Spatial acuity of the bottlenose dolphin (Tursiops truncatus) biosonar system with a bat and human comparison Free

Studies of toothed whale biosonar capabilities have historically focused on detection, discrimination, and classification capabilities (Au, 1993). Their biosonar systems rival and often outperform human engineered sonar in these capabilities, especially in shallow, turbid environments. Performance is often attributed to the hearing system of the animal and its ability to resolve time, frequency, and phase information (Branstetter et al., 2020; Finneran et al., 2002; Finneran et al., 2020, Moore et al., 1984). Far fewer studies have investigated how well these animals can perceive spatial information via biosonar. Observations of toothed whales using sonar to catch prey (Deruiter et al., 2009), navigate complex environments (Jensen et al., 2013), discriminate between closely spaced targets (Malinka et al., 2021; Wisniewska et al., 2012), and match objects based on shape (Pack et al., 2002; Pack and Herman, 1995) provide compelling evidence that their sonar system renders detailed spatial information about their environment. The need to resolve fine spatial details has been hypothesized as one of the primary evolutionary drivers for a narrow beam, high-frequency, short duration, broadband sonar signal (Branstetter and Mercado, 2006). This, in turn, drove the evolution of high-frequency, broadband hearing and fine temporal resolution, all properties unique to the hearing systems of echolocators (Branstetter and Mercado, 2006; Heffner and Heffner, 2008; Moore et al., 1984). Many toothed whales have pronounced cranial asymmetry (Fahlke et al., 2011), similar to what has been observed in other auditory predators (e.g., barn owl; Knudsen, 1981), which may have evolved to function as a position-dependent, spectral filter for sound localization in the vertical plane. Taken as a whole, the acoustic properties of the odontocete sonar signal, unique head morphonology, and specialized hearing capabilities provide compelling evidence that the need to perceive small spatial details was a primary evolutionary driver.

Spatial acuity can be defined as the ability to resolve two points in space. In biological sensory systems, spatial acuity has been measured in vision (Robson, 1966), audition (Renaud and Popper, 1975), somatosensation (Mancini et al., 2014), and olfaction (Jacobs, 2012). For sensory systems that perceive stimuli sources at range (i.e., vision, audition), spatial acuity is often measured as angular resolution. Like spatial acuity, angular resolution can be defined as the smallest angle between two objects where both objects are observed to be separate. In passive hearing, angular resolution is often called the minimum audible angle (MAA). Only a single study has measured the MAA in dolphins (Renaud and Popper, 1975). For tonal sounds, the MAA is between 2° and 4° depending on the frequency (Renaud and Popper, 1975). For click sounds similar to biosonar signals, the MAA is 0.9° and 0.7° in the horizontal and vertical planes, respectively (Renaud and Popper, 1975) [Fig. 1(D)]. In the horizontal plane, the dolphin presumably had access to interaural loudness and time differences to accomplish the task (Branstetter and Mercado, 2006; Moore et al., 1995). However, in the vertical plane, these cues are not available, suggesting the existence of well-developed head-related transfer function (HRTF).

There have been five studies that have attempted to measure horizontal or vertical angular difference discrimination in echolocating bats and dolphins, which is slightly different from angular resolution (Bel'kovich et al., 1970; Branstetter et al., 2003; Branstetter et al., 2007; Matrai et al., 2015; Simmons et al., 1983). In all of these studies, a comparison was made between two stimuli, each consisting of vertically or horizontally spaced rods (Fig. 1). For example, in Bel'kovich et al. (1970), the stimuli consisted of two arrays, each containing two vertical rods [Fig. 1(A)]. A common dolphin (Delphinus delphis) was required to interrogate the two arrays via echolocation and choose the array with the smaller spacing. The independent variable of the study was the angular difference (Δθ) between the two spatial separations,

