The evolution of novel animal signals is critical to the generation of biodiversity. Here, we explore how new sexual signals become established. This process is challenging to explain because if receiver preferences are coupled with existing signals, then most receivers should discriminate against new signals. We investigated an underappreciated hypothesis: relaxed receiver preferences facilitate novel signal evolution by allowing new signals to establish a foothold. Further, we probed the mechanistic underpinnings of relaxed preferences by combining field-based and common garden approaches, allowing us to investigate evolution and plasticity as mechanisms. We capitalized on the Pacific field cricket, Teleogryllus oceanicus, a species that has recently evolved multiple novel acoustic signals (e.g., purring and rattling) in response to an eavesdropping parasitoid fly only found in the crickets’ introduced range in Hawaii. To test the hypothesis that selection associated with high search costs in introduced populations leads to relaxed mating preferences and determine whether such relaxation is plastic, we conducted sound preference (phonotaxis) trials with females from the cricket’s native range (Australia and French Polynesia, where the fly is absent) and its introduced range (Hawaii, where the fly is present). We presented females with novel songs plus the typical, ancestral song. Differences in phonotactic behavior between the lab and field settings would indicate plasticity in preferences. Using Generalized Linear Mixed Models (GLMM) with whether the female cricket was phonotactic to a song (y/n) as a response variable, we found that Australian and French Polynesian females were quite plastic; they discriminated strongly against most songs in the field, but were much more phonotactic to rattling and the typical song in the lab. However, Hawaiian females exhibited little plasticity and were consistently highly responsive to the rattling and typical songs in the lab and field. This pattern points to a loss of ancestral plasticity in female preferences sometime after colonizing Hawaii, resulting in heightened responsiveness to all songs, allowing novel signals to establish. We then asked which specific preference function traits (tolerance, strength, and/or responsiveness) differed among regions to better understand what is ‘relaxed’ about preferences in Hawaii by generating individual and region-level preference functions for females from each region with respect to the purring, rattling, and ancestral stimuli. When examining whether females were phonotactic or not and how quickly they contacted stimuli, responsiveness, tolerance, and preference strength differed among regions. Finally, we developed a computational model to estimate the distances at which female T. oceancius from Australia and French Polynesia should be able to hear and use the purring, rattling, and typical calling songs. A model of mean peak frequencies and amplitudes among each song type revealed dramatic differences in effective hearing distances across song types. Our results provide insight into how novel signals gain a foothold when they initially invade, after which coupled preferences may eventually evolve, perhaps leading to reproductive isolation. Alternatively, relaxed preferences may remain longer term, facilitating the maintenance of signal diversity within populations.

Field-based Phonotaxis Tests

We collected reproductively mature adult female T. oceanicus at six field sites: two replicate sites in Queensland, Australia (Cairns and Daintree) in February of 2023, two in French Polynesia (Tahiti and Mo’orea) in December of 2023, and two in Hawaii (Hilo and Wailua) in December of 2022 (Figure S1, Table S1) using methods that are not biased with respect to song, sex, or life stage. Briefly, rather than locating individuals by sound, we swept by foot in the fields where the crickets are found, collecting animals visually such that all life stages, sexes, and morphs were encountered (following Tinghitella et al. 2018; Tinghitella et al. 2021). Upon collection, we took the animals to local field stations where we housed females in 15L plastic storage containers with rabbit food ad libitum, egg cartons, and cotton with water under natural day-night cycles (at a density of ~30-40 females per container). We acoustically isolated females from calling males for 24 hours before conducting phonotaxis trials.

