https://orcid.org/0000-0003-2923-1446
Abstract
Social foraging is a collective solution to the challenge of catching prey. A remarkable example involving different predator species with complementary hunting skills is Lahille’s bottlenose dolphins, Tursiops truncatus gephyreus foraging with net-casting human fishers to catch migratory mullet, Mugil liza. It remains unknown, however, to what extent dolphins coordinate their own actions when foraging with humans, and how intraspecific coordination impacts interspecific foraging success. Using drone-based tracking, we quantified dolphin group surfacing behaviors (diving synchrony, proximity, and heading angles between individuals) and tested the repeatability of these behavioral metrics across independent human–dolphin cooperative foraging interactions. We then quantified how the variance and consistency in these behaviors influenced the likelihood of fishers catching mullet. We found repeatable patterns in dolphin group proximity and heading angles across cooperative foraging interactions with fishers, and that fishers were more successful at catching mullet when dolphins approached them along different trajectories with consistent diving synchrony. These findings suggest that groups of dolphins can be, to some extent, coordinated when fishing with humans, thereby influencing the collective outcome of this cooperative foraging tactic. We consider hypotheses for the underlying mechanisms of synchrony and coordination within groups of dolphins and highlight solutions for overcoming the inherent challenges associated with quantifying collective foraging dynamics in natural settings.
A key driver of the formation of animal groups is increased foraging efficiency (Anderson and Franks 2001; Beauchamp 2014). The collective effort of hunting for prey is expressed in many ways, from loose foraging aggregations to collaborative and specialized groups (Beauchamp 2014) that vary in degrees of socialization, communication, specialization, resource sharing, and dependence (Lang and Farine 2017). Social foraging is widespread across taxa (Lang and Farine 2017; Duguid and Melis 2020) but is more common in species with complex cognition, social skills, and communication (Anderson and Franks 2001; Bailey et al. 2013; Wooster et al. 2023), which may facilitate within-group behavioral coordination to produce a mutually beneficial outcome when foraging together (Duguid and Melis 2020). Social foraging can also occur in mixed-species groups (Goodale et al. 2020; Cram et al. 2022), especially when species-specific foraging tactics complement one another, but in most cases, it remains unclear whether coordinated actions among the individuals of one species influence the foraging success of the other.
Iconic cases of mixed-species social foraging benefitting both species include hunting associations between coyotes, Canis latrans and badgers, Taxidea taxus targeting small mammals in the American West (Minta et al. 1992), and the coordinated hunting of groupers, Plectropomus pessuliferus and giant moray eels, Gymnothorax javanicus in the Red Sea (Bshary et al. 2006). These associations capitalize on complementary hunting skills that prove most efficient in different types of structures in their habitat. More rarely, interspecific social foraging involves human cooperation with wildlife, whereby individuals of each species rely on mutually understood cues and complex social skills to synchronize their actions (Cram et al. 2022; van der Wal et al. 2022). For instance, in sub-Saharan Africa, honeyguide birds, Indicator indicator and human honey hunters, Homo sapiens exchange acoustic signals to coordinate their joint foraging (Spottiswoode et al. 2016). Whereas the skilled birds lead the group to bee nests hidden in tree trunks, skilled humans use tools such as axes and smoke to access the nests, subdue the bees, and extract the honey, leaving behind the highly caloric beeswax, a prized resource for the honeyguides (Isack and Reyer 1989). However, these cases usually involve single foragers working together (e.g., between the eels and groupers) or leading groups of heterospecifics to food (e.g., honeyguides leading groups of human foragers). Rarer are cases in which groups of conspecifics coordinate their actions whereas also foraging socially with heterospecifics. A notable example involves joint fishing between artisanal human fishers and wild dolphins (Simões-Lopes et al. 1998).
This human–dolphin cooperation is a century-old tradition in southern Brazil, passed down through multiple generations of humans and dolphins (Peterson et al. 2008; Simões-Lopes et al. 2016). Lahille’s bottlenose dolphins, Tursiops truncatus gephyreus hunt migratory mullet, Mugil liza by herding their schools toward the shallow and murky estuarine waters, where net-casting fishers wait for cues from the dolphins. These cues, which include behaviors such as a sudden deep dive, tail slap or head slap, are interpreted by humans as the appropriate time and place to cast their nets over the mullet schools (Simões-Lopes et al. 1998). Fishers catch more mullet when they time their casts with the cues made by the dolphins’; in return, dolphins can optimize their foraging effort by targeting the mullet trapped in the nets (Cantor et al. 2023). Whereas recent research revealed how this interspecific foraging works—i.e., how individuals of both species synchronize their actions and reactions to catch prey together (Simões-Lopes et al. 1998; Cantor et al. 2023)—far less is known about the nature of the intraspecific interactions involved in this tactic. It remains unclear if and how individuals of each species coordinate their behavior among themselves when foraging cooperatively with individuals of the other species. On the human side, interviews reveal that artisanal fishers engage in both competitive and cooperative interactions among themselves when fishing with dolphins (Peterson et al. 2008; Santos-Silva et al. 2022). On the wildlife side, individual dolphins may form affiliative social groups (Machado et al. 2019), whereas agonistic and competitive interactions are apparently rare (Simões-Lopes et al. 1998; Machado 2024). However, when fishing with humans, do dolphins in groups coordinate their foraging efforts, or do individuals act solely of their own volition? And can dolphin group coordination, or lack thereof, impact the outcome of the interspecific cooperative foraging with humans?
