Squirrel monkeys were given either forward pairings of a bite-tube CS and shock US or backward pairings of these stimuli. Backward pairings produced stronger control of biting by the bite tube alone than did forward pairings. In a second experiment, subjects received backward pairings of US and CS with either a fixed ITI or a random IT!. Conditioned biting was obtained only when trials were presented with a fixed IT!. The magnitude of unconditioned biting was also significantly greater with the fixed IT!. It was argued that these results demonstrate that conditioning in this situation depends upon the degree to which biting predicts a relatively long shock-free period. When trials occur randomly in time, biting predicts no definite shock-free period; hence, it is not learned.In the course of discussing methods for analyzing the effects of punishment procedures upon behavior, Dunham (1971) has suggested that certain procedures used to present aversive stimuli to an animal can establish an "implicit" avoidance contingency for the response which is typically elicited by the aversive stimulus. Consider, for example, an hypothetical organism which exhibits only two behaviors, A and B. If a periodic electric shock is introduced into this organism's world, a third response, E, elicited by the shock, is generally added to the response repertoire. The elicited response, because it is sequentially dependent upon the occurrence of the shock, initially enjoys the unique property of predicting the absence of shock for a longer period of time than any other behavior in the repertoire (either response A or B). This "implicit" avoidance contingency would, of course, be eliminated if the intershock interval were short enough to permit the elicited behavior to be punished by the onset of subsequent shocks.There are at least two phenomena in the contemporary literature in which this "implicit" avoidance contingency could play an important role. The first has emerged from the context of research concerned with the effects of extended exposure to periodic shock upon the aggressive behavior of the squirrel monkey; the second emerged from research which attempts to demonstrate that aggressive behavior can be classically conditioned to previously neutral stimuli.With reference to the first of the above cases, O'Kelly and Steckle (1939) were apparently the first investigators to report that brief, painful stimulation will elicit aggressive behavior between pairs of rats exposed to such stimulation. The generality of this phenomenon, typically called pain-elicited aggression, has subsequently been demonstrated in a variety of experimental