The active space of sperm whale codas: inter-click information for intra-unit communication

Animal communication systems involve the interplay between intended interlocutors as well as the range over which communication signals can be received and decoded. This range is affected by the source properties of the emitted signals, ambient noise levels, sound propagation through the habitat, and the sensitivity and directionality of the receiver's auditory system (Bradbury and Vehrencamp, 1998). Hence, the resulting so-called ‘active space’, the range over which communication is possible, is the combined result of these different source properties, receiver hearing abilities and the properties of the environment in which the calls are made (Marten and Marler, 1977). Calls used in acoustic communication may be very weak to address only an audience within a body length, as for some insects, offering a very small active space (Nakano et al., 2009) or calls may be very powerful to facilitate communication across ocean basins, as in large baleen whales (Payne and Webb, 1971). Accordingly, the size of the active space of communicating animals offers critical insights into the size of the intended audience, and hence the social function of the communication system, in the face of eavesdropping predators, prey and competitors.

Whales have evolved complex acoustic communication where signals of different whale species have properties adapted to their geographic distribution, movement patterns and social structures (Tervo et al., 2012; Tyack, 1986): Large baleen whales communicate in tonal calls at very low frequencies which carry over vast swathes of the ocean to faraway conspecifics (Širović et al., 2007), while many species of dolphins communicate using high frequency whistles resulting in an active space of ∼1 km to address conspecifics nearby (Janik, 2000). Some groups of toothed whales, however, have adapted instead to communication using clicks at frequencies that are high enough to mitigate detection by predators at the expense of the active space, such as the burst pulse communication sounds made of narrow band high frequency clicks of harbor porpoises (Kyhn et al., 2013; Sørensen et al., 2018) or at source levels high enough to reach conspecifics tens of kilometers away with low frequency clicks, such as ‘slow clicks’ or ‘clangs’ produced by male sperm whales (Madsen et al., 2002a,b).

Sperm whales are a long-lived, socially complex species of toothed whale known for their deep diving behavior and extremely powerful biosonar system (Fig. 1; Møhl et al., 2003). To facilitate foraging for squid and other deep sea prey items, sperm whales have developed a large and highly specialized nose producing echolocation clicks that have the highest source levels of any animal (Cranford et al., 1996; Madsen, et al., 2002a,b; Møhl et al., 2003; Norris and Harvey, 1972). Unlike most other toothed whales, which make echolocation clicks with one set of phonic lips and tonal communication sounds with the other set (Ames et al., 2020; Madsen et al., 2013), sperm whales use their highly specialized sound production system with one phonic lip pair to emit clicks for both echolocation and communication (Fig. 2A) (Gordon, 1987; Madsen et al., 2002b; Mullins et al., 1988; Watkins and Schevill, 1977).

When clicking, sperm whales use their phonic lips to produce a single acoustic pulse (Madsen et al., 2023), which is partially transmitted into the water and partially reflected back and forth between the frontal and distal air sacs of the nose (Madsen et al., 2002a). These reflections form a multi-pulsed click as they are transmitted into the water (Fig. 2B) (Cranford et al., 1996; Gordon, 1991; Madsen et al., 2003; Møhl et al., 2003; Norris and Harvey, 1972; Zimmer et al., 2005). Owing to the relationship between the size-dependent spacing of an individual sperm whale's air sacs and the resulting inter-pulse intervals (IPIs), the pulse structure has been used to discriminate individual whales of different size in recordings, and sperm whales may similarly use the IPI of clicks to help identify conspecifics (Antunes et al., 2011; Gero et al., 2016a; Rendell and Whitehead, 2003; Schulz et al., 2011).