where θ_a and θ_b were the two angular separations within each array [Fig. 1(A)]. The resulting angular difference threshold of 0.028° is suspect since the rods were positioned on a plane rather than an arc, resulting in potential time of arrival cues the dolphin likely used to perform the discrimination [Fig. 1(A)]. For example, the S− array, with the larger angular separation would result in a larger time separation between the echoes of each rod, compared to the S+ array. Dolphins are extremely sensitive to small time separations between echoes (Branstetter et al., 2020; Moore et al., 1984) and likely used this cue to achieve the extremely small discrimination threshold. Simmons et al. (1983) measured horizontal angular difference discrimination in the big brown bat (Eptesicus fuscus) in a procedure similar to Bel'kovich et al. (1970), except the time-of-arrival confound was precluded by placing the rods at an equal radial distance [Fig. 1(B)]. The resulting angular difference discrimination threshold was 1.5°, which is consistent with the bat's ability to track moving targets with 1.6° of accuracy (Masters et al., 1985). Branstetter et al. (2003) and Branstetter et al. (2007) measured horizontal angular difference discrimination in an echolocating bottlenose dolphin (Tursiops truncatus). In these experiments, two vertical arrays were presented, where the angular separation between the rods within each array was a constant, while the space between the arrays varied [Fig. 1(C)]. Branstetter et al. (2003) used a four vs two array configuration [Fig. 1(C)], while Branstetter et al. (2007) used a two vs one configuration. The resulting discrimination thresholds were between 0.7° and 1.6°, which were in good agreement with the dolphin MAA (Renaud and Popper, 1975) and angular discrimination in echolocating bats (Simmons et al., 1983). Matrai et al. (2015) replicated the study of Branstetter et al. (2003) and Branstetter et al. (2007) but also measured vertical angular discrimination. The dolphin's horizontal and vertical angular difference discrimination thresholds were both 1.0°. This result is consistent with the findings of Renaud and Popper (1975), where horizontal and vertical auditory spatial acuity are almost identical. If both horizontal and vertical auditory localization capabilities are similar or equal, the auditory mechanisms enabling this feat have not been well defined.

Although fine echoic angular discrimination is expected in auditory predators, such as bats and dolphins, human echolocators appear to perform equally well. In a study by Teng et al. (2012), human subjects echolocated on two discs, one above the other, and had to determine if the top disk was offset to the left or right of the bottom disk [Fig. 1(E)]. Three of the six subjects had discrimination thresholds between 1.2° and 1.9°, performance that rivals both bats and dolphins. This is a significant finding, since the spatial resolving properties of bat and dolphin biosonar are often attributed to their broadband, high-frequency hearing and their narrow echolocation beam (Branstetter and Mercado, 2006; Malinka et al., 2021). Humans possess neither (Teng, 2013).

In a related experiment, the acoustic behavior of free-swimming harbor porpoises was measured while they performed a two-alternative forced-choice (2AFC) task (Malinka et al., 2021). Two blindfolded porpoises were presented with two targets (S+ and S−) and were required to swim and station on the S+. The targets were separated by distances between 13.5 and 108 cm. Both animals performed well for all spatial separations. The data indicated that for smaller spatial separations, both echoes would arrive well within the animal's auditory temporal window, and therefore resolving the targets in the range dimension was unlikely. The hypothesized mechanism was that the narrow echolocation beam allowed the animals to individually ensonify each target, thus functioning as a spatial filter. This hypothesis may be called the projector spatial filter hypothesis since the proposed mechanism for the spatial filter is located on the projector end of the biosonar system. An alternative hypothesis can be called the receiver spatial filter hypothesis, where the filter mechanism is on the hearing end of the biosonar system. Of course, the two hypotheses are not mutually exclusive since both the projector and receiver can conceivably function as complimentary mechanisms for spatial filtering.

In the current study, two experiments were conducted: true angular resolution was measured in two dolphins, performing an echolocation target discrimination task with two targets. The primary independent variable was the angle between the two targets, and the dependent variable was proportion correct discrimination. Unlike the previous experiments that measured angular difference discrimination, the current study aims to measure true angular resolution between two targets. In the second experiment, the projector spatial filter hypothesis was tested by replicating Malinka et al. (2021), but including spatial separations small enough to degrade performance, measure a psychometric function, and concurrently place both targets well within the dolphin's echolocation beam (i.e., there will be little to no measurable acoustic difference between received biosonar signals at each target). The projector spatial filter hypothesis predicts performance should degrade as acoustic differences between the targets diminish and approach chance levels when there are no acoustic differences. Thus, the hypothesis can be rejected if a dolphin is able to discriminate between two targets when both are well within the animal's biosonar beam.