Here, we investigate the extent to which dolphin groups coordinate their foraging efforts when herding mullets toward human net-casting fishers and whether such intraspecific coordination impacts the interspecific foraging success. We assess group coordination in dolphins by using drone-based tracking to measure surfacing behavior (diving synchrony, proximity, and heading angles among group members) during foraging interactions with net-casting fishers to test if these behaviors are consistent across independent foraging events. We then test whether synchronized behaviors (diving in time with each other, and swimming close together and parallel to each other) among dolphins influences the probability of fishers capturing mullets. We conclude by discussing hypotheses for the behavioral synchrony and coordination among dolphins, and how future studies can overcome the inherent challenges of quantifying collective foraging dynamics in the marine environment.
Materials and Methods
Observational and ecological sampling of interspecific foraging
We sampled human–dolphin foraging by recording the surface behavior of dolphins interacting with the net-casting artisanal fishers in the canal of the Laguna lagoon system, southern Brazil (Figure 1A). Our data sampling spanned 18 consecutive days in May and June 2018 during daylight hours. This period encompassed the peak of the mullet reproductive migration season, during which this primary target prey is most abundant and accessible to both dolphins and fishers (Simões-Lopes et al. 1998). For each hour of observation, we recorded in situ environmental conditions, which likely influence mullet migration (Herbst and Hanazaki 2014; Lemos et al. 2014): water temperature (underwater thermometer of 0.1°C precision at 1 m depth), wind speed and direction (anemometer Instrutherm AD-250), current speed and direction (from a weather station inside the lagoon system).
Sampling methods to investigate potential coordination among dolphins that fish with humans in southern Brazil. (A) At the main foraging site (green circle), on the edge of the canal connecting the lagoon system in Laguna to the Atlantic Ocean, we deployed (B) a multiplatform sampling system to record the foraging interactions between artisanal net-casting fishers (top) and wild bottlenose dolphins (center). We (i) conducted all-event observational behavioral sampling to count numbers of dolphins and fishers, time their foraging interactions and quantify the outcome of their foraging in terms of mullet caught; (ii) used adaptive resolution imaging sonar (blue cone) to sample prey availability (i.e., mullet schools); (iii) used photography to quantify dolphin group sizes through individual photo-identification; and (iv) used drone-based behavioral tracking to assess consistency in dolphin group behaviors (as a proxy for coordination), based on 3 surfacing behavior metrics: group proximity, group heading, and diving synchrony (definitions in Table 1). The map was created using the R package “maps” 3.4.1 (Becker et al. 2023).
Definitions of dolphin group surfacing behavior metrics. Surface behavior was measured from drone videos (Figure 1B) to investigate patterns, which suggest dolphin group coordination during the foraging interaction with artisanal human net-casting fishers
| Surfacing behavior metric | Defined as | Measured as | Visual concept |
|---|---|---|---|
| Group proximity | Distance between dolphins at the surface | Distance between dorsal fins of all dolphins at the surface (in meters) |  |
| Group heading | Angle between dolphins at the surface | Difference between the trajectories of all dolphins at the surface relative to the shore (in degrees) |  |
| Group diving synchrony | Time between individuals surfacing | Difference between surface times of individual dolphins (in seconds) |  |
| Surfacing behavior metric | Defined as | Measured as | Visual concept |
|---|---|---|---|
| Group proximity | Distance between dolphins at the surface | Distance between dorsal fins of all dolphins at the surface (in meters) | |
| Group heading | Angle between dolphins at the surface | Difference between the trajectories of all dolphins at the surface relative to the shore (in degrees) | |
| Group diving synchrony | Time between individuals surfacing | Difference between surface times of individual dolphins (in seconds) | |
Definitions of dolphin group surfacing behavior metrics. Surface behavior was measured from drone videos (Figure 1B) to investigate patterns, which suggest dolphin group coordination during the foraging interaction with artisanal human net-casting fishers
| Surfacing behavior metric | Defined as | Measured as | Visual concept |
|---|---|---|---|
| Group proximity | Distance between dolphins at the surface | Distance between dorsal fins of all dolphins at the surface (in meters) | |
| Group heading | Angle between dolphins at the surface | Difference between the trajectories of all dolphins at the surface relative to the shore (in degrees) | |
| Group diving synchrony | Time between individuals surfacing | Difference between surface times of individual dolphins (in seconds) | |
| Surfacing behavior metric | Defined as | Measured as | Visual concept |
|---|---|---|---|
| Group proximity | Distance between dolphins at the surface | Distance between dorsal fins of all dolphins at the surface (in meters) | |
| Group heading | Angle between dolphins at the surface | Difference between the trajectories of all dolphins at the surface relative to the shore (in degrees) | |
| Group diving synchrony | Time between individuals surfacing | Difference between surface times of individual dolphins (in seconds) | |
Each day from 09:00 to 17:00, we conducted land-based all-event behavioral sampling (Altmann 1974) to record foraging interactions between dolphins and fishers. These interactions were defined as instances when groups of dolphins approached the margins of the 100-m beach in the Laguna canals (Cantor et al. 2023) and performed behaviors that fishers recognized as cues to cast their nets (Simões-Lopes et al. 1998). Groups were defined by all dolphins actively foraging together within, but not swimming past the 100-m interaction beach site; we then used standard photo-identification protocols (Urián et al. 2015) to distinguish individual dolphins and define group membership. For each human–dolphin foraging interaction, we recorded (Figure 1B): the exact time when the dolphin performed the cue; the dolphin group size and number of individuals specifically foraging with fishers; the number of individual dolphins giving cues to fishers; the number of fishers active in the water to fish with dolphins; the number of nets, which fishers cast in response to each dolphin’s cue; and the outcome of the human–dolphin foraging interaction. We quantified the outcome of interaction when the fishers brought the nets to the beach, after each cast in response to the dolphin cue: We inspected all nets cast by all fishers and counted the number of mullets caught in each net. The human–dolphin foraging interaction was considered successful when mullets were caught or unsuccessful otherwise.
