Appendix 2—figure 1. M2-M3 co-operation during social learning.

Here we describe interesting social interactions between M2 & M3 during social training. To summarize, on Days 4 and 5, M2 was positioning himself in front of the touch screen, occupying the juice spout as usual, since M2 was dominant over M3. However, for some stretches, he allowed M3 to sit alongside closely such that M3 also had access of the touch screen. During these stretches, M3 performed the task for few trials (grey box), which included both correct and incorrect trials. Since M2 was occupying the juice spout, he got rewarded for these correct trials performed by M3. These interactions are detailed below. (A) Day 4, M2-M3 session: Shaded regions are showing trials where M2 and M3 co-operated in the task (M3 performed the task and M2 got juice). Red dots in shaded region are showing correct trials. The whole session is divided into non-overlapping bins (bin size is 15 trials except in the shaded regions). Each dot represents accuracy calculated on the total trials in that bin. Touching accuracy: percentage of trials initiated by M2. Response accuracy: percentage of responded trial (correct or incorrect) out of total trials. On this day, M2 was not touching the hold button much before the interaction trials (before trial 106), but after that M2 started initiating trials (Figure 4—figure supplement 1A ). He did not make any more progress. (B) Day 5, M2-M3 session: Correct response accuracy: percentage of total trials in which M2 made a correct response. Here bin size is 20 trials. All other conventions are same as (A). The arrow indicates the trial from which the hold time was changed (Day 1: 500 ms). From the beginning M2 was initiating the trials by touching the hold button but his response accuracy was very low (i.e. did not reach the two choices stage). He was able to maintain hold till response button appeared and made a response by dragging his hand through “Different button” for 13 trials before the interaction, out of which only four trials were correct. After this, M2 allowed M3 to perform the task for 14 trials (till trial 381) in which M2 received juice at a much higher rate (8 trials out of 14 were correct). After this interaction, M2’s response accuracy increased (Figure 4—figure supplement 1B) and he started making correct response at chance level, although this was largely due to only making the (DIFFERENT response).