Directional hearing sensitivity for 2–30 kHz sounds in the bottlenose dolphin (Tursiops truncatus)

Dolphins and other toothed whales (odontocetes) primarily use sound to hunt and interact with conspecifics underwater. The bottlenose dolphin (Tursiops truncatus) produces a variety of sounds including broadband (10–150 kHz) echolocation clicks and social communication sounds such as whistles (0.8–24 kHz) (Lilly and Miller, 1962; Lilly, 1967; Caldwell et al., 1990; Kaplan and Reiss, 2017). Research on the hearing abilities of odontocetes has primarily focused on ultrasonic frequencies (>20 kHz) because these animals depend on high-frequency biosonar to capture prey (Au et al., 1980; Nachtigall, 1980; Nachtigall et al., 1980). However, communication sounds less than 30 kHz are commonly used by many odontocete species (Lilly, 1962; Essapian, 1963; McCowan and Reiss, 1995; Schultz et al., 1995; Herzing, 1996; Corkeron and Van Parijs, 2001; Van Parijs and Corkeron, 2001; Simard et al., 2011). Communication sounds play a role in mating, feeding, and group social cohesion, and echolocation clicks might also broadcast information to conspecifics such as the direction of travel (Miller, 2002; Lammers and Au, 2003). Human activities such as commercial shipping, military exercises, and pile-driving produce noise that might interfere with odontocete communication (Richardson et al., 1995; Hildebrand, 2005; Erbe et al., 2016). To predict potential acoustic impacts on odontocetes due to anthropogenic noise, it is necessary to better understand how sounds below ∼30 kHz are received; especially how they vary with the orientation of the animal's body relative to the sound source, i.e., receiving beam pattern or directionality.

Although directional hearing and sound localization have been extensively studied in humans and other terrestrial mammals [see Gourevitch (1980) and Blauert (1988)], there have been few studies in marine mammals and those have mainly focused at higher frequencies. Au and Moore (1984) investigated directional reception of 30, 60, and 120 kHz tones by projecting masking noise at different angles relative to the signal in the horizontal and vertical planes. As the frequency increased, sound reception became more directed toward the front and below the midline. That study did not examine sound reception for frequencies below 30 kHz. In another study, the harbor porpoise (Phocoena phocoena) exhibited direction-dependent hearing sensitivity in the horizontal plane for 16, 64, and 100 kHz sounds (Kastelein et al., 2005), but the receiving beams were broader than the dolphin for the same frequencies. Popov and Supin (2009) also found a broader receiving beam for auditory-evoked potential thresholds in the beluga whale (Delphinapterus leucas) compared to the dolphin, which exhibited direction-dependent thresholds for 8 and 16 kHz sounds.

In an experiment by Schlundt and colleagues (2004), underwater behavioral hearing thresholds at 2, 8, and 12 kHz were measured when the location of the sound source was positioned either 1 m in front of the dolphin (on the midline axis) or 1 m below the location of the dolphin's ears. For 2 kHz tones, detection thresholds were lower when the sound was projected from below the ears as compared to in front; however, for 12 kHz tones, thresholds were lower for sounds projected from in front as compared to below (Schlundt et al., 2004). These differences in threshold based on the position of the sound source suggest that the dolphin's receiving sensitivity for frequencies lower than 30 kHz depends on the relative direction of the sound source.

The aim of the present study was to characterize the bottlenose dolphin directional sensitivity to sounds with center frequencies of 2, 10, 20, and 30 kHz, which were projected from different angles in 45° increments encompassing the 360° vertical and horizontal planes. It was hypothesized that dolphins would exhibit direction-dependent hearing sensitivity consistent with Au and Moore (1984), and that hearing thresholds for lower frequencies would be less dependent on source location compared to higher frequencies.

This study followed a protocol approved by the Institutional Animal Care and Use Committee of the Naval Information Warfare Center Pacific and the Navy Bureau of Medicine and Surgery and followed all applicable U.S. Department of Defense guidelines for the care and use of animals.