Observational and underwater sampling of prey availability
To account for the influence of prey availability on the outcome of the foraging interaction, we developed a proxy of mullet availability relative to environmental conditions as follows (full details in Cantor et al., 2023). First, we estimated the abundance of mullets in estuarine waters characterized by low visibility and high turbidity. We achieved this by deploying an adaptive resolution imaging sonar (ARIS 3000, SoundMetrics) from the bank of the canal where the fishers stand (Figure 1B). This sonar camera emits 128 sound beams at 3.0 MHz and transforms their echoes into digital images, covering the relevant area for this human–dolphin foraging (Cantor et al., 2023), which spans 3–20 m from where the fishers stand into the canal (Figure 1B). Next, we processed the continuous sonar footage to remove background noise, retaining only the mullet schools (see Tarling et al. 2022). We then converted the processed images into echograms, where pixel intensity indicated the presence of mullet, and classified each minute of the underwater footage as either containing or lacking mullet. Finally, we built a generalized linear mixed-effect model with a binomial error structure and logit link function to relate the proportion of minutes per hour with mullet detections as a function of the environmental conditions that can change throughout the day and affect local mullet abundance: wind speed and direction, tide state and velocity, water transparency and temperature (Herbst and Hanazaki 2014; Cantor et al. 2023). In this model estimating prey availability, days of observation were included as random intercepts to account for between-day variability, and the sonar range served as an offset to account for variation in the sampling effort (see Cantor et al. 2023). We used the predictions derived from this model as a proxy for the environmental conditions, which represent high and low availability of mullet at the interaction area (full details in Cantor et al. 2023).
Overhead sampling of intraspecific group coordination
We recorded human–dolphin foraging interactions using commercially available drones (DJI Phantom 3) to investigate potential coordination within groups of dolphins (Figure 1B). Drone operations were only carried out during favorable weather conditions (no rain, good visibility, and wind speed below 10 knots) and observing all safety flight guidelines (e.g., Fiori et al. 2017). To minimize any disturbance to the dolphins, we maintained the drone altitude at more than 60 m (Christiansen et al. 2016) and kept the drone stationary, with the camera pointing directly downward. All videos included 2 reference scales of known size: A 1-m stick placed alongside the fishers and the 110-m distance between the 2 rocky jetties defining the beach in the canal of the lagoon system.
Next, we played back all full 20-min drone videos at 10 times the normal speed to scan for the human–dolphin foraging interactions, i.e., the exact moment when a dolphin performs the foraging cue and fishers respond by casting nets. To ensure independence between these foraging interactions, we only considered interactions, which were separated by at least a 3-minute window before and after the exact moment of the cue performed by dolphins (as in Cantor et al. 2023). This way, an independent drone video sample was defined as a 3-min clip leading to the human–dolphin foraging interaction. After discarding foraging interactions with solitary dolphins, the dolphin group size (i.e., number of photo-identified individuals) across all interactions recorded by the drone ranged from 2 to 6 individuals (mean ± SD = 3.87 ± 1.23). We opted to discard the few (8.3%) foraging interactions with more than 5 dolphins due to the inaccuracy in manually tracking and distinguishing multiple individuals in the overhead drone videos, given that individual identification is made from the side, based on photos of natural marks of the dorsal fin (Urián et al. 2015), and that linking these 2 independent data streams remains challenging (Machado and Cantor 2022). Our final sample size was 45 3-minute drone videos containing independent human–dolphin foraging interactions, spaced out by at least 6 min.
To quantify the intraspecific coordination within groups of dolphins, we manually processed these 45 drone video samples using the BORIS event logging software (Friard and Gamba 2016). In the context of cooperative foraging, coordination refers to the synchronized actions and spatial relationships among individuals within foraging groups (Lang and Farine 2017). Based on a framework for studying social predation (Lang and Farine 2017), we defined and measured 3 surface behavioral metrics as proxies for coordination among the dolphins foraging in a group (Table 1; Figure 1B): (i) group proximity (how closely individuals surfaced to each other), (ii) group heading (how aligned their movement trajectories were when they surfaced), and (iii) group diving synchrony (how synchronized their surfacing intervals were).
