Responses of intended and unintended receivers to a novel sexual signal suggest clandestine communication

Abstract

Inadvertent cues can be refined into signals through coevolution between signalers and receivers, yet the earliest steps in this process remain elusive. In Hawaiian populations of the Pacific field cricket, a new morph producing a novel and incredibly variable song (purring) has spread across islands. Here we characterize the current sexual and natural selection landscape acting on the novel signal by (1) determining fitness advantages of purring through attraction to mates and protection from a prominent deadly natural enemy, and (2) testing alternative hypotheses about the strength and form of selection acting on the novel signal. In field studies, female crickets respond positively to purrs, but eavesdropping parasitoid flies do not, suggesting purring may allow private communication among crickets. Contrary to the sensory bias and preference for novelty hypotheses, preference functions (selective pressure) are nearly flat, driven by extreme inter-individual variation in function shape. Our study offers a rare empirical test of the roles of natural and sexual selection in the earliest stages of signal evolution.

Fig. 1. Map of the Hawaiian Islands indicating presence of three major cricket morphs (ancestral, purring, and silent) across six sampling populations.

Wing drawings adapted from Fig. 1 in Tinghitella et al. .

Fig. 2. PCA loadings of purring song characteristics (a–k) including frequency, temporal patterning, and broadbandedness (Table 2).

Points show the 46 songs used in the PCA. Songs chosen for phonotaxis in the exemplar experiment are indicated in red. Source data are provided as a Source Data file.

Fig. 3. Positive phonotactic behavior in female crickets and flies depends upon song type.

a Playbacks of the ancestral calling song elicited greater rates of phonotactic responses by female crickets than playbacks of the derived purring song or white noise control, while female crickets also responded more positively to purring song than to white noise. Data shown here come from phonotaxis responses of crickets in both the frequency and exemplar experiments. Dotted lines indicate the proportion of crickets exhibiting positive phonotaxis averaged across populations. b Fly traps playing ancestral song caught the overwhelming majority of flies in the field, with only a single fly caught at a trap playing the purring song, and none caught at traps broadcasting white noise (open points = average number of flies/trap; whiskers = SE; gray points = raw data). Colors of bars indicate populations of origin, with sample sizes indicated below the bars (N = numbers of individual phonotaxis trials and traps deployed for crickets and flies, respectively). Source data are provided as a Source Data file.

Fig. 4. Preference functions generated from behavioral responses of female T. oceanicus and O. ochracea to frequency-manipulated purring songs.

Panels a–c show whether T. oceanicus females (N = 114) displayed phonotactic behavior (a), whether or not females contacted the speaker (b), and distance traveled (c). Panels d and e show whether or not female O. ochracea (N = 37) moved during stimulus playback (d) and whether they contacted the speaker (e). Thin lines show individual-level splines fit by the program Pfunc to female responses (as identical splines are possible with binomial responses, some replicates overlap in panels a, b, d, e). Dashed lines show splines summarizing responses at the population-level (crickets; a–c) or species-level (flies; d and e). Source data are provided as a Source Data file.

Fig. 5. Fitness landscapes describing initial selection exerted by female crickets (N = 271) across the acoustic space of purring songs (exemplars) soon after the origin of purring.

Panel a shows whether the female cricket displayed phonotactic behavior, while b and c show whether or not females contacted the speaker and the distance they traveled, respectively. Surfaces depict female responses aggregated across populations (see Supplementary Table 3 for estimates of population random effects from GLMMs). Source data are provided as a Source Data file.