Subjects were two bottlenose dolphins: BLU (female, 52-yr old, ∼204 kg), and TYH (male, 36-yr old, ∼192 kg). Both dolphins had previous experience with behavioral hearing threshold collection. Prior to participating in this study, both subjects had some pre-existing hearing impairment. BLU showed a normal “U-shaped” audiogram consistent with age-related reduction in the upper-frequency cutoff, with an upper cutoff of hearing ∼50 kHz. TYH not only showed an age-related reduction in his upper-frequency cutoff (∼78 kHz), but also displayed elevated thresholds across all frequencies; however, his hearing was suitable for the frequency range of the sounds used in this study. Based on the animal ages and frequency pattern of hearing loss, it seems likely the hearing loss was cochlear in origin.

Dolphins stationed underwater on a plastic “biteplate” attached to polyvinyl chloride (PVC, schedule 80) and wood frame, which was secured to the deck of a 9 m × 9 m floating netted enclosure located in San Diego Bay, CA. The dolphin and biteplate were oriented either upright for horizontal measurements or rotated 90° clockwise for vertical plane measurements (see Fig. 1.). The biteplate was used to achieve consistent positioning across many experimental sessions. For TYH, a tail-rest was used during vertical plane data collection to control body positioning. The sound projector was located 1.9 m away from the dolphin's head at a depth of 1.4 m at any of the angular positions shown in Figs. 1(b) and 1(c). A coordinate system was used whereby the origin was the midpoint between the dolphin's two ears, and the x axis was aligned with the dolphin's longitudinal axis [see Fig. 1(a)] (American National Standards Institute, 2012). The sound projector housing was suspended from the deck of the floating pen from the eight fixed positions: 0°, 45°, 90°, 135°, 180°, 225°, 270°, and 315°. The projector positions were named in accordance with the horizontal plane angle, ϕ. In the horizontal plane, the 45° position was located to the dolphin's right side, and the 315° position was located to the dolphin's left side. The 0° position was directly in front of the dolphin and the 180° position was directly behind the dolphin. In the vertical plane, the 90° projector position (θ = 0°) was located above, and the 270° projector position (θ = 180°) was located below the dolphin's head.

Acoustic stimuli were 500 ms linear frequency-modulated upsweeps (10% bandwidth) with 10 ms onset–offset ramps to reduce transient distortion (Finneran and Schlundt, 2007). Center frequencies were 2, 10, 20, and 30 kHz. Sounds were generated digitally on a personal computer using custom LabVIEW-based software (National Instruments Corporation, Austin, TX), converted to analog at a rate of 500 kHz, attenuated if necessary (custom), filtered, and emitted through an underwater sound projector. Sounds less than 30 kHz were presented via a LL916 speaker (Lubell Labs, Whitehall, OH), while 30 kHz sounds were presented via an ITC1032 piezoelectric transducer (International Transducer Corp., Santa Barbara, CA). A piezoelectric transducer (TC4013, Reson Inc., Slangerup, Denmark) located at the underwater listening position (between the ears, without the dolphin present) was used to calibrate the sound projector before and after each test session, and record and listen to sounds during sessions [Figs. 1(c) and 1(d)]. Calibrations were performed for each projector position, frequency, and plane combination before and after each session. If post-session calibrations differed from the pre-session calibrations by less than 2 dB, no adjustments were made. If the difference between pre- and post-session calibrations was 2–5.5 dB, then thresholds were corrected by adding 1/2 of the pre-post difference or excluded from further analysis if the difference was greater than 5.5 dB. At least three thresholds from each frequency at each angle and plane were collected.

Directional hearing thresholds were determined psychophysically using a “go/no-go” detection task, where the dolphin reported the presence of the acoustic stimulus by producing a conditioned whistle (hit) or withholding the response in the absence of a stimulus (correct rejection). Failure to respond to a stimulus was classified as a miss. Whistles produced in response to stimulus-absent trials (catch trials) or prior to presentation of the stimulus in signal-present trials were classified as false alarms. The highest acceptable false alarm rate was 33%. The dolphins were given equal amounts of fish reinforcement for hits and correct rejections and were recalled to the trainer on a miss or false alarm.

