Spike avalanches of 40-80 units in six awake mouse brain divisions did not meet the crackling-noise scaling relation; the gap fell with synchrony and did not need spike timing below 50 ms

Question: Fontenele et al. (2019) reported that, in anesthetized rat cortex, spike avalanches satisfy the crackling-noise scaling relation gamma = (tau_t - 1)/(tau - 1) only at an intermediate population CV, which they gave as evidence for criticality between cortical states. Does this hold in awake mice and in other brain divisions? Approach: We used two public Neuropixels data sets (Steinmetz et al. 2019, 39 sessions, 10 mice; IBL brain-wide map spontaneous epochs, 84 sessions). We made populations of 40 or 80 good units per area or division (isocortex, hippocampal formation, thalamus, midbrain, striatum, other), sorted 10-s windows by CV of 50-ms counts, and calculated Delta = predicted minus fitted gamma per CV group. A surrogate moved each spike inside its own 50-ms bin, which keeps the CV of every window. A driven branching network gave a positive and a negative control. Results: Delta stayed positive. In 204 Steinmetz populations of 40 units it fell from 0.43 (95% CI 0.40-0.46) in the lowest to 0.25 (0.23-0.29) in the highest CV decile, and changed sign in only 7.8% of populations. The smallest deviation was usually in the most synchronized windows, which were the quiet windows (less face and wheel movement). 80-unit populations and IBL data gave the same pattern. Jittered data gave nearly the same Delta profiles (median profile correlation 0.92-0.97; mean abs Delta difference 0.006, -0.021 and 0.000). A network with no propagation and slow common drive reproduced the data pattern. Limitation: a near-critical network (m = 0.995) met the relation only at weak drive (1.3 Hz per unit); at recorded rates (3.7-4.5 Hz) near-critical networks also gave Delta 0.14-0.22 and no jitter effect. Thus, with this sampling, the scaling-relation test cannot separate near-critical propagation from slow common drive in awake mice; the null result is not evidence against criticality. Fit ranges are short (about 1.5 decades).

neuronal avalanches · criticality · crackling noise · scaling relation · cortical state · Neuropixels · subsampling · spike jitter surrogate · branching network · awake mouse · replication · brain dynamics