Figure 4. Sequential, transient peaks in the firing rate span the delay period.
(A) Raster plot (top) and firing rate (bottom) for an example neuron (aligned to the onset of the delay period). Shaded area represents ± 1 SEM. (B) Heat-maps represent the z-scored firing rates for all units for the 10 s, 5 s, and 1 s delays, aligned to the onset of the delay period. Each row is a single unit. Rows in all three plots were sorted by the peak firing rate time in the 10 s delay trials (left-most plot). (C) Same as A, but spike times were randomly shifted relative to behavioral timestamps.
Figure 4—figure supplement 1. Presence of sequential activity when controlling for position, and when aligning to event markers other than delay period onset.
(A) Z-scored firing rates when aligned to first nose port arrival. Each row corresponds to one neuron. Left-most plot was generated by only using data from times that the rat was at the nose port. White area corresponds to when the rat was not at the nose port for a sufficient amount of time to estimate firing rates (fewer than 25 left trials and 25 right trials). Middle plot was generated using data from all times regardless of the rat position, but time bins were masked to match masking of left plot. Right-most plot was generated using all data without any masking. Rows are ordered based on the time of their peak firing rate in right-most plot. (B) Peak activity time when rat was in the nose port (left plot of panel A) was significantly correlated with peak activity time calculated using all data (middle plot of panel A) (r = 0.87, p<10–6 Pearson correlation test). Neurons that had their peak activity fall in the masked time bins (white area in middle heat-map of panel A) were excluded (n = 37). (C) Same as panel A, but aligned to delay period onset (sample lever press). (D) Same as panel B, but aligned to delay period onset (sample lever press). Peak activity time when rat was in the nose port (left plot of panel B) was significantly correlated with peak activity time calculated using all data (middle plot of panel B) (r = 0.74, p<10–6 Pearson correlation test). Neurons that had their peak activity fall in the masked time bins (white area in middle heat-map of panel A) were excluded (n = 35). (E) Histogram of peak times for 10 s-delay trials for non-shifted data (Figure 3B) and randomly shifted data (Figure 3C). Comparison of histograms reveals the peaks were biased towards the beginning of the delay period (p<0.02; Wilcoxon rank sum test). (F) Normalized firing rates of all neurons triggered to different events of the task. Rows in each plot are sorted by time of peak firing rate. Notice that aligning to sample press time is the same as aligning to delay period onset for all delay period lengths.
Figure 4—figure supplement 2. Ridge-to-background analysis to quantify the presence of firing rate sequences.
(A) Ridge-to-background ratios of delay period activity during 1 s delay trials (left), 5 s delay trials (middle), and 10 s delay trials (right). Green bars represent ridge-to-background ratios from the real data and correspond to the three heat-maps in Figure 4B. Blue bars represent the mean ridge-to-background ratio obtained by circularly shifting the spike times by random values 1000 times. For all three delay periods, ridge-to-background ratio was significantly larger than that expected by chance (*p<0.001; one tailed test using the ratios from the randomly shifted data as the null distribution). (B) Ridge-to-background ratios of delay period activity for the time that the rat is at the nose port for 10 s delay trials only, aligned to delay onset (left) and aligned to time when the rat first arrives at nose port (right). Green bars represent ridge-to-background ratios and correspond to the left-most heat-maps in Figure 4—figure supplement 2A. Blue bars represent the mean ridge-to-background ratio obtained by circularly shifting the spike times by random values 1000 times. For both cases, ridge-to-background ratio was significantly larger than that expected by chance (* p<0.001; one tailed test using the ratios from the randomly shifted data as the null distribution). (C) Ridge-to-background ratios of activity aligned to sample lever press, nose-poke, choice lever press, and reward port entry. Green bars represent the ridge-to-background ratios and correspond to the four heat-maps in Figure 4—figure supplement 2F. Blue bars represent the mean ridge-to-background ratio obtained by circularly shifting the spike times by random values 1000 times. For all four cases ridge-to-background ratio was significantly larger than that expected by chance (*p<0.001; one tailed test using the ratios from the randomly shifted data as the null distribution). Ridge-to-background analysis was calculated on a time window spanning 5 s before event to 15 s after event. In all panels, error bars are ± 1 standard deviation.