Stimulus levels were adjusted using an up–down modified staircase procedure (Cornsweet, 1962). Stimulus presentation at the test frequency began 15–30 dB above the dolphin's predicted hearing threshold (based on previously collected behavioral audiogram data and the relative presentation angle). The initial step size was 4 dB, followed by a 2 dB step size after the first reversal, defined as the transition from hit to miss or miss to hit. The dolphin's response window began 50 ms after stimulus onset and ended 2 s after the stimulus onset. The trial order was based on a pseudorandom sequence with a 70% chance of a signal present trial. Threshold was defined as the mean stimulus sound pressure level (dB re 1 μPa) over the final six reversal points after reaching a plateau. Sessions were rejected from analysis if the standard deviation of the final six reversals was 3 dB or greater.

The order of conditions tested were neither systematic nor completely random. Typically, 6–9 thresholds were collected from each subject per experimental day. The angular position and plane were not randomly varied because it required relocating the bite plates and sound projector. Within and between experimental days, the projector position was progressively moved away from the front (ϕ = 0°) and toward the back (ϕ = 180°). Unique combinations of projector position, plane, and frequency were not repeated in a single day, so all averaged thresholds were calculated using data that were collected across at least three different days.

The dominant sources of ambient noise in San Diego Bay were snapping shrimp (Synalpheus parneomeris), other dolphins, and vessels. During data collection, the experimenter monitored ambient underwater sound subjectively (within the audible human frequency range) by listening to an in-air speaker connected to the TC4013 hydrophone, which was located inside the experimental pen during sessions. Experimental sessions were paused when the experimenter noticed approaching vessels or when noise from other dolphins was elevated. Sessions were resumed when audible noise subsided. After data collection, noise was estimated objectively using two methods. First, sound recorded from the TC4013 hydrophone during stimulus-absent trials from each session was analyzed in 1/3-octave bands. Second, a SoundTrap (HF300, Ocean Instruments, New Zealand) located outside of the experimental pen recorded 15 s of ambient background noise once every 5 min. Recordings from both the TC4013 and the SoundTrap (576 kHz sample rate, 16-bit resolution) were used to estimate noise levels, and sessions were eliminated if they occurred during a period of excess noise, i.e., levels greater than 1.5 times the interquartile range above the upper quartile. Thresholds were eliminated from further analysis when either the 1/3-octave noise band surrounding the stimulus frequency or the broadband root-mean-square noise from the session met the criterion of excess noise.

The DI describes the directionality of a receiver relative to an omnidirectional receiver in the same acoustic conditions. The DI is defined as

where D is the directivity factor, also called the integrated beam pattern. D is the ratio between the maximum, on-axis sensitivity of a directional receiver and the sensitivity of an omnidirectional receiver in an equivalent sound field (a DI value of 0 dB indicates a uniform, or omnidirectional sound reception field). DI does not consider beam pattern shape and assumes symmetry about the major axis. According to the American National Standards Institute (2012), D is defined as

where $b(θ,ϕ)$ is the beam pattern. Because it is not possible to measure the hearing thresholds for all directions, the DI must be approximated. This was accomplished using the method of Tylka and Choueiri (2016) where equally spaced measurements (thresholds obtained every 45°) were weighted according to the surface area of the portion of the sphere that they represent. The approach used to calculate and normalize weights is described in detail elsewhere (Kendig and Mueser, 1947; Tylka and Choueiri, 2016). To summarize this approach, the surface area of the portion of the sphere represented by each averaged threshold can be visualized as a strip around the sphere for non-zero angular positions, and as a circle for the endpoints of the sphere. The weight attributed to each measurement depends on the separation between measurements and the resulting area of the unit sphere. The angle where the lowest threshold was measured for each subject-frequency combination re-set the origin of the coordinate system. Here the approximated DI (in dB) is given by

where w_n is the direction-dependent weight, which was calculated for each measurement (i.e., averaged threshold), and N is the number of measurements taken along each of the two orbits m (m = 0, 1 for the horizontal and vertical planes, respectively) (Tylka and Choueiri, 2016). The response of the receiver, or the averaged behavioral hearing threshold obtained in this study is H(ω, ϕ$,θ$⁠) where ω is the angular frequency in rad/s, ϕ is the horizontal angle and $θ$ is the vertical angle, and H is in units of pressure (Pa). The on-axis response is denoted by H_0,0(ω) = H(ω, 0, 0). The approximated DIs were calculated using Eq. (3) for each subject-frequency combination.

