Back

Group-level phenotypes are idiosyncratic yet unpredictable over development in a clonal fish

Perkes, A.; Jolles, J.; Doran, C.; Wolf, M.; Bierbach, D.; Laskowski, K.

2026-02-05 animal behavior and cognition
10.64898/2026.02.03.700030 bioRxiv
Show abstract

Behavioral differences among groups can have profound impacts on individual fitness1, but the ontogeny of group-level differences is poorly characterized. Group-level differences may originate from genetic differences and/or subtle variation among individuals2-5, but they are also shaped over time by social feedback mechanisms within and between groups, leading to potentially dramatically different group-level behavior6-8. Disentangling these drivers of group phenotypes is inherently difficult, particularly because we usually observe groups after they have already formed. Here we take advantage of a powerful study system, the naturally clonal Amazon molly, to form replicate groups of genetically identical individuals and continuously track group-level behavior from birth throughout the first 55 days of life. We find that groups, on average, vary in a predictable fashion during this early developmental phase. In stark contrast, while we also find that groups differ in their collective phenotypes from very early in life, these among-group differences are highly unstable and thus inherently unpredictable over longer timescales. It appears that groups undergo a reshuffling of behavior in mid-development, around 4-5 weeks of age. These results suggest that different processes generate group-level differences throughout ontogeny: immediately after birth, groups differ predictably because individuals display consistent individual differences, but as individuals mature and become more socially responsive, social feedbacks push groups along divergent, less predictable behavioral trajectories.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.