Our first question was whether female crickets from Australian, French Polynesian, and Hawaiian populations differ in their phonotactic responses to the typical song and newly evolved songs in Hawaii. We addressed this in the field using standardized phonotaxis trials in which we tested each female’s response to the following stimuli: a loop of five purring calling songs (played in immediate succession), a loop of four rattling calling songs, a loop of four typical calling songs, and a silent negative control. All loops were previously used in Tinghitella et al., 2018, 2021, and Gallagher et al. 2022. We randomized the order in which the stimuli were played. We conducted all trials between 2-8 hours after sunset (the active period when females search for males) at 24-26°C and under red light (following Tinghitella et al, 2021). In each trial, we placed a single female cricket under a plastic cup at one end of an arena (50 cm wide × 195 cm long × 25 cm tall) 1m away from an AOMAIS Sport II Bluetooth speaker that was positioned in the arena’s center. After the female adjusted to the arena, we played the first stimulus track, lifting the cup and allowing her to walk about the arena. The typical T. oceanicus loop was broadcast at 70 dBA, the rattling loop at 60 dBA, and the purring loop at 53 dBA (at one meter away), reflecting biologically realistic amplitudes (Tinghitella et al. 2018; Gallagher et al 2022, 2024). We confirmed amplitudes using a class 1 PCE-430 sound level meter set to “A” weighted measurements with fast integration time. We played each stimulus for a maximum of one minute or until the female contacted the speaker. If the female demonstrated positive phonotactic behavior, but did not contact the speaker in the first minute, we gave her a second minute of playback to allow potential contact. We considered a female to be positively phonotactic when she moved in the direction of the speaker in a classic zigzag pattern without following the wall or circling the arena. In addition to noting whether the female was positively phonotactic, we recorded the maximum distance traveled and the latency to contact the speaker for each stimulus (following Tinghitella et al. 2021). One observer relayed behavioral observations to a recorder while wearing earplugs so they were blind to the stimulus. Given the low levels of phonotaxis exhibited by Australian females to our stimuli, we created a supernormal stimulus (Supplemental Methods) to confirm that females were indeed sexually mature and phonotactic (Figure S2).

Lab-based Phonotaxis Tests

To investigate the impacts of environment and experience on female preferences, we repeated our phonotaxis trials in the lab after rearing animals to at least the F2 generation (to avoid transgenerational effects). We reared animals in a common garden, but separated by population, in 15 L plastic containers in a clock-shifted, temperature and humidity-controlled room set to a 12:12 light:dark cycle and 25 °C. Each 15 L container housed a mixed sex group with access to ad libitum food (Fluker’s Cricket Chow for juveniles and Kaytee rabbit chow for adults), water from moistened cotton, and an egg carton. Prior to trials, we haphazardly chose females from each of the six lab reared populations at the antepenultimate instar, the instar prior to which they develop auditory organs (Kämper, 1992), and isolated them in a sound-insulated incubator inside of 1.89 L containers; thus, females were reared in silence (eliminating song experience as a source of variation in preference behavior). We ensured their virginity at the time of trials and checked isolated females twice a week for eclosion to adulthood. We conducted phonotaxis trials on adult females during their scotoperiod when they were 7-21 days post-eclosion. All equipment and the phonotaxis protocol were identical to the field trials except that we conducted lab-based trials in a temperature-controlled room set to 25 °C. All sample sizes can be found in Table S1.

Statistical Analyses

To address the hypothesis that relaxed preferences facilitate the success of novel sexual signals and gain insight into the underlying mechanisms, we first asked how phonotactic behavior depended on region (Australia, French Polynesia, Hawaii), song type (purring, rattling, typical, silence), and whether trials were conducted in the field or lab. Using the lme4 package (Bates 2015) in R Studio version 2024.12.1, we ran a Generalized Linear Mixed Model with a binomial distribution in which phonotaxis (yes/no) was the response variable and region, song type, and lab vs field plus all two-way interactions were included as main effects. Individual ID was a random effect to account for repeated testing of females with the four stimuli. In this model, a significant interaction between region and song type would indicate that there are differences in female preference among regions. As the lab vs. field effect indicates whether preferences are plastic, finding an interaction between region and lab vs. field would indicate that regions differ in the extent to which their preferences are plastic, as would occur if plasticity had evolved. This could occur in several ways. For instance, plasticity might be selected for in Hawaii if it facilitates mating where search costs are high, or ancestral plasticity might be lost if it is costly to maintain, leading Hawaiian females to be consistently highly responsive. And, a significant song type by lab vs field effect would support the hypothesis that plasticity is stimulus-dependent.

Next, we constructed preference functions using our field phonotaxis data to examine which preference traits (see below) differed among regions in order to gain insight into the manner in which preferences are relaxed in Hawaii. Preference functions describe signal attractiveness as a function of variation in signalling traits (Neelon et al. 2019) and can take on many shapes (e.g., open-ended or closed) described by quantifying preference traits (Rodríguez et al., 2006, 2013; Fowler-Finn & Rodríguez 2012a, 2012b; Kilmer et al. 2017; Neelon et al., 2019). We investigated three preference function traits: tolerance (acceptance of trait values that deviate from the peak preference), responsiveness (mean response level across all signal trait values), and strength (how much attractiveness declines as signal values deviate from the peak). An individual with a relaxed preference might, for instance, have high responsiveness, be more tolerant, and/or have low preference strength relative to an individual with a preference strongly coupled to a particular signal (Figure 1). We used PFunc (Kilmer et al. 2017) to visualize the shape of individual and region-level preference functions for the three song types (typical, rattling, and purring songs). Note that because novel songs are categorically different from the typical song, differing in multiple signal traits (e.g., dominant frequency, bandwidth), we visualized preference functions across song types. Preference function traits (tolerance, responsiveness, and preference strength) were extracted from individual-level functions fit to two measures of female responses: phonotaxis (yes/no) and latency to contact. General linear models compared these values across regions.