In each instance where dolphins surfaced, we chose one clear frame and measured (i) proximity as the distance between the tips of their dorsal fins in pixels, then converted to meters using both the 1-m reference stick and the 110-m beach length as reference scales (e.g., Machado and Cantor 2022). Group proximity for each frame was calculated as the median distance in meters between all individual dolphins at the surface. In each instant where dolphins surfaced, we chose one clear frame and measured (ii) heading as the angle created by the dolphin’s dorsal median line (from rostrum to fluke) perpendicular to the beach directly. The group heading for each surfacing instance was calculated as the median angle difference between all pairs of dolphins at the surface (if dolphins were heading in the same direction, i.e., left or right, we subtracted the smaller angle from the larger one; if they were heading in different directions, we summed the angles). Finally, to measure (iii) diving synchrony, we calculated the mean difference in surfacing time between individual dolphins in each surfacing instance by subtracting the time when the first dolphin’s dorsal fin broke the water’s surface from the time when each subsequent dolphin surfaced.
Repeatability of dolphin group behavior
We estimated the repeatability of dolphin group surfacing behavior—proximity, heading, and dive synchrony (Table 1)—during their foraging with humans. The repeatability analyses (Nakagawa and Schielzeth 2010) allowed us to determine the proportion of variation in these behavioral metrics, which could be attributed to each video sample containing a human–dolphin foraging interaction. A low repeatability estimate would indicate a low level of variance (Nakagawa and Schielzeth 2010; Wolak et al. 2012) and thus a high level of consistency in these surfacing behaviors across foraging interactions, which we took as a proxy for group coordination.
To estimate the repeatability of dolphin group surfacing behavior, we fitted linear mixed models using each of the 3 surfacing behavior metrics as response variables (Stoffel et al. 2017). We specified the number of surfacing dolphins in each measurement as a fixed effect and treated human–dolphin foraging interactions as random intercepts. Repeatability is then expressed as a ratio of the variance components, ranging from zero to one, with higher repeatability values indicating that more of the variance can be attributed to the random effect (Nakagawa and Schielzeth 2010; Wolak et al. 2012), which in our case is the human–dolphin foraging interactions. To assess the uncertainty of repeatability estimates, we used parametric bootstrapping (1,000 iterations) to calculate 95% confidence intervals. In addition, we used the likelihood ratio to test for the statistical significance of the grouping factor (i.e., the human–dolphin foraging interaction) by comparing fits of models with and without this factor.
Influence of dolphin group behavior on human–dolphin foraging
To test our hypothesis, which coordinated dolphin behaviors contribute to joint foraging success with humans, we evaluated whether variance and consistency in group behavior across foraging interactions increase the probability of catching mullet. To estimate group behavior variance, we summarized the dolphin behavior metrics for each human–dolphin foraging interaction by extracting the coefficients of the random intercepts from the linear mixed models used to estimate repeatability. To estimate group behavior consistency, we used the absolute values of these coefficients, as the positively valued coefficients indicate exactly how much the behavioral metrics in each human–dolphin foraging interaction deviated from the global mean (regardless of direction). A zeroed coefficient would indicate that the behavioral metric showed a consistent pattern with the global mean, whereas a non-zeroed value would indicate how much the behavioral metrics in a foraging interaction deviated from consistency (hereafter “consistency deviation”).
A strong relationship between the absolute values of these coefficients and foraging outcome would suggest that an increase or decrease in the behavioral metrics would increase or decrease the probability of a successful human–dolphin foraging interaction. By contrast, a strong negative relationship would suggest that the closer a foraging interaction is to the global mean (which is zero and represents the most repeated pattern, i.e., consistency), the higher the probability of a successful foraging interaction. In other words, a consistent pattern in dolphin group behavior (suggesting coordination) could increase the probability of a successful human–dolphin foraging interaction. Any relationship different from this suggests that consistency is not important for the outcome of the human–dolphin foraging tactic.
We then built a set of generalized linear models (GLMs) with a binomial error distribution and logit link function to test if a consistent pattern in the dolphin group behavior, represented by coefficients close to zero, affects the outcome of the human–dolphin foraging whereas accounting for group size and prey availability. The response variable was the outcome of the foraging accounting for fishing effort (number of nets cast), represented by the proportion of fishers’ net casts that capture at least 1 mullet in each foraging interaction. The predictors were the 3 coefficients of the random effects as proxies for consistency when close to the global mean for all dolphin group surfacing behavioral metrics (proximity, heading, and dive synchrony). We also controlled for prey availability and dolphin group size by introducing 2 other fixed effects into these models: The predictions from the model fitted for mullet availability based on environmental variables (described above) and dolphin group size (i.e. the number of foraging individual dolphins identified through photo-identification). We included these predictors as quadratic terms to account for the possibility that foraging success is optimized at intermediate levels of prey availability and number of dolphins, in which both lower and higher levels lead to decreased human–dolphin foraging success. Due to the relatively small size of our dataset, we did not include 2-way interaction terms to avoid convergence issues and unreliable model fits. Using a stepwise backward selection, we departed from the full model, dropping 1 predictor at a time based on Akaike’s information criterion (AIC) values and a chi-squared test for an analysis of deviance. We built a set of eight models, with an additional null model containing only the intercept, and selected the most parsimonious model based on the lowest AIC corrected for small samples (AICc) (Johnson and Omland 2004).
To ensure the validity of the best-fitting model, we checked for collinearity among predictors by computing the variance inflation factor, assessed its goodness-of-fit, and visually inspected the model residuals (Zuur et al. 2009). To validate the assumptions of this model, we followed the protocol for linear models (Bolker et al. 2009) and simulated residuals for 1,000 datasets from the fitted model to test if the distribution of the scaled residuals deviated from the expected distribution (Hartig and Lohse 2022). All data analyses were carried out in R version 4.1.0 (R Core Team 2021) using the native package stats, and the rptR (Stoffel et al. 2017) and DHARMa (Hartig and Lohse 2022) packages.