Behavioral hearing thresholds were obtained for 2, 10, 20, and 30 kHz at horizontal and vertical angles surrounding the dolphin in 45° increments (see Fig. 2). Between September 2017 and March 2018, 537 behavioral hearing thresholds were obtained. Noise measurements across all sessions averaged 79, 77, 75, and 72 dB (re 1 μPa²/Hz) for the 1/3-octave bands surrounding 2, 10, 20, and 30 kHz, respectively. A total of 27 thresholds were eliminated from the analysis due to excessive ambient background noise, or for failing to meet the other criteria (see Sec. II), and all 510 remaining thresholds were used in the analysis. Thresholds at all frequencies were calculated at each of the 16 positions for a total of 128 averaged thresholds (see Fig. 2, Table I for BLU and Table II for TYH). False alarm rates calculated over the final six reversals within individual sessions ranged between 0% and 33%. The false alarm rate averaged across all subjects and sessions was 2%, thus the animals in this study adopted a conservative response bias (Schusterman, 1974).

The intraclass correlation coefficient (ICC) was estimated to assess between-subject differences (using RStudio Version 1.1.383 – 2009–2017 RStudio, Inc., Package ICC Version 2.3.0, ICCest function). Between subject reliability was poor [ICC = 0.24, 95% confidence interval (CI) = (0.06, 0.99) α = 0.05, k = 254.8], so data from the subjects were examined individually when appropriate in the subsequent analyses.

Hearing threshold asymmetry was found in the vertical plane when thresholds collected from angles above the midline [θ = 45°, 0°, and 315°, see Fig. 1(c)] were contrasted with thresholds obtained from angles below the midline (θ = 225°, 180°, and 135°). For this analysis, data from on-axis positions θ = 90° and θ = 270° were omitted. An analysis of variance (ANOVA) accounting for error due to frequency and subject indicated that whether sound was projected from above or below had a significant effect on threshold (F = 63.7, p < 0.001). The median threshold measured across all subjects and frequencies was 99 dB for sounds projected from above and 90.5 dB for sounds projected from below. An average of a 7.7 dB higher threshold was obtained across frequencies and subjects when the projector was situated above as compared to below. The above–below threshold differences for each dolphin and frequency are shown in Fig. 3(a), and pooled differences are reported in Table III.

Hearing threshold asymmetry was found in the horizontal plane when thresholds collected from angles in front of the dolphin [ϕ = 0°, 45°, and 315°, see Fig. 1(b)] were contrasted with thresholds obtained from angles behind the dolphin (ϕ = 135°, 180°, and 225°). For this analysis, data from ϕ = 90° and ϕ = 270° were omitted, but data from the vertical plane where θ = 90° and θ = 270° (equivalent to horizontal angles ϕ = 0° and ϕ = 180°, respectively), were included. An ANOVA accounting for error due to frequency and subject indicated that whether sound was projected from in front or behind had a significant effect on threshold (F = 200.7, p < 0.001). The median threshold measured across all subjects and frequencies was 90 dB for sounds projected from in front and 101.5 dB for sounds projected from behind. An average of a 9.3 dB higher threshold was obtained across frequencies and subjects when the projector was situated behind as compared to in front. The behind–in front threshold differences for each subject and frequency are shown in Fig. 3(b) and pooled differences are reported in Table III.

