Testing socially mediated estrous synchrony or asynchrony in wild baboons

Abstract

Social structure of animal groups is affected by the spatial and temporal distribution of females. In particular, the extent to which fertile periods of females are temporally overlapped has been deemed as a crucial factor to determine the structure of social groups in primate species. Dominant males are less able to monopolize fertile females when there are more females being in estrous simultaneously. This provides a potential opportunity for females to manipulate their defensibility by dominant males by modifying the level of estrous overlap. Previous studies that have attempted to detect socially mediated synchrony or asynchrony of estrous cycles have produced mixed results, some of which have been questioned on methodological grounds. Here, we address this issue using an exceptionally large dataset of daily reproductive states in four troops of wild anubis baboons (Papio anubis) over 14 to 24 years (77 troop-years). We compare observed levels of estrous synchrony with null distributions obtained by a randomization procedure under the assumption that estrous cycles are mutually independent among females in the same troop. We do not find any evidence supporting synchrony or asynchrony of estrous cycles. Based on our result and those of previous studies in other species of baboons, we conclude that socially mediated synchrony or asynchrony is unlikely to play a significant role in structuring social groups in baboons. In addition, our analysis points out that care should be taken when applying randomization procedures to a dataset with missing observations.

Significance statement

Females in some animal species have been suggested to synchronize or desynchronize their estrus cycles. This phenomenon may have a significant impact on the structure of animal societies because when there are more females simultaneously in estrus, even a dominant male is less able to defend them from other males. Previous attempts to detect non-random estrous cycles in primates have produced mixed results, and some of them have been criticized on the methodological grounds. We investigate whether wild female anubis baboons exhibit non-random estrous cycles using a larger dataset than previous studies. For a statistical analysis, we use a randomization procedure, taking a biasing effect of missing observations into consideration. Consistent with earlier studies on other species of baboons, our analysis does not find any evidence supporting estrous synchrony or asynchrony in anubis baboons.

Appendix

Additional simulations to study the effect of missing observations

The following analysis was done in order to check whether missing observations can affect our test of non-independent estrous cycles. We generated artificial annual sequences of reproductive states for 10 females by randomly assigning each female a reproductive state for each day. For simplicity, reproductive state in a given day was chosen independently of those in preceding days. Possible reproductive states were either E or NE, and state E was chosen with probability 6/37, which approximates the observed proportion of days in which a cycling female is in state E. We then randomly added missing observations to the artificial data by replacing some assigned reproductive states by state U. Days for which the assigned reproductive states were modified were chosen in two distinct ways: in one set of simulations, each female had missing observations in different days, while in the other, all females had missing observations in the same missing days.

In artificial samples generated in this way, reproductive states are by definition independent of each other among females. Hence, by applying our test of non-independent estrous cycles on the artificial samples, we can investigate whether missing observations affect the outcome of the test in the sense that the p values are systematically biased toward large or small values even under the null hypothesis. Specifically, we obtained p values from 1000 artificial samples for each of various numbers of missing observations to see if the average p value varies with the number of missing observations.

In the absence of missing observations, the mean p value of 1000 simulations was approximately 0.5. This result is as expected and assures that the test is unbiased in the absence of missing observations. When the days lacking observations were chosen independently for each female, the mean p values were approximately 0.5 for all numbers of missing observations per female examined (Fig. 4a). In stark contrast, when the days lacking observations were the same for all females, that is, missing observations always occurred as missing days, the mean p value increased with increasing number of missing days (Fig. 4b). These results indicate that outcomes of our test of non-independent estrous cycles are affected by the presence of missing days, but not by missing observations per se, in a way that p values are systematically biased toward large values.

Fig. 4

Means and standard deviations of p values of tests of non-independent estrous cycles applied to artificial samples with various numbers of missing observations. The mean and standard deviation were obtained from 1000 simulation runs for each case. a Days lacking observations were chosen independently for each female. b Days lacking observations were the same for all females, that is, missing observations always occurred as missing days. c Missing observations occurred as missing days were removed before the randomization procedure

Our simulation work also suggests that missing days do not cause a bias in the test if they are removed before the randomization procedure. The mean p values were approximately 0.5 for all numbers of missing days examined when the missing days were deleted and the resultant gaps were closed up (Fig. 4c).