Fig. 5: Change in quantity and activity of different dyad types as the experiment progressed.

The mean number of events per dyad during each period of the experiment is displayed for same- (orange line) and different-class (blue line) a non-affiliate and b affiliate dyads. The number of same-class dyads (dark grey shading) is also shown alongside the combined number of same- and different-class dyads (light grey shading). Different-class affiliate dyads and both types of non-affiliate dyad were first observed shortly after the start of the experiment, but same-class affiliate dyads were only observed after approximately 200 association events had taken place. For affiliate dyads, the balance of events per dyad across same-class and different-class dyads shifted as the experiment progressed, indicating an increase in activity within same-class affiliate dyads during the last third of the experimental period. However, due to a greater number of different-class (11) as compared to same-class (7) affiliate dyads, this increase in within-dyad activity did not translate into an overall difference in likelihood of observation (308 observations of same-class affiliate dyads, 340 observations of different-class affiliate dyads). In contrast, same-class and different-class non-affiliate dyads displayed close to equal activity through the first approximately 10% of the experiment (300 events: 60 same-class non-affiliate dyads, 69 different-class non-affiliate dyads, 140 same-class observations, 144 different-class observations), but same-class activity exceeded different-class activity across the remainder of the experiment (2817 events: 333 same-class non-affiliate dyads, 339 different-class non-affiliate dyads, 1279 same-class observations, 906 different-class observations).