Results
Dolphin group foraging behavior
In the 45 independent human–dolphin foraging interactions recorded by the drones, proximity within dolphin groups ranged from 0.3 to 61.9 m (mean ± SD = 4.62 ± 8.17, n = 277 observations); differences in group heading ranged from 0.0 to 178.8° (mean ± SD = 28.8 ± 28.68, n = 277 observations); and dive synchrony ranged from 0 to 22 s (mean ± SD = 1.76 ± 2.41, n = 180 observations). These data were moderately imbalanced in terms of number of surfacing instances in the drone frames (median = 13, SD = 5.86), ranging from 4 to 35 instances per human–dolphin foraging interaction (Figure S1). To address this imbalance and avoid issues when calculating variance within groups of dolphins, we disregarded foraging interactions with fewer than 4 dolphin behavioral measurements (Figures S2 and S3).
Next, we omitted the interactions lacking data on foraging outcomes and were left with 9 successful (at least 1 fisher-caught fish) and 20 unsuccessful (no fish caught) human–dolphin foraging interactions. There were overall differences in the dolphin surface behavioral metrics between successful and unsuccessful foraging interactions, despite variation within foraging interactions. On average, successful foraging interactions involved slightly fewer individuals (i.e., smaller dolphin group sizes: successful = 3.52 individuals ± 0.99 SD; unsuccessful = 3.80 ± 1.25; mean number of dolphins at the surface: successful = 1.58 ± 0.74; unsuccessful = 1.74 ± 0.84). Successful foraging interactions also showed slightly lower group proximity (i.e., larger distances: successful = 6.86 meters ± 12.30 SD; unsuccessful = 4.17 ± 6.66), slightly larger angle differences (i.e., greater difference in group heading: successful = 39.63 ° ± 52.45 SD; unsuccessful = 24.80 ± 32.91), and slightly larger time intervals between group members at the surface (i.e., lower diving synchrony: successful = 2.06 s ± 3.08 SD; unsuccessful = 1.68 ± 2.08) in comparison to unsuccessful foraging interactions.
Dolphin group behavior can be consistent when foraging with humans
Low repeatability occurs when the coefficients consistently stay close to the zero axis (global mean), which happens for group heading and diving synchrony but not for group proximity. The distribution of the random intercepts of these models, and how each deviated from their respective mean random intercept, are given in Supplementary Figure S4, whereas the linear mixed models estimating repeatability of dolphin group surface behavior are summarized in Figure 2. The first model (Figure 2A) indicated that dolphin group proximity was moderate (R = 0.328, SE = 0.071, 95% CI bootstrap = 0.176–0.454), but significantly, repeatable (likelihood ratio test, proximity: logLik = -955.772, D = 35, df = 1, P = 1.65 × 10–9). This model then suggested some consistency in group proximity across human–dolphin foraging interactions but with some interactions deviating from this consistent pattern (Supplementary Figure S4). The second model (Figure 2B) indicated that the group heading had low (R = 0.093, SE = 0.046, 95% CI = 0.003–0.186) and significant repeatability (logLik = –1397.915, D = 5.1, df = 1, P = 0.012). This model suggested that despite some variation across foraging interactions (Supplementary Figure S4), there was some consistency in the dolphin group heading across human–dolphin foraging interactions. By contrast, dolphin diving synchrony (Figure 2C) exhibited very low (R = 0.0103, SE = 0.024, 95% CI = 0–0.083) and nonsignificant repeatability (logLik = –642.755; D = 0.097, df = 1, P = 0.378), probably due to very low variation in the data. In summary, the significant repeatability of group proximity and group heading indicated that each of these metrics tended to be similar across human–dolphin foraging interactions, meaning that foraging interactions explained little of the variance in group proximity or heading. Therefore, dolphin groups tended to exhibit repeated patterns in proximity and heading, but less so in diving synchrony, as they approached the fishers.
Repeatability of dolphin group behavior across human–dolphin foraging interactions. Repeatability estimates of the 3 dolphin surfacing behavior metrics: (A) group proximity, (B) group heading, and (C) group diving synchrony (definitions in Table 1). Blue circles and dashed lines indicate the empirical repeatability index; whiskers represent the 95% confidence intervals of the bootstrap repeatability distribution. The repeatability index ranges from 0 (low variance between human–dolphin foraging interactions) to 1 (high between-foraging interaction variance).
Dolphin group behavior can influence the outcome of foraging with humans
Finally, we investigated the influence of dolphin group behavior on the outcome of the human–dolphin foraging interactions. The best-fitting binomial GLM (Supplementary Table S1) suggested the outcome of the interactions was influenced by prey availability, by variance in dolphin group heading angles, and by consistency deviation in diving synchrony across human–dolphin foraging interactions (Figure 3A and B). The variance in group heading had a positive effect on the proportion of nets successfully catching mullet (β = 0.134; z = 3.015; P = 0.003, Figure 3C), whereas the consistency deviation in group diving synchrony had a negative effect (β = –17.240; z = –2.175; P = 0.027; Figure 3D; Supplementary Table S2). Foraging interaction success also showed a quadratic relationship with availability of mullet (Figure 3E): The proportion of successful nets was higher at low mullet abundance (β = -2.480; z = –1.995; P = 0.046), decreased at intermediate levels, and exhibited a slight but nonsignificant increase at higher abundance (β = 2.624; z = 1.777; P = 0.076; Supplementary Table S2).
