Emergent collective behavior evolves more rapidly than individual behavior among acorn ant species

Abstract

Emergence is a fundamental concept in biology and other disciplines, but whether emergent phenotypes evolve similarly to nonemergent phenotypes is unclear. The hypothesized process of emergent evolution posits that evolutionary change in at least some collective behaviors will differ from evolutionary change in the corresponding intrinsic behaviors of isolated individuals. As a result, collective behavior might evolve more rapidly and diversify more between populations compared to individual behavior. To test whether collective behavior evolves emergently, we conducted a large comparative study using 22 ant species and gathered over 1,500 behavioral rhythm time series from hundreds of colonies and isolated individuals, totaling over 1.5 y of behavioral data. We show that analogous traits measured at individual and collective levels exhibit distinct evolutionary patterns. The estimated rates of phenotypic evolution for the rhythmicity of activity in ant colonies were faster than the evolutionary rates of the same behavior measured in isolated individual ants, and total variation across species in collective behavior was higher than variation in individual behavior. We hypothesize that more rapid evolution and higher variation is a widespread feature of emergent phenotypes relative to lower-level phenotypes across complex biological systems.

Keywords: behavioral evolution; complex systems; synchronization; temnothorax; ultradian rhythms.

Fig. 1.

(A) Representative time series of collective activity from colonies of six different ant species. (B) Representative activity time series from isolated individuals of six different ant species. All of the visualized time series depict the raw data after being rescaled to fall between 0 and 1. (C) The phylogeny of species used in this study, with boxplots showing the data for each of our behavioral traits next to the corresponding species. Colony-level data are potted in red, and data from isolated individuals are plotted in blue. Each colony-level (red) data point represents the mean trait value for a unique colony. Each individual-level (blue) data point represents the data from a unique individual. (D) Estimates for the rates of evolution (σ²) for the matching collective and individual ultradian traits. The distribution of rate estimates for each trait was obtained using a bootstrapping approach (Methods). Dots represent rate medians and error bars depict the 95% CI.

Fig. 2.

(A) Behavioral phenospace of ant ultradian rhythms across 22 different species. Each data point represents a different species. Solid points falling within the pink polygon correspond to species’ colony-level rhythms, and open circles within the blue polygon correspond to species’ individual-level rhythms. The individual-level and colony-level data points associated with Temnothorax rudis and Temnothorax obturator are provided as examples. Specimen images are to scale and were extracted from images from www.antweb.org (T. rudis: casent0005689, T. obturator: casent0104756) Each corner of the phenospace is ornamented by a synthetic time series generated by a noise-driven FitzHugh–Nagumo model (SI Appendix, SI text) that qualitatively illustrates the features of time series from their respective region of the phenospace. (B) comparison of the phenotypic disparity (measured as the sum of variances) between species’ colony-level and individual-level behavior spaces. The sum of variances was calculated by bootstrapping the phenospace data. Dots represent medians, and error bars depict the 95% CI. (C) Behavioral phylomorphospaces for individual-level vs. colony-level period and (D) individual-level vs. colony-level rhythmicity. The dotted lines show what a 1:1 relationship between the variables would be.