Cooperating elephants mitigate competition until the stakes get too high

Abstract

Cooperation is ubiquitous in the animal kingdom as it aims to maximize benefits through joint action. Selection, however, may also favor competitive behaviors that could violate cooperation. How animals mitigate competition is hotly debated, with particular interest in primates and little attention paid thus far to nonprimates. Using a loose-string pulling apparatus, we explored cooperative and competitive behavior, as well as mitigation of the latter, in semi-wild Asian elephants (Elephas maximus). Our results showed that elephants first maintained a very high cooperation rate (average = 80.8% across 45 sessions). Elephants applied "block," "fight back," "leave," "move side," and "submission" as mitigation strategies and adjusted these strategies according to their affiliation and rank difference with competition initiators. They usually applied a "fight back" mitigation strategy as a sanction when competition initiators were low ranking or when they had a close affiliation, but were submissive if the initiators were high ranking or when they were not closely affiliated. However, when the food reward was limited, the costly competitive behaviors ("monopoly" and "fight") increased significantly, leading to a rapid breakdown in cooperation. The instability of elephant cooperation as a result of benefit reduction mirrors that of human society, suggesting that similar fundamental principles may underlie the evolution of cooperation across species.

Fig 1. Cooperation rate of Asian elephants in the test under the 2-tray and the 1-tray phase.

Cooperation rate was calculated as frequency of successful cooperation divided by frequency of all attempts (i.e., all trials in a session). Each point represents the cooperation rate in each session. Gray areas represent 95% confidence intervals. In the 2-tray Phase I, cooperation was maintained at a high level throughout testing, while cooperation broke down entirely after the 17th session in the 1-tray Phase II. The data used to generate this figure can be found in S1 Data.

Fig 2

Change in competition behaviors over time in the 2-tray Phase I (A) and 1-tray Phase II (B). Each point represents the frequency of a competition behavior in each session. Different colors represent different types of competition behaviors, while gray shading represents 95% confidence intervals. The perceived cost of each type of competition behavior increases from left to right, with approach being least costly and fight being most costly. While the cooperation rate remained stable across the 2-tray Phase I, it decreased rapidly in 1-tray Phase II, which contributed to the decreasing frequency of each competition behavior across Phase II (i.e., when cooperation happened less frequently or not at all, there was less or no food, respectively, for which to compete). The data used to generate this figure can be found in S1 Data.

Fig 3

Mitigation strategies used by elephants in Phase I based on their rank difference with a partner (top) or their level of affiliation (bottom). Rank difference was calculated using the ranking of the initiator elephant minus the ranking of the target elephant. A positive rank difference means the initiator was dominant to the target elephant, while a negative value indicates the initiator was of lower rank. The only exception to this is that the rank difference under the “rope pulling” competition type was calculated as the absolute value of the rank difference between initiator elephant and target elephant, because we could not tell which individual initiated the behavior when 2 elephants pulled a single rope. The maximum affiliation index was 126. A solid line means that the probability of a mitigation strategy being used by an elephant can be predicted by rank difference (top plots) or affiliation (bottom plots), while long dashed lines represent nonsignificant patterns. The solid red line (no response) was set as a reference level in the models. The data used to generate this figure can be found in S1 Data.

Fig 4. Comparison of competition behavior frequencies between 2-tray and 1-tray phases.

The bold line inside each box is the median frequency per session of each competition behavior. The top whisker represents the maximum frequency, while the bottom whisker represents the minimum frequency (excluding outliers). Upper and lower quartiles are represented by the top and bottom edge of the box, respectively. The black data points above the top or below the bottom whisker are outliers. Asterisks represent significant differences between the frequency of behaviors between the 2-tray and 1-tray conditions (*** P < 0.001; n.s. P > 0.05). The data used to generate this figure can be found in S1 Data.

Fig 5. Mitigation strategies used by elephants in Phase II based on their rank difference with a partner (top) or their level of affiliation (bottom).

Rank difference was calculated using the ranking of the initiator elephant minus the ranking of the target elephant. A positive rank difference means the initiator was dominant to the target elephant, while a negative value means the initiator was of lower rank. The only exception to this is that the rank difference under the “rope-pulling” competition type was calculated as the absolute value of the rank difference between the initiator elephant and the target elephant because we could not differentiate between the initiator and target when the 2 elephants were pulling a single rope. The maximum affiliation index was 126. A solid line means that the probability of a mitigation strategy being used by an elephant can be predicted by rank difference (top plots) or affiliation (bottom plots), while long dashed lines represent nonsignificant patterns. The solid red line (no response) was set as a reference level in the models. The data used to generate this figure can be found in S1 Data.

Fig 6. The loose-string cooperation apparatus in this study.