Predicted effects of the repeatable dolphin behavior on the outcome of the foraging with artisanal fishers. (A) A group of 2 dolphins, in close proximity and with similar headings, give a behavioral cue that artisanal net-casting fishers identify as the right time and place to cast their nets (photo by F.G. Daura-Jorge). (B) The outcome of a human–dolphin foraging interaction can be measured by the success of fishers in catching mullets, their shared prey with dolphins (photo by D.R. Farine). (C) The most parsimonious GLM (Supplementary Table S2) indicates that the success of human–dolphin foraging (based on the proportion of casting nets that caught mullet) tends to increase with disparities in dolphin trajectories (i.e., larger heading angle differences between group members) as they herd mullet and approach the fishers. The same model predicted that (D) increased consistency deviation in group diving synchrony (i.e., a greater time differences between dolphins surfacing) and (E) increased mullet availability had overall negative effects on the success of the human–dolphin foraging.
This best fitting model met all assumptions (Supplementary Figure S5) and showed moderate improvement over the null model, with a pseudo-R2 value of 34.2%. No other models received sufficient support from the data (delta AIC > 2.00; Supplementary Table S1), reinforcing that dolphin group size, consistency in heading and proximity, and variance in diving synchrony and proximity did not significantly affect the outcome of the human–dolphin foraging interactions. In summary, these findings suggested that foraging interactions where dolphin groups approached fishers with large heading angles between group members and maintained some consistency in diving synchrony were associated with higher chances of fishing success, particularly when mullet abundance was relatively low in front of the fishers.
Discussion
By investigating how dolphins coordinate their foraging behavior when engaging in unique interspecific cooperation, we found repeatable patterns in the within-group proximity and heading of dolphins as they herd fish schools toward artisanal net-casting fishers. These data suggest that the chance of their joint success increases when dolphins approach fishers in groups whose individuals move in different trajectories and with consistent diving synchrony. In what follows, we place these findings in the context of coordinated foraging in cetaceans, exploring hypotheses related to the mechanisms of synchrony and coordination within groups of dolphins that forage with fishers. Additionally, we address the challenges and potential solutions associated with quantifying the collective foraging dynamics of these marine predators.
Coordinated foraging in cetacean groups
As predators in a 3-dimensional marine environment, whales and dolphins face the challenge of finding and catching ephemeral and patchy prey (Benoit-Bird 2023). This challenge often promotes collective decision-making (e.g., Zwamborn et al. 2023) and a diverse array of social foraging tactics that require synchronization, coordination, and/or collaboration among group members (Bailey et al. 2013; Lang and Farine 2017). For instance, baleen whales such as humpbacks, Megaptera novaeangliae can engage in a highly coordinated production of bubble nets to trap prey (Wiley et al. 2011), whereas blue whales, Balaenoptera musculus can call one another on finding large prey aggregations (Cade et al. 2021). Among toothed whales, spinner dolphins, Stenella longirostris coordinate group movements to collectively encircle large fish schools in deep pelagic waters and take turns feeding on the tightly packed prey (Benoit-Bird and Au 2009). Tightly coordinated group foraging can also be found among killer whales (Orcinus orca). Some ecotypes move and dive in fine synchrony to create waves that wash their desired pinniped prey from sea pack ice (Visser et al. 2008), whereas others coordinate as entire groups to intentionally strand on and take prey from beaches (e.g., Lopez and Lopez 1985). In all such cases, group members move and dive in similar trajectories, often almost synchronously, suggesting that multiple individuals attempt to align their actions in both time and space, thereby concentrating on performing similar actions. Whereas in many cetacean social foraging tactics the collective success appears to depend on such a high level of within-group “synchrony” and “coordination” (sensu Boesch and Boesch 1989; Bailey et al. 2013), other social foraging tactics may involve individuals performing different (e.g., Boesch 2002) or complementary actions (“collaboration”, sensu Boesch and Boesch 1989; Bailey et al. 2013).
This apparent lack of surface behavior coordination or synchrony is well illustrated by some of the innovative social foraging tactics employed by bottlenose dolphins—such as the one we studied here. When hunting large, fast-moving schools of mullet, bottlenose dolphins have come up with diverse social foraging tactics (Wells 2019), some of which include complementary actions by the group members. In Cedar Key, Florida, a “driver” dolphin herds mullet schools towards “barrier” dolphins and all group members can benefit from catching mullet that attempt to escape by leaping into the air (Gazda et al. 2005; Hamilton et al. 2022). In other bottlenose dolphin populations, that physical barrier is made of suspended substrate. For example, in Florida Bay and parts of the Caribbean, a “ring-maker” dolphin strikes its tail near the seafloor to create a mud ring that entraps the fish school for the benefit of all group members who will feed on leaping mullet (Torres and Read 2009; Engleby and Powell 2019; Ramos et al. 2022). Whereas at the surface it may appear that the group’s actions and movements are unsynchronized or uncoordinated, these foraging tactics instead involve role specializations (e.g., Gazda et al. 2005). Our data from southern Brazil suggest that dolphins in groups can coordinate to some extent and raise the hypothesis that group members could also perform complementary actions when foraging with artisanal fishers.
