Can rat cortical cultures predict what comes next, unpredictable or learned?
In the authors' words
In humans, physiological activity recorded before an unpredictable stimulus reportedly varies slightly with what the stimulus will be, an effect termed predictive anticipatory activity. Whether reduced neural preparations show it is unknown. We re-analyzed a public dataset in which twelve dissociated rat cortical cultures on high-density multi-electrode arrays performed a hidden-source inference task (DANDI 001611: 27 experiments, 691,200 trials). Two features of the dataset limit what it can test. Each experiment replays one fixed 256-trial source sequence in all 100 sessions, so the upcoming source is unpredictable only in session 1. One stimulation volley per trial arrives on an exact 1-s clock, so every pre-stimulus window lies in the previous trial's after-effects. Two questions were fixed before a confirmatory run on held-out experiments. Q1 asked whether, in session 1, a stimulation-free 250-ms window before each trial's volley carries information about the upcoming sources beyond what trial history explains. Q2 asked whether this information grows over the 100 repetitions, as learned expectation would predict. Q1 (13 experiments, 9 cultures): decoding above a shifted-label null, {Delta}AUC +0.003, 95% CI -0.026 to +0.033, randomization p = 0.73, equivalent to zero within +/-0.059 (the human effect expressed as AUC) and narrowly within +/-0.03. Q2 (12 experiments, 8 cultures): growth {Delta}G +0.004, 95% CI -0.006 to +0.013, p = 0.29, equivalent within +/-0.02. Pre-stimulus activity thus carried no detectable information about the stimulus, and none emerged with repetition.
Appeared: Thursday, November 19. bioRxiv. Preprint, not yet peer-reviewed.