the attainment of concepts do not support unequivocally one or another cluster of theories. Some of the more important conditions are reducing the number of irrelevant attributes (e.g., Bulgarella and Archer, 1962;Keele and Archer, 1967;Kepros and Bourne, 1966); increasing the saliency of the relevant attributes (Archer, 1962;Trabasso, 1963); presenting a great number of examples of the concept, in close contiguity, (Bourne et al, 1965), most of the examples being positive instances incorporating all the relevant attributes (Haygood and Devine, 1967;Hovland and Weiss, 1953), some being negative instances which help to define the boundaries of the concept (Huttenlocher, 1962;Wells, 1967) particularly when the concept is based on a disjunctive, conditional or another principle more complex than the conjunctive (Bourne and Guy, 1968). …”

Relevant redundancy in disjunctive concept learning

“…First, all of the studies have been done with affirmative or conjunctive concepts, to the exclusion of concepts based on rules other than conjunction, such as disjunctive and relational concepts. Numerous studies of nonconjunctive concepts have been done and, in many cases, the outcomes differ radically from those found with conjunctive rules (e.g., Bower & King, 1967;Haygood & Devine, 1967). Thus, it seems useful to demonstrate the existence of a redundancy gain with other than conjunctive concepts.…”