Fig 7. Together, total history dependence and the information timescale show clear differences between neural systems.

(A) Embedding-optimized Shuffling estimates (d_max = 5) of the total history dependence R_tot are plotted against the information timescale τ_R for individual sorted units (dots) from four different neural systems (raster plots show spike trains of different sorted units). No clear relationship between the two quantities is visible. The analysis shows systematic differences between the systems: sorted units in rat cortical culture (n = 48) and rat dorsal hippocampus layer CA1 (n = 28) have higher median total history dependence than units in salamander retina (n = 111) and mouse primary visual cortex (n = 142). At the same time, sorted units in cortical culture and retina show smaller timescale than units in primary visual cortex, and much smaller timescale than units in hippocampus layer CA1. Overall, neural systems are clearly distinguishable when jointly considering the total history dependence and information timescale. (B) Total history dependence R_tot versus the autocorrelation time τ_C shows no clear relation between the two quantities, similar to the information timescale τ_R. Also, the autocorrelation time gives the same relation in timescale between retina, primary visual cortex and CA1, whereas the cortical culture has a higher timescale (different order of medians on the x-axis). In general, neural systems are harder to differentiate in terms of the autocorrelation time τ_C compared to τ_R. Errorbars indicate median over sorted units and 95% bootstrap confidence intervals on the median.