Potential coordination in dolphin groups foraging with humans
In human–dolphin cooperative foraging, dolphin groups could in theory behave in a range of ways when herding mullet schools toward net-casting fishers: From highly coordinated groups—whereby dolphins approach fishers with aligned trajectories, tight group cohesion, and synchronized dives—to highly uncoordinated groups. In highly coordinated groups, individuals would closely relate in time and space to each other’s behaviors, thereby performing equivalent foraging actions; in uncoordinated groups, by contrast, individuals act independently (e.g., Boesch and Boesch 1989; Bailey et al. 2013). Our findings on repeatable patterns in within-group proximity and heading angles among dolphins fall between these extremes. Our models indicate that 2 out of 3 group behavior metrics—proximity and heading—exhibit significantly low repeatability across human–dolphin foraging interactions and that successful interactions are associated with groups of dolphins moving along different trajectories with some consistency in diving patterns (i.e., dolphins diving at different angles and coming up to breath roughly at the same time). Whereas a group of dolphins acting identically may indicate individuals working together toward a shared foraging goal, coordinated foraging can also manifest when individuals complement each other’s actions (Bailey et al. 2013; Lang and Farine 2017). Thus, a hypothesis emerges that variability in individual trajectories just before the key moment in this interspecific foraging tactic (i.e., the dolphin cue perceived by fishers as an indication of prey location) suggests that dolphins within a group may perform complementary actions when foraging with humans.
The disparity in individual dolphin trajectories, as they approached the net-casting fishers, could represent a collective effort to herd and attack the prey from multiple directions. Notably, a group of dolphins attacking a fish school from different angles has been shown to be an efficient hunting tactic (e.g., Benoit-Bird and Au 2009). Whereas the current data are suggestive of individual dolphins performing complementary actions that contribute to their collective foraging success, we cannot yet rule out other nonexclusive explanations. A similar pattern in surface behavior could emerge in a competitive context in which group members opportunistically take advantage of the foraging efforts of others. For instance, if one dolphin takes the lead in herding the mullet school and cues the fishers, and other scrounger dolphins attack the school from different angles to benefit from the first dolphin’s efforts, the different approaching trajectories could be misinterpreted as collective complementary actions. Whereas these hypotheses remain to be properly tested, our current study suggests that such a pattern of group foraging seems to be important for the success of the human–dolphin foraging tactic as a whole. Next, we explore these unresolved questions and propose avenues for future research to address them.
Caveats and ways forward
Whereas our approach generates hypotheses regarding dolphin group coordination and synchrony, it currently lacks a means to directly test their underlying mechanisms. One caveat of this study is that we could not control for potentially relevant individual covariates, such as sex, age class, and differences in foraging skill. Individual dolphins in Laguna vary in their abilities to herd fish schools toward fishers and cue them in (e.g., Daura-Jorge et al. 2012; Cantor et al. 2018; Machado 2024), as long recognized by the net-casting fishers who distinguish between “good” and “bad” dolphins based on this ability (Peterson et al. 2008). Differences in sex and age could further reflect differences in foraging skills; for instance, younger male dolphins could be generally less experienced in interacting with fishers than adult females. Although we have photo-identified the dolphins as they fish with humans, and estimates of the abovementioned traits exist for some individuals (e.g., Machado et al. 2019), a persistent challenge in our current study was linking these photo-identifications (based on natural markings of the dorsal fin as shot from the side) to the individual dolphins in the drone videos (seen from above) (Machado and Cantor 2022), thereby limiting our ability to make any individual-based inference on group coordination. Analyses accounting for these individual traits will reveal the extent to which individual dolphins participate in more vs. less coordinated groups when foraging with fishers, which will enable testing the hypothesis on the existence of within-group complementary actions. This will further allow testing whether these actions are indicative of foraging role specialization, and whether such roles are randomly or consistently assumed by specific individuals (Lang and Farine 2017)—as seen in other dolphin foraging tactics (e.g., Gazda et al. 2005; Torres Ortiz et al. 2021). A further challenge arose in manually tracking several dolphins with certainty as it was nearly impossible to tell individual dolphins apart as they dove in the murky waters. This limitation of quantifying the surface behavior of larger groups may be alleviated in the future with the advance of new deep learning tools for automatizing multiple objects tracking in drone videos (e.g., Koger et al. 2023).
Another caveat of our study was inferring group coordination based on intermittent behavior seen at the water surface. Shedding further light on the potential cooperation among dolphins as they fish with humans would require an additional, continuous metric for group coordination. Given that it remains challenging to continuously track dolphins from drone footage in Laguna because of the very low estuarine water visibility (Cantor et al. 2023), a potential solution would be tracking the foraging behavior of individual dolphins acoustically (e.g., King et al. 2021; Hamilton et al. 2022; Macaulay et al. 2022). In the bottlenose dolphin population with clear evidence for foraging role specialization (Gazda et al. 2005), individual dolphins emit more whistles just before engaging in the driver-barrier social foraging tactic (Hamilton et al. 2022). These findings raise the possibility that whistles may be used to assist in group coordination (King et al. 2021), or that individuals may use whistles to signal their motivation and recruit others to participate in social foraging (Duguid and Melis 2020). Among the Laguna dolphins, individual-based acoustic analyses could reveal whether community-level differences in whistle production (see Romeu et al. 2017) scale down to the individual level and are used as communication signals involved in group coordination when fishing with humans.
