To mitigate potential risks and develop trust, strangers may “test the waters” by gradually escalating the type of social investment from low cost to high cost. We introduced four unfamiliar groups of feral monk parakeets together and observed the sequence of social behaviors that occurred as relationships developed. We then tested the effect of relationship status (stranger vs familiar) on the probability of dyads following predicted sequences. We also tested whether strangers who progressed their relationships maintained higher rates of no-contact proximity compared to dyads that did not. We found that stranger dyads, but not familiar dyads, were more likely to (1) approach each other without contact before making contact and (2) follow predicted sequences of affiliative behaviors. Strangers that progressed to contact also had higher rates of associations than did birds that never made contact.

We provide raw data and an R markdown file with code to reproduce all analyses and figures. 

Social behaviors and predicted sequences

We categorized the first observed instances of behaviors based on perceptions of relative risk and investment into 3 levels: Low (no-contact proximity), Moderate (shoulder contact, allopreening, beak touching), or High (allofeeding, copulating), and visualized these behaviors using a Sankey plot.

Estimating the effect of relationship status on the sequence of behaviors

To assess whether the predicted sequences occurred more often in stranger dyads than in familiar dyads, we used the R package brms (version2.220) to fit two generalized multi-membership models (GLMM) with a Bernoulli outcome (logit link) for whether or not dyads followed the expected sequences.

To simulate how often predicted sequences would occur if stranger and familiar dyads had exhibited behaviors in random order, we shuffled (1,000 iterations) the timebin in which behaviors were observed among dyads.

Testing for relationship progression among strangers

Using the same permutation-based reference model described above, we also used nonparametric permutation tests to assess whether the sequences occurred more than expected if stranger dyads had exhibited behaviors in random order.

Finally, we assessed whether each bird’s mean hourly proximity rate with stranger dyads observed in affiliative contact was greater than the mean hourly proximity rate with stranger dyads not observed in affiliative contact, using a permuted paired t-test (5,000 iterations).