Sperm whales communicate in patterns of clicks called codas (Watkins and Schevill, 1977). Different coda types have been defined by the variation in the rhythm, tempo and number of clicks (Fig. 2A). Importantly, sperm whales have both long-term group members nearby and preferred inter-group associations (Christal et al., 1998; Gero et al., 2015; Whitehead and Weilgart, 1991), potentially creating a need for both short- and long-range communication via codas. While adult male sperm whales are known to roam over oceans to forage and are mainly solitary when not attempting to mate (Rice, 1989), female sperm whales live in highly stable, mostly matrilineal groups called units (Christal et al., 1998; Gero et al., 2008; Konrad et al., 2018; Mesnick, 2001; Mesnick et al., 2003). Units associate with other units within their vocal clan, a group comprising units sharing a common and distinct repertoire of codas that can exist sympatrically with other vocal clans (Gero et al., 2016b; Rendell and Whitehead, 2003; Whitehead and Weilgart, 1991). Within units, individual whales in the Pacific commonly organize into a ‘rank’, in which small clusters of one to three whales are spread out by a few hundred meters over a single axis facing the direction of travel. Ranks off the Galápagos Islands were estimated to be approximately 1 km long for all clusters visible from the research vessel (Whitehead, 1989, 2003). Gero et al. (2014) found similar unit spread amongst animals in the Caribbean, with standard deviation of left–right displacement of individuals from the mean trajectory of the unit being ∼500 m either side; however, individuals within units are frequently observed spanning larger distances as well. The distance over which the individuals in a unit are generally spread is thus no more than a few kilometers.

Early field studies of sperm whale communication estimated that codas have an active space of ∼600 m, based on practical experience of tracking units of female whales at sea using vessel-mounted hydrophones (Weilgart and Whitehead, 1997). These field experiences suggest that codas primarily serve to mediate behavioral transitions and facilitate within-unit individual and unit-level bonds, which have been shown to persist over long periods (Gero et al., 2015). Such a small active space is, in turn, incompatible with the proposed long-range coda communication for broadcasting clan identity beyond the unit (Gero et al., 2016a,b; Rendell and Whitehead, 2003).

Across ocean basins, units can be assigned to clans based on ‘identity codas’, which are specific coda types produced frequently within, but rarely outside, the clan (Hersh et al., 2022). Rendell and Whitehead (2003) suggest that identity codas could mediate the social segregation observed between clans by broadcasting a unit's clan identity to other units over larger distances before engaging in a close-proximity social interaction. However, the functional use of codas as potential communication signals among conspecifics within units versus long-range radiation of clan identity remains untested. Such a test requires investigating the click properties that define the active space, thereby providing quantification of the range at which receivers are able to decode the coda information.

Here, we quantify the active space, and hence the intended audience of codas, by combining measures of coda source levels with ambient noise levels and sound propagation modelling. Specifically, we test the hypothesis that the active space of coda communication is on par with the acoustic footprint of a single unit, estimated to be at least 1 km, to serve in intra-unit communication. We also test the hypothesis that the IPI information of a single click is not conserved over the active space of the coda as a whole, limiting its use as a marker of individual identity.

Fieldwork was conducted off the western coast of the island of Dominica (15°18′N, 61°23′W) between 13 April and 25 May 2018 from a 12 m auxiliary sailing vessel as a part of The Dominica Sperm Whale Project (see Gero et al., 2014). Social units of adult female and juvenile sperm whales (Physeter macrocephalus L.) were located acoustically using a towed two-channel hydrophone array and individuals were visually spotted at the surface via their blows.

Once near sperm whales, acoustic recordings were made with a vertical hydrophone array if two or more adult female sperm whales were either at the onset of a foraging dive or socializing at the surface while the vessel engine was turned off. Five bouts of coda production were successfully recorded on the array during 42 days at sea. The vertical hydrophone array consisted of five ST300 SoundTraps (Ocean Instruments, NZ) equally spaced at 20.0 m apart, with the top hydrophone at a depth of 10 m and the bottom hydrophone at a depth of 90 m. The SoundTraps had calibrated clip levels of 172 dB re. 1 µPa, and sampled sound at a rate of 288 kHz with 16-bit resolution. To allow for accurate time-syncing of the audio recordings from the five SoundTraps, non-acoustic synchronization pulses were sent out once per second by the topmost hydrophone to the other hydrophones through a cable connecting all of the SoundTraps (see Malinka et al., 2020 for detailed time synchronization methodology). At the surface, three floating buoys attached to the top of the array allowed it to float freely and at the bottom of the array, a 10 kg weight kept the array vertical in the water column. DST tilt sensors (Star-Oddi, Iceland) were placed at the top and bottom hydrophones to record the depth, tilt and temperature to calculate the linearity and tilt angle of the array, and to estimate the sound speed. A 219 MHz VHF radio transmitter (Advanced Telemetry Systems, MN, USA) aided array recovery via a 3-element Yagi antenna and a R1000 radio receiver.