Having found that Hawaiian females are more responsive to novel songs than those from elsewhere across the crickets’ range, particularly in field studies, we next explored one mechanism, enhanced locomotor behavior, that may underlie that pattern. More locomotory females may be more likely to wander close enough to the quieter purring or rattling males to hear their songs and then pursue them as mates. We asked whether females from different regions differed in the distance they travelled during the silent negative control trials. These data were non-normally distributed, so we ran Kruskal-Wallis tests on distance traveled during silence by region in both the field and lab and conducted pairwise post-hoc comparisons with Mann-Whitney-Wilcoxon tests.

Additionally, given that rattling males are only present in Hilo (not Wailua), and purring males are only present in Wailua (not Hilo), we investigated potential local adaptation, using a GLMM comparing phonotaxis to these two songs (rattling and purring) between these two populations with female ID as a random effect. If local adaptation has occurred, we expect females to be more responsive to the novel song found in their respective populations. If coevolution with silent flatwing males is critical, Kauai females (where the population was more than 95% silent for at least 60 generations; Bailey et al. 2024) should be the most relaxed of all.

Model of Effective Hearing Distance

Finally, because the pattern of female responses to purring and rattling songs across the crickets’ range differed, with phonotaxis to purring being rare overall (see results), we developed a computational model to determine how far away female T. oceanicus are likely to be able to hear and use purring, rattling, and typical songs in the context of long-distance mate location. Using previously published neural audiogram data in T. oceanicus (Atkins & Pollack, 1986; Fullard et al., 2010) and the sound intensity levels and peak frequencies of different types of cricket calling songs, we built a model estimating effective hearing distances for each song type. We based the model on peak frequencies of the purring, rattling, and typical songs reported in (Gallagher et al. 2022) and sound intensity levels from (Wikle et al. 2025), modeling responses to the average dominant frequency +/- 1 standard deviation and the average amplitude +/- 1 standard deviation. The model accounts for non-frequency-dependent damping of sound with distance and frequency-dependent attenuation due to atmospheric absorption. Detailed modeling methods can be found in (Wikle et al, 2025), where we previously modeled the effective hearing distances to these songs for the parasitoid fly. There is substantial inter-individual variation in spectral characteristics (dominant frequency and bandwidth) and amplitude of novel songs (Gallagher et al. 2023). All of these likely influence effective hearing distances, and our model examines two important axes of variation (amplitude and dominant frequency) that characterize differences among song types, but we are unfortunately not able to account for variation in bandwidth in this model.

Several previously published neural audiograms based on ON1 (omega neuron 1- responsible for directional processing of cricket songs) recordings from T. oceanicus have been collected from lab-reared animals that originated from Australia (Atkins & Pollack, 1986; Kostarakos et al., 2009; Fullard et al., 2010) and Mo’orea (Fullard et al, 2010). Since purring and rattling songs contain frequencies that extend well above 10 kHz (Tinghitella et al, 2021; Gallagher et al. 2022), we digitized neural audiograms from (Atkins & Pollack, 1986) and (Fullard et al, 2010), both of which describe auditory tuning of ON1 up to 40 kHz (whereas (Kostarakos et al. 2009) only examines neural responses from 3-6kHz). We used Meazure 2.0 (written by Baron Roberts, C Thing Software) to capture data points across three audiograms. As the neural audiograms from these two studies were conducted for different frequency sampling points, we imported the captured data into MATLAB, used linear interpolation between frequency sampling points, and averaged across the three interpolated audiograms to produce an average T. oceanicus audiogram (Figure S3). We determined thresholds for song peak frequencies from the interpolated average neural audiogram. Following the approach described in (Wikle et al. 2025), we calculated attenuation curves over distance starting at peak amplitude for nine different frequency and amplitude combinations for each song type and determined the distance at which a particular song's sound pressure level reached the threshold identified from the average neural audiogram required to elicit a neural response. We calculated these estimated hearing distances for each song type.