Effects of extended training (nine sessions, 50 trials each) on two-way avoidance response latencies were studied. For each rat, auditory and visual warning signals (CS) were presented on separate trials, either according to a Gellermann series or in 25-trial blocks. Intermittent presentation of the two kinds of trials yielded shorter latencies and higher levels of avoidance performance in response to the noise CSthan in response to the darkness CS. Presentation of trials in blocks revealed stronger response-eliciting properties of stimuli presented during the second half of the session than of those presented during the first half. A significant decrease of avoidance performance in the early portions of the delay period, an index of inhibition of delay, was obtained on auditory trials presented in the second half of the sessions. Results indicate that strong fear of the warning signal is required for inhibition of delay of avoidance responses. Two-wayavoidance training typically results in a gradual increase in the proportion ofcrossing responses performed during the warning-signal period, and a corresponding decrease in responses during subsequent aversive stimulation. At the outset of training, the rat changes compartments in response to painful stimulation. Fleeing, an innate species-specific defense reaction of rats (Bolles, 1970(Bolles, , 1971, soon becomes the dominant mode of behavior to pain and also to the events signaling painful stimulation. The data showing that escape and avoidance response latencies do not change in the course of training conducted in the shuttlebox (Bolles, Moot, & Nelson, 1976) were in agreement with such an understanding of instrumental defensive responding.A number of experiments yielded results that opposed the assertion of Bolleset al. (1976) concerning the invariance of response latencies in shuttlebox avoidance learning. Shortening of the two-way avoidance response latency was noted during a single training session (Bond, 1981;Galvani & Twitty, 1978). In Starr and Mineka's (1977) experiments, the mean avoidance response latencyThe present research was supported by a statutable grant from the State Committee for Scientific Research to the Nencki Institute. We thank W. Lawicka and 8. Zemicki for their comments on an earlier draft of this paper. Correspondence should be addressed to K. Zielinski, Nencki Institute of Experimental Biology, 3 Pasteur St., 02-093 Warsaw, Poland (e-mail: ziel@nencki.gov.pl.).-Accepted by previous editor. Vincent M. LoLordo shortened by half a second between the first three and the last three avoidances in a series of 27 consecutive nonshock trials. In the W. R. McAllister, D. E. McAllister, Dieter, and James (1979) study, significant shortening of the avoidance response latency was observed during the second session in rats trained with strong, but not with weak, shock.It seems, however, that during extended training, more complex changes in avoidance response latencies occurred. In our recent experiments on rats trained in twoway avoidance, a sh...