The third and fourth cells did not fire in block 1 and formed distinct timing patterns in block p38 MAPK pathway 2. Finally, the fifth cell was active both early and late in the delay in block 1, and its response to lengthening the delay was to maintain both times, one relative to beginning and the other relative to the end of the delay. Note that retiming typically did not occur immediately when the delay was increased. Comparisons of firing rates within the “time-fields”

across trials after the delay was increased showed that retiming did not happen immediately but occurred after a variable number of trials, either suddenly or gradually, in different cells (Figure S4). Neurons that showed absolute and relative timing, as well as retiming, were observed in simultaneously recorded ensembles, ranging 29%–54% for absolute and relative timing versus 45%–71% for retiming, suggesting that each neuron coded moments in the delay independently of the others. In addition, two

rats were returned to their standard delay during block 3, allowing us to assess whether neurons that retimed returned to the pattern of activity that was observed when the standard delay was reintroduced. Volasertib supplier The cross-correlation analysis indicated that most neurons (90%, 46/51) that retimed in block 2 maintained the altered pattern through block 3, similar to the hysteresis reported for partial remapping of place cells (Leutgeb et al., 2005a). The remaining five neurons appeared to return to a firing pattern in block 3 that resembled that in block 1. Examples of both types

of responses in block 3 are presented below in Figure 6B. One possible explanation for retiming is that the performance of the rat deteriorated when the delay was lengthened. For two rats, changing the delay had no apparent effect on performance, and this was confirmed by comparing performances in each block (two-sample t tests, all p values >0.17). A third rat did show a transient decrease in performance from block 1 to the first third of block 2 trials (two-sample t test; t58 = 3.25; p = 0.002). However, its performance recovered during the last two-thirds of block 2 (two-sample t test; t71 = 2.07; p = 0.04) and was otherwise stable throughout the recording session (for all remaining comparisons: two-sample t tests, all p values >0.18). Note also that, whereas performance for all rats was equally strong in blocks 1 and 3, when the lengths of the delays were equal, retiming that occurred in block 2 often persisted into block 3. Thus, retiming appears unrelated to changes in task performance. It is also possible that retiming might be secondary to changes in the locations the rat occupied during sequential time segments when the delay was lengthened. To address this possibility we compared second-to-second spatial firing rate maps for the early part of the delay across all trial blocks.