main effect (F = 7.82; df= 5,80; p< .01) and the Blocks by Schedule interaction (F = 3.11; df= 5,80; p< .05) were significant; none of the effects involving measure attained significance. For the low-MA Ss,in addition to the blocks main effect (F = 7.92; df= 5,80; p< .01) and the Blocks by Schedule interaction (F = 5.31; df= 5,80; p< .01), the Blocks by Measure by Schedule interaction attained significance (F = 3.94; df= 5,80; p< .01). Subsequent analyses in terms of blocks and reward schedule, performed separately for the starting speeds and movement speeds ofthe low-MA Ss, indicated a significant Blocks by Schedule interaction (F = 8.02; df= 5,80; p< .01) for the movement speed dataonly; the corresponding interaction for starting speed was not significant (F = 1.53; df= 5,80;p> .10). D1SCUSSION The major results of the reanalyses ofthe Watson et al (1%7) data may be summarized briefly as fonows: (a) for high-MA Ss, asymptotic partial-reward superiority was evident on both starting speed and movement speed;(b) forlow-MA Ss, asymptotic partial-reward superiority was evident on movement speed but not on starting speed.These results confirm in full the predictions detaiJed earlier in this paper, and support the notion that an attentional factor is operative in the starting response but not the movement response. They further suggest that an attentional deficiency eliminates the performance increment resuIting from nonreward, when the response in question has a large attentional component. The present data do not permit determination of whether the attentional deficiency resuIts in (a) decreased reactivity to frustrative nonreward, (b) a relative fai/ure to establish the reward expectancy which is assumed to underlie frustrative nonreward, or (c) both. Further research should be directed toward this question.A second implication of these findings is that frustration can be specific to one of several contiguous responses (see Watson, 1968). The difference in performance between partial and continuous reward may be thought of as due either to primary frustration, which persists from a preceding trial to a subsequent trial, or to conditioned frustration, wh ich is elicited at the beginning of a trial. In the case ofthe low-MA Ssofthe present paper, who demonstrated the 50%-100% difference on movement but not starting speed, it is evident that conditioned rather than primary frustration was respon· sible, since primary frustration would have elevated both the starting and movement speeds ofthe 50% group.