Data were initially audited using Adobe Audition CC 2018 (San Jose, CA, USA) and codas were marked for analysis using a custom program, CodaSorter, which used a combination of a MATLAB R2017b script (Mathworks, MA, USA) and LabVIEW (v. 2015) program (National Instruments, TX, USA). Within CodaSorter, individual coda clicks on the hydrophone at 30 m were manually identified to avoid erroneously identifying surface reflections as clicks. This hydrophone was selected for initial auditing to minimize surface reflections. This process resulted in a dataset of 3316 clicks constituting 646 codas. These codas were then classified into known coda types using a custom MATLAB script (using similar methods to Gero et al., 2016b, after Ankerst et al., 1999). Codas which were not classified as a known coda type were excluded from the analysis to avoid erroneous inclusion of non-coda clicks misclassified as codas by CodaSorter.

Using custom scripts in MATLAB, single clicks were isolated from the full .wav files by extracting 2 s windows containing the focal click and surrounding noise. These windows were filtered with a 4-pole 0.5 to 20 kHz band-pass Butterworth filter. The signal-to-noise-ratio (SNR) of each click was measured using 0.5 s of noise sampled from 0.6 s to 0.1 s before the coda, and a window around the click from 0.5 ms before and 4 ms after its peak amplitude. Only clicks with an SNR of greater than 10 dB were included in the analysis of click. A Tukey window with taper/duration ratio of 0.1 was applied to each click segment to reduce spectral leakage before computation of spectra parameters. For each click, the duration of 90% of the energy within the click window, and the peak-to-peak received level (RL_P–P), RMS received level (RL_RMS), and received energy flux density (RL_ERD) within the 90% energy window were measured along with the RMS bandwidth, centroid frequency, and peak frequency. These values were averaged over all clicks in each coda to provide one representative value per coda.

In order to evaluate how intra-click information of whale identity based on IPI is preserved over distance compared with the active space of the main clicks in a coda, we calculated the ratio of the amplitude of the first pulse to the second for each click in the codas. To compute this amplitude decay ratio, the 2 s windows of raw click recordings were filtered with a 4-pole 3 to 20 kHz band-pass Butterworth filter to better reveal the second pulse of each click. The first two pulses were located by first identifying the maximum amplitude in the click file, then finding a second maximum at least 1.5 ms later than the first maximum, the smallest likely inter-pulse interval as described in Tønnesen et al. (2018). The log₁₀ transformed ratio of the peak amplitude of the first pulse to that of the second pulse was calculated to determine the decay on a dB scale between the pulses (sensuMadsen et al., 2002a).

For each deployment, the separate hydrophone channels of the array were time-synced and their recordings were combined into multitrack .wav files via the non-acoustic synchronization pulse using a custom MATLAB library (see Malinka et al., 2020). Click detection and classification on the other channels of the array were done in PAMGuard (www.pamguard.org) following the manual audits on the channel from the hydrophone at 30 m. Localization was done using the PAMGuard Large Aperture 3D Localizer module, which used time of arrival differences between the different channels of the array to determine the location of the whale in 2D space relative to the vertical array at the time the click was emitted. The ‘Mimplex’ algorithm used within the PAMGuard module is a combination of a simplex localization algorithm and a Markov Chain Monte Carlo simulation, described in detail in Macaulay et al. (2017). Only clicks localized to within 450 m, which is half the theoretical limit for the range that can be reliably estimated from an array of this length, were included in the following analysis. Localization to less than five times the array aperture leads to an extremely conservative localization error, with resulting SLs subject to <2 dB of error for this array type (Madsen and Wahlberg, 2007).