Across subjects and frequencies, the largest directional threshold differences were observed when thresholds collected directly in front of the dolphin (ϕ = 0° and θ = 90°) were subtracted from thresholds collected directly behind the dolphin (ϕ = 180° and θ = 270°). Across subjects and frequencies, thresholds collected behind and on-axis were an average of 12.6 dB higher compared to in front. These on-axis threshold differences are plotted for each subject and each frequency in Fig. 3(c). The 2 kHz stimulus did not produce a shadowing effect for BLU, but TYH's 2 kHz thresholds increased by 10 dB for sounds projected from directly behind as opposed to directly in front. The magnitude of attenuation increased with frequency until a maximum effect was observed for BLU at 20 kHz (18.5 dB) and TYH at 30 kHz (17 dB) [see Fig. 3(c)].

Thresholds for each frequency varied with the horizontal angle at which sound was projected relative to the dolphin. On average, hearing thresholds increased 3.5 dB for every 45° increment away from ϕ = 0° in the horizontal plane. The results shown in Fig. 2 suggested that threshold differences might exist between left (ϕ = 315°, 270°, and 225°) and right (ϕ = 45°, 90°, and 135°) sides of the horizontal plane. All thresholds collected from each side of the horizontal plane were compared, and data from ϕ = 0° and ϕ = 180° were omitted. Individual subjects were analyzed separately, and Wilcoxon signed rank tests for paired samples showed significant right–left asymmetry across frequencies for BLU (W = 444.5, p < 0.001), but not TYH (W = 1365, p = 0.1). TYH exhibited 5.2 dB right–left asymmetry for 2 kHz sounds but showed 9.6 and 7.6 dB higher average thresholds on the left for 10 and 30 kHz sounds, respectively. TYH's 20 kHz thresholds were less than 1 dB different between right and left. BLU's right–left threshold differences are reported in Table III. Tables S1 and S2 show angle-specific right–left asymmetries for each subject.¹

The DI estimates for each subject and frequency are shown in Fig. 4. The DIs averaged across subjects were 3.2, 7.9, 13, and 11.6 for 2, 10, 20, and 30 kHz stimuli, respectively. The DI estimates averaged across subjects from this study are plotted with data from Au and Moore (1984) in Fig. 4. A linear regression model was fit (RStudio Version 1.1.383 – 2009–2017 RStudio, Inc., Version 2.3.0, lm function) to the log-transformed data, including DIs from Au and Moore (1984) and each individual subject's estimated DI from this study (see Table IV). The results showed a positive linear-log relationship between DI and frequency (R² = 0.74, p = 0.0007). This was a good correlation between DIs calculated here and those by Au and Moore (1984) despite different subjects, environments, stimuli, and data collection/computational methods.

The results presented here show that bottlenose dolphins receive frequencies between 2 and 30 kHz in a direction-dependent manner. Frequency-dependent shadowing was found in both the vertical and horizontal planes and left–right asymmetry was observed. The DIs and vertical plane asymmetries were generally consistent with the results of previous studies in dolphins (Au and Moore, 1984; Schlundt et al., 2004; Popov and Supin, 2009). These results have implications—beyond basic biological considerations—for acoustic impact modeling in environmental conservation efforts and identify areas for future research such as spatial release from masking.

Dolphin hearing was more sensitive for sounds projected from underneath [see Fig. 3(a), Table III], and both subjects showed similar frequency-dependent dorsoventral asymmetries. These results concur with a previous study which demonstrated that hearing sensitivity for 30, 60, and 120 kHz sounds declined more rapidly for angles above the head than below (Au and Moore, 1984). The present results showed that both subjects had the best sensitivity to the 30 kHz stimulus when sound was projected from the midline (θ = 90°); however, both subjects were most sensitive to 2, 10, and 20 kHz stimuli when they were projected from below the midline at θ = 135°. This suggests that the main axis of the receiving beam, and perhaps also sound propagation pathways to the ear, are different for sounds of 20 kHz and lower.