Two bottlenose dolphins participated in this study. The dolphins SHA (female) and WHP (male) were 41 and 17 years old, respectively, and had previous experience in biosonar detection and discrimination tasks. Both SHA and WHP had good broadband hearing, measured via evoked potential audiometry (Strahan et al., 2020) with upper-frequency cutoffs of 128 and 113 kHz, respectively (Fig. 2). Experiments were conducted in a 9 m × 9 m floating, netted enclosure in San Diego Bay, at Naval Base Point Loma. The study was approved by the Institutional Animal Care and Use Committee of the Naval Information Warfare Center (NIWC) Pacific and the Navy Bureau of Medicine and Surgery. The study followed all applicable United States Department of Defense guidelines for the care of laboratory animals.

Two dolphins were trained to perform a 2AFC task (Branstetter et al., 2003). At the beginning of each experimental session, silicone rubber suction cups (eye cups) were placed on the dolphin's eyes to prevent the use of vision. Once the eye cups were on the dolphin, a training assistant would prepare the S+ and S− targets for the next trial (see Sec. II B 1 for more details). Upon a tactile command from the trainer, the dolphin would swim and position itself within the stationing hoop with its pec fins flush against the hoop (Fig. 3). Approximately 0.5 m directly in front of the dolphin was a 165 cm × 165 cm netted gate. The depth (approximately 80 cm) and horizontal placement of the stationing hoop were designed to place the dolphin at the center of the netted gate. Beyond the netted gate was the open water of San Diego Bay. The netted gate allowed sound to travel freely through the pen. When the dolphin was properly positioned, each target would be at a 4 m radial distance from the dolphin's blowhole. At a verbal command from the trainer, a research assistant would lower the stimuli in the water, and the dolphin would interrogate the stimuli via echolocation, back out of the hoop, and touch a left or right paddle with its rostrum to indicate the location of the S+ target (left or right). Response paddles were positioned 1 m to the left and right of the stationing hoop and were constructed of polyvinyl chloride (PVC) with a small float on the lower end to provide an echoically reflective target for the dolphin's sonar sense. The trainer, whose back was turned to the stimulus apparatus and could not observe the targets, served as a blind observer and would verbally report the dolphin's response. A research assistant would then inform the trainer of a correct or incorrect paddle press. Correct responses were rewarded with fish reinforcement. Incorrect responses were not rewarded. After the dolphin's response, the targets were removed, and two research assistants would reposition the targets on the wooden rack according to a printed schedule (see below). This procedure was repeated until the end of the session. Each session began with 10 warmup trials, followed by 24 test trials, followed by five cooldown trials. For the warmup and cooldown trials, the angular separation was always 10°. Performance on warmup trials served as an indicator of the animal's willingness or ability to participate in that session. If the dolphin missed more than two warmup trials (<80% correct) the session was aborted. A new session could be attempted after a minimum of a 1-h break. Cooldown trials served to increase motivation for future sessions by ending the session with “easy” high-probability trials that ended in fish reinforcement. Cooldown trials also served as an indicator that the dolphin was still motivated and did not switch strategies (e.g., developing an extreme left or right bias) during the session. If the dolphin missed more than one cooldown trial (<80% correct) or three cumulative warmup and cooldown trials (<80% correct), data from that session were excluded from analysis. A modified method of constant stimuli with a pseudo-randomization procedure was used to select the angular separation for each trial and assign the locations of the S+ and S− targets to the left or right position. The randomization algorithm ensured that the S+ target never appeared in more than three consecutive trials on the same side and that each angular separation had an equal number of right and left S+ placements within a session.