Finally, tracking dolphin acoustic behavior during cooperative foraging with humans would also provide insights into the foraging success of the group members. So far, our study has relied on the mullet caught by fishers as a proxy for the joint success of humans and dolphins; further studies recording dolphin echolocation efforts, especially “terminal buzzes,” will indicate prey capture attempts by the dolphins (Pirotta et al. 2013; Cantor et al. 2023). This way, tracking terminal buzzes at the individual level could shed light not only on the foraging success on the dolphin side but also on variation in success within a group of dolphins. With this approach, one could quantify the extent to which individual dolphins benefit or compete when foraging in groups with fishers—a first step at disentangling whether individual dolphins act independently in proximity to, or with intentional coordination with, other group members (sensu Duguid and Melis 2020).
Closing remarks
Coordination and synchrony among wild foraging animals can be logistically challenging to test empirically (e.g., Duguid and Melis 2020), especially when trying to understand the cognition and intent of the individuals involved (e.g., Melis and Raihani 2023; Wooster et al. 2023). These challenges become more complex in the constraints of a wild marine system (e.g., Hansen et al. 2022), particularly in murky estuarine waters. Despite such challenges, our study provides novel insights into a unique foraging tactic between wild dolphins and humans: Our data suggest that groups of dolphins can, to some extent, be coordinated when fishing with fishers, with implications for the success of their cooperative foraging tactic. However, given the current caveats, we acknowledge that concrete evidence for coordination and cooperation currently only exists for the interspecific part of this foraging tactic—that is, among dolphins and fishers (Cantor et al. 2023). Novel conceptual frameworks (e.g. Melis and Raihani 2023; Zwamborn et al. 2023) and technological advances for collecting and processing multiple streams of behavioral data in animal collectives (e.g., Machado and Cantor 2022; King and Jensen 2023) are poised to shed further light on the movement and communication among dolphins and, by extension, on the underlying mechanisms of their fine-scale coordination in foraging tasks (Duguid and Melis 2020). These further efforts will reveal the extent to which the success, and consequently the persistence, of this rare human–wildlife cooperation relies on the dolphins’ collective behavior.
Acknowledgments
We thank the dedicated scientists who were on the ground in Brazil collecting data and facilitating field logistics (P.V. Castilho, B. Romeu, C. Bezamat, J. Valle-Pereira, P.C. Simões-Lopes, and others), and the insightful comments by our colleagues at Universidade Federal de Santa Catarina, Oregon State University, Université de Lausanne (C.R. Teixeira, T.A. Hersh, and C.T. Beck), and the 2022 UCC Marine Biology Master’s cohort.
Funding
This study is part of the Long-Term Ecological Research Program “Sistema Estuarino de Laguna e Adjacências” (SELA-PELD) funded by the Conselho Nacional de Pesquisa e Desenvolvimento Tecnológico (CNPq 445301/2020-1). It received support from the PROBRAL Research Program (23038.002643/2018-01) funded jointly by Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) and Deutscher Akademischer Austauschdienst; and from the Deutsche Forschungsgemeinschaft, Centre for the Advanced Study of Collective Behaviour, Germany’s Excellence Strategy (EXC 2117–422037984). M.C. received support from CNPq (53797/2016-9), CAPES (88881.170254/2018-01), Department for the Ecology of Animal Societies at the Max Planck Institute of Animal Behaviour; National Geographic Society (WW210R-17 and NGS-101549R-23); and the Marine Mammal Research Program Fund at Oregon State University. F.G.D.-J. received research funding from CNPq (308913/2022-0), CAPES (88887.374128/2019-00) and Fundação de Amparo à Pesquisa do Estado de Santa Catarina (2021TR387 and 2021TR000581). D.R.F. received funding from the European Research Council European Union’s Horizon 2020 (850859), and the Swiss National Science Foundation Eccellenza Professorship (PCEFP3_187058).
Authors' Contributions
K.M.: Conceptualization; Data curation; Investigation; Methodology; Visualization; Writing—original draft, review & editing. F.G.D-J.: Conceptualization; Data collection; Formal analysis; Software; Funding acquisition; Writing—review & editing. A.M.S.M.: Conceptualization; Data collection and curation; Formal analysis; Investigation; Methodology; Software; D.R.F.: Data collection; Funding acquisition; Investigation; Methodology; Project administration; Writing—review & editing. E.R.: Investigation; Resources; Supervision; Writing—review & editing. M.C.: Conceptualization; Data collection and curation; Formal analysis; Funding acquisition; Investigation; Methodology; Project administration; Resources; Software; Supervision; Validation; Visualization; Writing—original draft, review & editing.
Conflicts of Interest
None.
Ethics Statement
All data were collected using non-invasive methods (observations and multimedia recordings) without capturing animals or collecting biological tissues. Research permits for recording dolphin behavior were obtained from the Brazilian Ministry of Environment (SISBio #47876-1, #64956-1), and all national legislation and guidelines for observing and recording wild cetaceans were followed. Behavioral sampling involving human subjects was also collected non-invasively, ensuring confidentiality of any sensitive information, and only after the study’s goals were explained and informed consent was received. Research permits for recording artisanal fishers’ behavior and their fishing success were issued by the Committee for Ethics in Research with Humans of the Universidade Federal de Santa Catarina, Brazil (CEPSH#52308116.9.0000.0121, #06457419.6.0000.0121).
