Yi et al. investigated how interictal spikes (IS), which are brief abnormal discharges seen between seizures in temporal lobe epilepsy, can predict working memory. The authors used hippocampal recordings from epileptic mice during a delayed alternation spatial task, and the researchers found that the rate of IS did not predict performance; however, the location and the timing of the IS did. When the IS occurs broadly spatially during periods of running, it is associated with worsened performance. On the other hand, an IS clustered in the reward zones was associated with improved performance. Machine learning analysis also revealed that an IS during the reward period had larger waveforms than other IS across the maze. The analysis also showed that spatially clustered IS preserved hippocampal replay, which is useful for memory consolidation, while dispersed IS disrupted hippocampal replay. Overall, these findings illustrated that the spatial context of an IS, not the frequency, determines if an IS is harmful or helpful towards working memory performance.