The response of phasic neurons is often measured in terms of average spiking rates over arbitrarily selected time intervals. Averages taken over inappropriate intervals may severely reduce the information content of data, as we show here using response data from female American cockroach peripheral olfactory cells. We demonstrate that a 100-ms period around the peak response contains the best information for discriminating among odors at moderate to high concentrations. Further, the 100-ms post-peak response period contains the best information at low concentrations, as well as in situations where it is important to minimize errors in misidentifying the quality of an odor. Averaging the data over the full 0.5 s stimulus period degrades the odor separation qualities of the data.