Midnight-to-midnight average pollutant concentration has been widely used as the exposure index in many epidemiological studies. This index, however, does not take account of the variability of time of death, because it simply takes 24-h average concentration from midnight to midnight, and this is equivalent to assuming that all deaths happened at the end of the day. This assumption is clearly inappropriate in the real situation, and might be a significant information bias in the exposure assessment. In order to take account of the variability of time of death, a new exposure index, the number-of-death-weighted 24-h average, is introduced in this study. This new index is calculated for each subject by taking 24-h average concentration preceding the time of death. In this study, this new index is applied to the suspended particulate matter (SPM) exposure assessment in Japan, and compared with the result obtained from the conventional index, the midnight-to-midnight average. There are quite large differences in the SPM average concentration between these two indices. Moreover, risk ratio calculated by the new index at time lag 0, where the mortality data and the average concentration data on the same day are used, is a lot larger than that obtained by the conventional one. In addition to the difference in its magnitude, time lag structure (from 1-5 days) also shows the different patterns between these two indices. This study suggests that the conventional exposure index, the midnight-to-midnight average, may evaluate the risk of SPM incorrectly, because of its inadequacy of capturing the temporal variability.