Three auditory identification experiments were designed to specify the prime-response retrieval model of negative priming (S. Mayr & A. Buchner, 2006), which assumes that the prime response is retrieved in ignored repetition trials and interferes with probe responding. In Experiment 1, shortly before (in Experiment 1A) or after (in Experiment 1B) the prime, a cue signaled whether participants were to respond ("go" trials) or not ("no"-go" trials) to the prime. Negative priming was found in either case. A "prime-response retrieval effect"--an increase in prime response errors to the probe targets of ignored repetition trials--was found for go trials only. In Experiment 2, prime trials with go cues always demanded a response, whereas the response to no-go trials depended on motor discrimination: For left- (right-) hand responses, the response had to be withheld (valid no-go); for right- (left-) hand responses, the response had to be executed (invalid no-go). The prime-response retrieval effect was present only for go and invalid no-go trials. This implies that execution of the prime response is a precondition for prime-response retrieval, whereas a response preparation plan and a response description in task-specific terms are not sufficient. (Contains 2 footnotes, 3 tables and 7 figures.)