Under the assumption that the auditory system of sperm whales, like other toothed whales (Madsen et al., 2006), has a hearing sensitivity that is better than ambient noise levels at the frequencies where they communicate, we computed ambient masking noise as a lower proxy for the detection threshold of the coda clicks. Based on our measured distributions of the peak frequency and RMS bandwidth of coda clicks (Table 1), we used the octave band centered at 4 kHz (frequency range from 2.8 kHz to 5.7 kHz) as an estimate of the masking band for sperm whale coda clicks. Ambient noise (NL, dB re. 1 µPa, rms) was accordingly calculated using custom scripts in MATLAB in this octave band in 1 s bins. We computed these masking noise levels from individual SoundTraps deployed within the study area at three depths: 100 m, 450 m and 1000 m. Noise levels were calculated at 100 m to represent surface noise, 450 m to represent foraging depth and 1000 m to represent the maximum depth female sperm whales are likely to dive to at this location. We analyzed a total of 144 h of noise data from the three deployment depths to render probability density functions for the masking noise.

Because noise varies with depth, the detection threshold and hence active space also varies with depth and the average active space of a coda is accordingly defined by the probability of a listening sperm whale being at a particular depth. To estimate the probability of a listening sperm whale to be at a particular depth, we formed a cumulative distribution function for sperm whale time-at-depth from existing animal-borne sound and movement tag data (DTAG3; Johnson and Tyack, 2003) of individuals tagged off Dominica (our unpublished data; n=23 tag recordings of least 5 deep dives; Table S1). We used inverse transform sampling to obtain a sample of 100,000 depths reflective of this observed distribution.

Finally, for each sampled depth of sperm whale diving tag data, we estimated the extent of the coda active space: the maximum range-from-source at which the models predicted transmission loss to be below the ETL. This process resulted in a distribution of active space values consistent with the sampled habitat and the whales' observed depth distribution. We repeated the active space estimation procedure for three different source depths (2, 25 and 216 m) representing 5%, 50% and 95% of coda production depths in the DTAG data combined with the 5th, 50th (median) and 95th percentiles of ETL.

Twelve bouts of codas were recorded on the vertical hydrophone array. The duration and frequency parameters (Table 1) were calculated from the coda clicks individually identified on the channel located at 30 m depth for each deployment. Five deployments allowed for full synchronization and were used for coda click localization in PAMGuard. Only clicks localized to a range of <450 m from the array were included in further analysis to ensure lower errors (<3 dB TL) of localization ranges (Madsen and Wahlberg, 2007). This resulted in a final total of 1598 clicks from 444 different codas of 24 different known coda types. For summary statistics, a median value was taken from each coda's clicks, then a mean was taken again over all the codas, in order to avoid biasing the parameters towards coda types with a larger number of clicks (Table 1). Multiple whales were likely to be producing codas during each recording and the individual contribution of each whale is unknown, but as codas from different whales are easy to distinguish, all evaluations were calculated on the level of a single coda and all whales were in the same behavioral setting.

The overall median RMS apparent source level (ASL_RMS) was 161 dB re. 1 µPa (IQR of 14 dB between 153 to 167 dB re. 1 µPa; Fig. 3A). The masking noise level (NL) at three different depths was calculated to estimate the detection thresholds limiting active space at different biologically relevant depths (Fig. 3B). The 100 m noise level, selected to simulate noise levels near the surface, was 87 dB re. 1 μPa RMS (IQR of 84 and 92 dB re. 1 μPa). The 450 m noise level, representing noise levels at the approximate foraging depth of sperm whales in this area, was 82 dB re. 1 μPa (IQR of 80 to 85 dB re. 1 μPa). The 1000 m noise level, selected to represent the deepest depths sperm whales are likely to use in this area, was 73 dB re. 1 μPa (IQR of 6 dB between 70 and 76 dB re. 1 μPa). Median excess transmission loss, resulting from the convolution of ASL_RMS and NL distributions (Fig. 3C), at 100 m was 69 dB (5th percentile, 47 dB; 95th percentile, 97 dB; Fig. 3C). At the 450 m noise level, excess transmission loss was 74 dB (5th percentile, 55 dB; 95th percentile, 90 dB) and at the 1000 m noise level, excess transmission loss was 83 dB (5th percentile, 64 dB; 95th percentile, 99 dB).

