Comment on 'Parasite defensive limb movements enhance acoustic signal attraction in male little torrent frogs'

Abstract

Zhao et al. recently reported results which, they claim, suggest that sexual selection produces the multimodal displays seen in little torrent frogs (Amolops torrentis) by co-opting limb movements that originally evolved to support parasite defense (Zhao et al., 2022). Here, we explain why we believe this conclusion to be premature.

Keywords: Amolops torrentis; co-option; ecology; evolutionary biology; multimodal display; parasite defense; sexual selection.

Figure 1.. Re-analysis of whether male little torrent frogs (Amolops torrentis) produce limb displays in the presence of parasites.

(A) Proportion of different limb displays observed passively in a population of males (n=69) either in the presence of parasites (blue bars) or in the absence of parasites (orange bars). Note that these data are weighted by the number of limb movements each male produced, which were highly skewed in the original dataset. In other words, in the first analysis by Zhao et al., some males produced >90 displays, whereas other males produced zero (Zhao et al., 2022). See Methods for details about how we weighted values. For all subsequent analyses (G-tests for goodness of fit), the proportion of toe trembling produced in the presence and absence of parasites was used as the null hypothesis, setting our expectation of how often displays should be produced by chance in the presence or absence of parasites (see Methods for justification). (B–F) Density plots of the boot strapped chi-squared (χ²) statistics from the G-test of goodness of fit analysis. On the y-axis is the density of chi-squared (χ²) statistics after 1,000 iterations, and on the x-axis is the chi-squared (χ²) value. Solid green lines denote mean chi-squared statistics associated with each distribution of values, whereas solid red lines represent the cut-off for statistical significance (P<0.05) with 1 degree of freedom. If the green line falls on the right side of the red line, then the result is statistically significant (i.e., male frogs appear to perform the given display in the presence of parasites more than we might expect by chance, as determined by the null model set through toe trembling). By contrast, if the green line falls on the left side of the red line, then the result is not significant (i.e., male frogs do not perform the given display in the presence of parasites more than we might expect by chance). We found that (B) toe trembling (TT) was (as expected) not statistically significant (χ²=0.084, P=0.772), nor was (C) hind foot lifting (HFL; χ²=0.487, P=0.485) or (D) arm wiping (AW; χ²=2.772, P=0.096). Importantly, these were the behaviors the females supposedly preferred, though see the main text for a discussion of the limitations associated with this assay. We found that (E) limb shaking behavior (LSA) was statistically significant (χ²=5.0314, P=0.025, denoted with asterisk), as was (F) wiping (W) (χ²=4.212, P=0.040, denoted with asterisk). These latter two behaviors (LSA and W) were not preferred by females in the behavioral assay. Note that when comparing A to both E and F (LSA and W, respectively), the proportions in A would suggest that the effect reported in F would be more robust, compared to the effect in E. However, there were several males that did not wipe (0 values), which may have broadened the Chi Squared curve and decreased the statistical power in the analysis.