Between blocks, lengthening or returning the delay

to its standard length brought about robust changes in temporal firing patterns, even though the rats occupied the same ABT-888 in vivo locations at comparable times in all trial blocks. These results show that retiming is not attributable to differences in behavior during delays of different lengths but, rather, is caused by altering a highly salient temporal parameter that characterizes the delay event. Combining these findings, changing the duration of the delay revealed that, while a minority of neurons encode absolute or relative time, the majority form qualitatively distinct representations when the critical temporal cue was altered, and most of these maintain the new patterns when the delay is shortened to the original length. In order to assess whether a neuronal ensemble tracked the passage of time at each trial period, we used a two-way ANOVA using factors lag and trial period

to compare the similarity of the population vector at different lags during the object, odor, and first 1.2 s (early) and last 1.2 s (late) phases of the delay period. This analysis revealed a main effect of lag (F(4, 20) = 34.74; p < 0.001), trial period (F(3,15) = 9.94; p = 0.001), and an interaction between the two factors (F(12,60) = 3.17; p = 0.002). Separate one-way ANOVAs confirmed a main effect of lag (all p values <0.002) and a significant linear component (all p values <0.03) Protein Tyrosine Kinase inhibitor such that the population vector became less similar as lag increased during all trial periods, indicating temporal coding throughout the trial. Furthermore, a comparison of the change in the similarity of the population vector between lag 1 and lag Parvulin 5 (ΔL) indicates that time is coded at higher resolution early in the

trial (F(1, 11) = 23.81, p < 0.001; ΔL for delay early and object compared to ΔL for delay late and odor in Figure 4B). We also conducted GLM analyses to directly compare the extent to which time and location influence firing during the object and odor periods; these analyses do not consider other behavioral variables. Unlike the delay neurons, the activity from almost three-quarters (72/99 or 72%) of the neurons active in the object period was best explained by space or time, but not both variables. For 43 (60%) of these 72 object neurons, the inclusion of space without time in the model provided a more parsimonious account of the data. In 29 neurons (40%), time by itself was sufficient to explain neural activity, and the proportion of these neurons was different than that explained by space (χ21 = 4.70; p = 0.03). For the remaining 27 out of 99 object neurons, activity was explained best by both time and space, and the STIC from 13 of these neurons favored time while that of 14 neurons favored space. The results obtained from neurons active during the odor period were similar.