The event-related potential (ERP) correlates of sound detection are attenuated when eliciting sounds coincide with our own actions. The role of attention in this effect was investigated in two experiments by presenting tones separated by random intervals. In the homogeneous condition of Experiments 1 and 2, the same tone was repeated, whereas in the mixed condition of Experiment 1, tones with five different frequencies were presented. Participants performed a time-interval production task by marking intervals with keypresses in Experiment 1, and tried to produce keypress-tone coincidences in Experiment 2. Although the auditory ERPs were attenuated for coincidences, no modulation by the multiplicity of tone frequencies in Experiment 1, or by the task-relevancy of tones and coincidences in Experiment 2, was found. This suggests that coincidence-related ERP attenuation cannot be fully explained by voluntary attentional mechanisms.