The effects of 20 min structured relaxation immediately prior to retrieval were observed for word lists learned 48 hours previously. Error probabilities and response latencies were scored at acquisition and retrieval for paired associates at two levels of response dominance. Frontalis isometric muscles tension was measured just prior to and during retrieval. Relaxation markedly reduced EMG activity associated with muscle tension. This effect did not extend to the retrieval period. Error probabilities were significantly lower for retrieval preceded by relaxation but the interaction between relaxation and response dominance proved non-significant. These findings suggest that states of low arousal do not bias memory search to readily accessible information as appears to be the case for high arousal.