This seems counter to the original findings with mistakenly large smoothing windows. It seems for Both-Rewarded events and Win events, most of the neurons are mainly significant only during the Alert phase, and not during the Dispense phase. However, for Lose events, most of those were significant during the Dispense phase.
Those Win & Lose plots are actually coming from different mice, because no mice during the Both-Rewarded trials both won and lost >5 times during a recording, which is required for a Wilcoxon test. The Both-Rewarded plot comes from all of the mice.
So mice that typically won, had win-sensitive neurons during the alert phase, and mice that typically lost, had lose-sensitive neurons during the dispense phase. Is this maybe because all mice have a reward expectation signal during the alert phase, except for the mice that typically lost (because they knew they weren’t going to get a reward), and the dispense phase of the lose neurons are actually encoding the cagemate getting rewarded.
So does this mean we should only be looking at the first 5 seconds instead of the whole 10 second event? Well let’s compare it the same results to whole 10-second significance, and see if significance in either phase will result in a 10 second significance.
Although Alert & Dispense might not overlap too much, they both overlap with the Whole 10 significance, so I think it’s still safe to continue mainly looking at the Whole 10. If we wanted to ask specifically about reward expectation or post-reward dispense, we could, but if we’re considering an “event” like “winning” or “losing”, I think it’s fair to look at the Whole 10 seconds.
I think it’s also pertinent to note the quantity of neurons that showed significance for the Whole 10, but not for either Alert or Dispense. I don’t know what exactly to make of that.