Active space estimates, parameterized as the maximum range at which the models predicted transmission loss below the ETL, were calculated for 2 m source depth representing animals producing codas at the surface, 25 m source depth representing the median depth of coda production in our tag data, and 216 m source depth representing the depth above which only 5% of codas were emitted (Fig. 4). The ETL values were resampled based on the time spent at depth of listening whales from tag data and the probability distribution function of the estimates was plotted (Fig. 5). At a source depth of 2 m, using the 50th percentile ETL, the estimated active space was 4.3 km (5th percentile ETL, 0.1 km; 95th percentile ETL, 20.2 km). At a median source depth of 25 m, using the 50th percentile ETL, the estimated active space was 4.4 km (5th percentile ETL, 0.1 km; 95th percentile ETL, 21 km). At the very rare source depth of 216 m where less than 5% of codas were produced, using the 50th percentile ETL, the mean estimated active space was 19.6 km (5th percentile ETL, 0.3 km, 95th percentile ETL, 48.4 km).

To estimate the difference in active space between the dominant pulse in a coda click and the next pulse, we computed the amplitude decay between the first and second pulse for coda clicks in which the second pulse was obvious (Table 1; n=436 of 1598 total clicks). The mean decay was 9±4 dB, so IPI information is available for much shorter range (−9 dB) than the first pulse, corresponding to an active space reduction of ∼70%, assuming spherical spreading. Thus, median intra-click information in coda clicks probably only has an active space of approximately 1 km.

Here, we sought to investigate the intended audience and hence primary function of sperm whale communication by estimating the active space of their coda clicks. To do so, we combined in situ measurements of source levels and noise levels with transmission loss modeling and tag data on coda production depths and diving behavior. Our estimate for the median active space for coda communication is ∼4 km at the prevailing near-surface source depths of coda production, median receiver depths and natural ambient noise levels, which strongly supports our hypothesis that codas are used primarily for within-unit communication (Weilgart and Whitehead, 1997).

The median ASL_RMS of coda clicks in this study was measured at 161 dB re. 1 μPa (Table 1), which is considerably louder than most tonal calls of smaller toothed whales [Tursiops sp.: 146.7±6.2 dB re. 1 µPa (RMS) (Jensen et al., 2012); Pseudorca crassidens: 115–130 dB re. 1 µPa (peak–peak) (Thode et al., 2016); Sousa chinensis: 138.5±6.8 dB re. 1 µPa (RMS) (Wang et al., 2019)]. This places sperm whales among the loudest communicators within the toothed whales, but ∼20 dB weaker than the songs of male humpback whales (Au et al., 2001) and the moans of blue whales (Cummings, 1971), which also produce much longer communication sounds with much more energy for the same peak pressure level. It follows that the higher frequencies and weaker source levels of sperm whale codas offer a median active space of ∼4 km that is several orders of magnitude smaller than that of large baleen whales, such as fin and blue whales, that use powerful and very low frequency signals to communicate over hundreds to thousands of kilometers (Širović et al., 2007). Notably, some baleen whales, such as humpback whales also produce weaker non-song calls that are thought to be intended for closer-range communication and are produced at lower source levels [158.4 dB re. 1 µPa (RMS) (Dunlop et al., 2013); 137 dB re. 1 µPa (RMS) (Fournet et al., 2018)].

