Function of a multimodal signal: A multiple hypothesis test using a robot frog

Abstract

Multimodal communication may evolve because different signals may convey information about the signaller (content‐based selection), increase efficacy of signal processing or transmission through the environment (efficacy‐based selection), or modify the production of a signal or the receiver's response to it (inter‐signal interaction selection). To understand the function of a multimodal signal (aggressive calls + toe flags) emitted by males of the frog Crossodactylus schmidti during territorial contests, we tested two hypotheses related to content‐based selection (quality and redundant signal), one related to efficacy‐based selection (efficacy backup), and one related to inter‐signal interaction selection (context). For each hypothesis we derived unique predictions based on the biology of the study species. In a natural setting, we exposed resident males to a robot frog simulating aggressive calls (acoustic stimulus) and toe flags (visual stimulus), combined and in isolation, and measured quality‐related traits from males and local levels of background noise and light intensity. Our results provide support to the context hypothesis, as toe flags (the context signal) are insufficient to elicit a receiver's response on their own. However, when toe flags are emitted together with aggressive calls, they evoke in the receiver qualitatively and quantitatively different responses from that evoked by aggressive calls alone. In contrast, we found no evidence that toe flags and aggressive calls provide complementary or redundant information about male quality, which are key predictions of the quality and redundant signal hypotheses respectively. Finally, the multimodal signal did not increase the receiver's response across natural gradients of light and background noise, a key prediction of the efficacy backup hypothesis. Toe flags accompanying aggressive calls seem to provide contextual information that modify the receiver's response in territorial contests. We suggest this contextual information is increased motivation to escalate the contest, and discuss the benefits to the signallers and receivers of adding a contextual signal to the aggressive display. Examples of context‐dependent multimodal signals are rare in the literature, probably because most studies focus on single hypotheses assuming content‐ or efficacy‐based selection. Our study highlights the importance of considering multiple selective pressures when testing multimodal signal function.

How signals interact in multimodal displays: Insights from a robotic frog

“…Alternative hypotheses and key predictions for the function of multimodal (acoustic + visual) agonistic displays in C. schmidti , tested in Caldart et al (2021). Terminology follows Hebets and Papaj (2005).…”

“…To understand the current function of multimodal agonistic displays in C. schmidti , Caldart et al (2021) built on a detailed understanding of the natural history of this species (Caldart et al, 2011; Caldart et al, 2014). Males of C. schmidti are diurnal and territorial.…”

“…Caldart et al (2021) used an electromagnetic robot, programmed to emit acoustic and visual signals identical to those produced by living males, and focused on the responses of receivers, their quality proxies and local environmental settings to evaluate the support for content‐based selection, efficacy‐based selection and inter‐signal interaction (Figure 1). Robotic models have been successfully employed in studies of animal communication and multimodal signalling over the last two decades (Narins et al, 2003; Taylor et al, 2008, 2011).…”

“…The responses to multimodal displays in C. schmidti are shaped by the interaction between acoustic and visual components. First, Caldart et al (2021) showed that the probability and latency of response is similar whether resident males are exposed to acoustic or multimodal signals, suggesting that visual signals in a multimodal display do not act as a backup for signal transmission. Moreover, using two quality traits in anurans, body size and body condition, Caldart et al (2021) found no correlation between the quality trait and the rate of emission of signals of either modality in resident males.…”

“…Finally, an important step beyond assessing the current function of a multimodal display is to understand the sources of between-and within-individual variation in the use of alternative signal components. The results of Caldart et al (2021) Here, we expect that the probability and latency of receiver responses vary along environmental gradients only when exposed to unimodal displays. If the stimulus is multimodal, these response metrics should be consistently highly across environments.…”