On inferring selective association: Methodological considerations

The theoretically appropriate means for demonstrating selective association are discussed and shown to be empirically necessary. Following the acquisition of an unsignaled instrumental avoidance baseline, dogs received either CS-contingent shocks (CS+) or random, independent CS/shock presentations. The CS was either a tone or a flashing light. When the CSs were subsequently presented during avoidance responding, only the tone-CS+ group showed absolute facilitation of response rate. However, both tone- and light-CS+ groups showed facilitation relative to their respective random controls due to the nonassociative inhibitory effects of the light. A bidirectionally sensitive dependent variable enabled the detection of this pattern of effects. Thus, a demonstration of selective association requires (1) appropriate controls for nonassociative effects and (2) selection of a dependent measure that is sensitive to both excitatory and inhibitory influences.