Another study with this species showed that dolphins have 9–14 dB lower thresholds for 2 kHz sounds when presented from below (θ = 180°) as compared to in front (θ = 90°) (Schlundt et al., 2004). The present data show that 2 kHz thresholds obtained at θ = 180° were only 2 dB lower than those obtained from θ = 90° (see Tables I, II, and Fig. 2). The discrepancy between the two studies in these in-front vs below differences might be attributable to stimulus duration or distance from the acoustic source. The stimuli were twice as long in this experiment and the underwater sound projector was located farther away. It is also possible that ambient noise during data collection was higher than when data were collected by Schlundt et al. (2004).

Vertical plane asymmetries in odontocete hearing are likely the result of an acoustically sensitive lower-jaw (Brill et al., 1988; Supin and Popov, 1993). There are multiple proposed pathways for peripheral sound conduction in odontocetes. Specialized “acoustic fats” within or associated with the lower jaw create a pathway to the middle ear (Varanasi and Malins, 1971; Cranford et al., 1996; Au et al., 1998; Aroyan, 2001). It has also been demonstrated that while higher-frequency sounds are received best via the lower jaw (Norris, 1968; Brill et al., 1988; Supin and Popov, 1993), the dolphin receives sounds of around 30 kHz and lower in frequency at a location near the auditory meatus (Renaud and Popper, 1975; Brill et al., 1988; Kastelein et al., 1997; Ketten, 2000; Branstetter and Mercado, 2006; Popov et al., 2008). Popov et al. (2006) found that the horizontal angle where the lowest-threshold responses (auditory-evoked potentials) were measured varied with frequency, which provided further evidence of multiple sound-conduction pathways. Recently, Sysueva et al. (2017) used electrophysiological methods to measure auditory sensitivity on different areas of the bottlenose dolphin jaw, and found a maximally sensitive area at the meatus for a 32 kHz stimulus. While the present study was not designed to assess optimal sound reception pathways, our results show that the auditory receptive field in the vertical plane was directed toward the lower jaw and/or the “gular pathway” below and between the left and right lower mandibles (Cranford et al., 2008).

Asymmetry was found for sound reception in the horizontal plane in this study. BLU exhibited significantly lower thresholds on the left side across frequencies (see Fig. 2, Table III, Table S1¹); monaural hearing loss in BLU's right ear, which might have contributed to asymmetry in her receiving beam (Popov et al., 2006), could not be ruled out. Brill et al. (2001) tested acoustic sensitivity in one female bottlenose dolphin and also found that acoustic sensitivity was lower on the left for 10, 30, 60, and 90 kHz sounds (Brill et al., 2001; Cozzi et al., 2017). In the future, it would be interesting to compare behavioral and physiological thresholds from each ear in a single individual. Horizontal receiving beam asymmetry has also been described in the harbor porpoise (Kastelein et al., 2005), and might be explained by different sensitivities of the right and left ears and/or asymmetric skull morphology. Popov and Supin (2009) found that the dolphin receiving beam is more directional than the beluga whale and posited that anatomical differences in skull and head morphology might contribute to this difference. Previous work also showed that the harbor porpoise, which lacks the elongated skull of the dolphin, also has a less directional receiving beam than the dolphin (Kastelein et al., 2005).

Figure 4 shows the DIs obtained in this study with those obtained from Au and Moore (1984). In the present study DI was approximately 3.2, 7.9, 13, and 11.6 dB for the 2, 10, 20, and 30 kHz stimuli, respectively. These results were consistent with the general trend reported previously (Au and Moore, 1984; Kastelein et al., 2005; Popov and Supin, 2009) in that DI decreases with increasing frequency (see Fig. 4). The data presented by Au and Moore (1984) would suggest that the dolphin's directional hearing might approach a DI of zero around 10 kHz. However, with the new data presented here, the trend shows a more gradual slope to the linear-log relationship between frequency and DI, suggesting that dolphins have more directional hearing for lower frequencies than was expected based on earlier studies.