Sperm whales have the capacity to produce the highest sound pressure levels in the animal kingdom, with source levels of echolocation clicks measured in excess of 230 dB re. 1 μPa (RMS) with a directionality index of 27 dB (Møhl et al., 2003). In contrast, owing to their low-frequency emphasis, coda clicks are nearly omnidirectional (Madsen et al., 2002a). If echolocation clicks are the loudest clicks that sperm whales can make, then source levels of the loudest possible omnidirectional click for coda communication would be the source level, at their highest measured ∼230 dB re. 1 μPa (RMS), minus the directionality index of 27 dB, resulting in a maximum omnidirectional estimate of ∼200 dB re. 1 μPa (RMS). However, the RMS ASLs measured for coda clicks in this study were ∼161±10 dB re. 1 μPa (RMS), four orders of magnitude less intense. Even considering differences between sexes and body sizes, the fact that the whales are theoretically capable of making echolocation clicks that radiate with an energy 40 dB higher than the coda clicks measured here implies that the source levels of coda clicks are not limited by the biomechanics of sound production, but rather selected for to address a nearby audience. Indeed, if the primary intended audience of coda clicks is nearby conspecifics, using very high source levels of omnidirectional codas may give rise to exposure of the receiving auditory systems that are unpleasantly high at the typical few-body-lengths spacing during social interactions at the surface. It may also be disadvantageous to loudly radiate the position of a social unit with calves near the surface to potential eavesdropping predators, such as killer whales (Aguilar De Soto et al., 2012). Finally, if the precise timing or spectral tuning of clicks within codas carry intra-coda information, it may also be that coda click loudness is traded for acute motor control of the click timing in different coda types. It remains to be seen if the potential for making louder codas is harnessed to defend the realized active space in higher ambient noise levels from, for example, a passing cargo vessel (Hermannsen et al., 2014). Unless compensated for by louder codas, a cargo ship at 3 km from a sperm whale listening for codas will easily lead to a halving of the active space (Findlay et al., 2023), meaning that sperm whales, like many other marine mammals, probably face increasingly smaller active spaces in increasingly industrialized oceans (Richardson and Würsig, 1997).

The notion that codas serve to address a nearby audience is reinforced by the observation that sperm whales very rarely make codas at great depths where the active space is significantly larger than at the surface (Fig. 4). Thus, the finding that SLs that are much lower than physiologically possible and the shallow depths at which codas are produced both support the interpretation that sperm whales do not seek to maximize their active space. Instead, sperm whales may use these relatively low-powered coda clicks because it is only necessary for codas to propagate over the footprint of a sperm whale unit. In that context it is important to note that at 4 km, only half of the codas will be detected on average, so the active space of the full coda repertoire is smaller than the typical footprint of a sperm whale unit.

Although the typical spatial distance between sperm whale units at sea remains unclear and may vary based on the population, sperm whales in the Pacific, as well as off Dominica in the Eastern Caribbean, have been estimated to travel up to 50 km per day (Gero et al., 2014; Whitehead, 2001), implying a large degree of spatial separation between units of typically hundreds of kilometers. This supposition is further supported by the fact that units from different clans are not commonly observed on the same day in the same area, while members of different units of the same clan may frequently associate within meters of one another in the waters west of Dominica (Gero et al., 2014; 2015; Whitehead, 2003). These large spacings of units from different clans imply that between-clan communication would call for a median active space of codas several orders of magnitude higher than the median 4 km estimated here. Given that clan identification often requires detection and classification of coda repertoires, clan decoding from distant codas would thus require not only detection of one loud coda, but likely several of them to decode the clan affiliation, again reducing the range at which that is possible. Thus, our data show that sperm whale codas are not suited for long-range acoustic communication to radiate clan identity to other distant sperm whale units.