The frequency-specific anterior–posterior threshold differences observed here (see Fig. 4) suggests that dolphins might receive relatively dampened sound levels when they are oriented 180° away from a source. Furthermore, dolphins were less sensitive to sounds projected from above. This suggests that the angular orientation of animals with respect to the sound source could be included in acoustic impact modeling. The mechanism for anterior–posterior sound attenuation is likely reflection from air spaces such as the lungs and other structures inside the body such as tendon or bone that have considerably different densities and acoustic properties than the soft tissues (Aroyan, 2001). If the lungs and/or body structures reflect sound energy, then it follows that the size of the animal would change the amount of attenuation. In our study, thresholds were collected at a depth of 1.4 m, but animals might not experience similar attenuation at greater depths when the lungs are compressed and occupy a smaller volume (Ridgway et al., 1969). This is a topic that warrants further investigation.

Directional hearing allows the listener to separate a signal from noise based on the spatial location of the signal and noise sources (spatial release from masking). Human-generated (anthropogenic) sounds can potentially mask the wide variety of social and foraging sounds that are critical for the survival of dolphins in their natural environment (Richardson et al., 1995; Hildebrand, 2005; Erbe et al., 2016). Naval, oceanographic, construction, and seismic exploration activities can produce very high-amplitude, low-frequency sounds (Mulroy, 1991), but the effects of these activities on marine mammals are often difficult to predict because the existing data do not include directional considerations (Clark et al., 2009). The current study was not designed to measure masking, but the evidence presented here shows that the dolphin's receiving directivity is more acute for lower frequencies than previously thought, which suggests that dolphins might undergo some spatial release from masking, even for lower frequencies. Furthermore, spatial hearing is important for sound source localization, which is essential for animals to successfully avoid or approach sources that are biologically significant. The frequency-dependent spatial filtering observed here (see Fig. 2) likely leads to enhanced direction sensitivity, although perhaps not to the same degree that interaural level cues are available for high frequency for sonar signals.

The San Diego Bay environment in which this study was conducted featured ambient noise that could not be controlled and might have affected thresholds and variability in our dataset. Ambient noise spectral densities in San Diego Bay at 1 kHz were approximately 90 dB re 1 μPa²/Hz and decreased linearly with the logarithm of frequency up to about 40 kHz. Since ambient noise spectral densities in this environment generally increase with decreasing frequency, it is possible that bottlenose dolphins could exhibit more directional hearing sensitivity to 2 kHz sounds than found here; but noise would likely affect all angles equally. Relative threshold differences between sound projection angles were more important than absolute thresholds for this study, and behavioral hearing thresholds collected under excessively loud ambient background noise were eliminated from the analysis in order to minimize potential effects (see Sec. II). Furthermore, each averaged threshold consisted of threshold measurements that occurred on different days of the experiment, which decreased the likelihood that day-to-day variations in ambient noise influenced the averaged thresholds. Other limitations included that the two dolphin subjects had pre-existing hearing losses, and each individual subject was a substantial source of variability in the dataset. Hearing sensitivity in one subject (BLU) generally increased with increasing frequency, as might be expected from a normal dolphin's audiogram. However, TYH's data did not clearly exhibit this relationship, likely because of pre-existing hearing impairment. Additionally, BLU exhibited clear right–left asymmetry whereas TYH did not. The two subjects were comparable in mass but BLU was female and TYH was male, so sex-related differences are also possible. Dissimilarity between individual animals is common, but it complicates efforts to make general observations about hearing sensitivity.

It is widely accepted that the auditory system of bottlenose dolphins is specialized for directional reception of ultrasonic biosonar echoes, but growing concern about anthropogenic sound demands that more research be conducted on the full range of cetacean hearing to best predict impacts on marine mammals. This study showed that sounds lower than 30 kHz in frequency are received in a direction-dependent manner. Future research should specifically investigate (1) spatial release from masking at frequencies of 30 kHz and lower (2) whether anterior–posterior shadowing depends on depth or body size, and (3) directional sound reception for frequencies lower than 2 kHz.

The authors thank the National Marine Mammal Foundation dolphin training staff for their dedication during data collection, P. Moore and D. Houser for helpful discussions, and funding from the U.S. Navy awarded to A.K. Jenkins. We also thank two anonymous reviewers for comments that improved the manuscript. This is scientific contribution number 236 of the National Marine Mammal Foundation.