Sperm whale clicks are made up of a series of pulses of decreasing amplitude that are produced as a click is reflected off surfaces within the nose (Møhl et al., 2003). Many studies have used the IPIs between these pulses to discriminate between clicks from different sperm whales, suggesting that IPI might be similarly harnessed by the whales to also distinguish individuals. However, Schulz et al. (2009) found that the multi-pulsed structure of coda clicks is subject to pulse degradation off the acoustic axis, decreasing the probability that the IPIs would be distinguishable to conspecifics as an identity signal across long distances. Bøttcher et al. (2018) and Beslin et al. (2018) found significant IPI variation within individuals that was not fully explained by differences in recording aspect or automatic detection. To estimate the active space over which such IPI information may be communicated if sperm whales can solve the problem of forward masking of these closely spaced pulses in their auditory system, we measured a median decrease of 9 dB from the first to the second pulse in coda clicks. This implies that listening whales must experience a transmission loss that is smaller by the same 9 dB, meaning a decrease in active space by 70%. While potentially available for use by conspecifics at very close proximity when socializing at the surface, this decrease, coupled with the intra-individual variation and degradation off axis, makes it unlikely that IPI could be used as an identity signal at the kilometer-scale distances at which codas might otherwise be used for communication between whales and at which passive acoustic monitoring is frequently used.

Examining the features of a communication system can provide important insight into the social organization of the species who uses it, because the focus of the most important social signals should reflect the most important relationships (Kragh et al., 2019; Tyack, 1986). Humpback whales lack individualized calls and instead invest in songs that are shared over the population level, reflecting the relative unimportance of individual bonds compared with group bonds (Tyack, 1986). Bottlenose dolphins produce unique signature whistles which serve to identify them on the individual level, reflecting the importance of individual relationships within the population (King et al., 2014). Sperm whales appear to be somewhere in the middle, regarding the importance of individual relationships. Our investigations of pulse decay within coda clicks show that it is unlikely that information within click pulses are identity signals over longer ranges, but spectral features of their coda clicks may serve this purpose. Studies demonstrating the importance of individual relationships within units and the behavioral context of coda production in those relationships, coupled with the small active spaces found here, strongly imply that codas have an intra-unit communication function (Antunes et al., 2011; Gero et al., 2008; Schulz et al., 2008).

Some coda types have been identified in Pacific populations of sperm whales to serve as possible symbolic markers of clan identity (Hersh et al., 2022). Similar coda repertoire differences, which can be used to distinguish sympatric vocal clans, have been identified in the Eastern Caribbean population analyzed here (Gero et al., 2016b). However, all the recordings in this study came from contexts in which unit members were socializing in close contact with one another, while assumed not to be in acoustic contact with other units. Thus, it is possible that under behavioral contexts different from the one of this study, certain coda types may be produced at higher source levels better suited to long range communication between units. Further data collection during a wider variety of behavioral contexts, including when two units are known to be in acoustic contact, could shed light on the changes that may occur in coda production when units interact. However, if these hypothetical longer range, inter-unit focused codas of much higher SLs are produced, with resulting much larger active spaces, these codas should dominate coda repertoires in passive acoustic monitoring owing to the much higher probability of recording them, and there is currently no evidence of that in either tag or array data.

Interpreting the functional use of codas to mediate social interactions within or between units, to coordinate unit movement, or to exchange information about their environment, is a complex challenge requiring a detailed understanding of not only the acoustic characteristics of codas but also the wide range of behavioral contexts in which they are used. However, this study takes a new step in our understanding of the communication capacity of sperm whales in the field by quantifying the range over which these exchanges occur, to be understandable, and over which vocal cues might be interpretable by conspecifics. This, in turn, informs our understanding of the social interactions between sperm whales in the same and different units, shedding light on the unique social structures of these animals that face rising noise levels in Anthropocene oceans.

We dedicate this paper to the late Dr Roger Payne; a champion of large whales. We thank the Chief Fisheries Officer and the Dominica Fisheries Division officers for research permits and their collaboration in data collection, as well as Dive Dominica, Al Dive, and W.E.T. Dominica for logistical support while in Dominica. Thank you to all crews of the R/V Balaena and the Dominica Sperm Whale Project, including Hal Whitehead, Stef Weilgart-Whitehead, Taylor Hersh, Fabien Vivier and Felicia Vachon. The authors would also like to thank the Marine Bioacoustics Lab for their knowledge and feedback and the Section for Zoophysiology, Department of Biology, Aarhus University, which provided technical materials and expertise for hydrophone construction.