One target, the S−, was made from a hollow steel tube (Vigoro, Atlanta, GA) with a plastic coating [length (L) = 150 cm, outer diameter (OD) = 1 cm, wall thickness (WT) ∼ 0.5 mm]. The S− target was water filled. Another target, the S+, was made of the same hollow steel tube, except the tube was air filled and had a copper tip (length 20 cm, OD = 1.2 cm). Initial discrimination training proved difficult for both dolphins, so additional solid cylindrical aluminum rod bars (L = 13 cm, OD = 0.65 cm) were attached to the front of the tube facing the dolphin and were spaced every 5 cm. The addition of the small rods presumably provided greater difference in the returned echoes while keeping the horizontal spatial profile the same (1 cm). The ends of each aluminum rod were cut at a 45° angle to provide a hydrodynamic, bubble-free entry through the water surface. The targets were clamped to a custom wooden arc. The wooden arc was rectangular, except the side where the targets were connected, where an arc was shaped from a circle with a 4 m radius. Markings on the arc allowed the targets to be accurately placed at any required angular separation. Additional plastic spacers were fabricated at each of the angular separations. The spacers could be placed between the targets to ensure they were at the correct angular separation along the lengths of both targets. The spacers were removed before each trial. Once the targets were attached to the wooden arc, PVC cylinders allowed the arc to be placed on the wooden rack, where the tips of the targets were submerged a few centimeters below the surface. Upon a command from the trainer, the targets would be lowered with a smooth, “guillotine-like” motion, for the dolphin to interrogate. The tips of both targets were at the same depth, which was equal to that of the dolphin. The rack and wooden arc were above the water's surface, and only the targets were submerged.

Psychometric functions were fit to proportion correct vs angular separation data, using the quickpsy package (Linares and López-Moliner, 2016) in R (R Development Core Team, 2012). The general form of the psychometric function can be expressed as

where $ψx;θ$ was the probability of a correct response when $x$ was presented, θ = (x;α,β,γ,λ) were model parameters, and F was a sigmoidal logistic function in the form

Model parameters α and β are position and scale parameters, and x is the independent variable, which in this case was the angular separation between the S+ and S− stimuli. Because the dolphins performed a 2AFC procedure, the “guess rate,” or lower asymptote γ, was fixed at 0.5 (i.e., chance performance was 50% correct). The lapse rate λ represents the upper asymptote (1 − λ). The lapse rate is considered to originate from stimulus-independent reporting errors (e.g., the animal knew the correct answer but hit the wrong paddle due to memory or attention issues). Model parameters were estimated using maximum likelihood. Threshold values corresponded to the angular separation associated with the 0.75 proportion correct point on the psychometric function, and 95% confidence intervals (CIs) for the thresholds were bootstrapped using 200 iterations (Linares and López-Moliner, 2016).

One dolphin (WHP) participated in a free-swimming, target discrimination task similar to Malinka et al. (2021). The targets were the same ones used in the angular resolution measurements, except the dolphin's task was to swim and touch the S+ with his rostrum [Fig. 3(C)]. At the start of a trial, eye cups were placed on the dolphin. The targets were prepared in the same manner as in experiment I; however, they were lowered into the water before the start of each trial. Upon a command from the trainer, the dolphin would swim the length of a 9-m enclosure, swim through an open gate, and then swim the length of another 9-m enclosure to where the targets were located [Fig. 3(C)]. The dolphin would then attempt to select the S+ target by touching it with his rostrum. Correct responses were informed with a whistle bridge and rewarded with fish. The same warmup and cooldown trial series was used for this study. The same spatial separations from experiment I were used in this experiment, but the separations are reported in cm rather than degrees due to the variable distance between the targets and the moving dolphin. For a subset of trials, a hydrophone (Reson TC4013, Slangerup, Denmark) was attached at the very bottom of each target. Each hydrophone was coupled to a Reson VP1000 preamplifier (gain = 20 dB, high-pass filter cutoff = 1 kHz), and the signals were digitized with a National Instruments USB-6251 (Austin, TX) data acquisition device at 500 kS/s. When the dolphin was visually observed to swim through the gate that separated the pens, 10 s of acoustic data acquisition was manually triggered by a human operator. Each trial resulted in a two-channel (one for the S+ and one for the S− hydrophones), 10-s recording. A video camera (GoPro Hero 3 Black Edition, Carlsbad, CA) was placed just beneath the surface, above the targets, to verify the animal's response. An additional GoPro was attached via suction cup to the dolphin's head, behind his blow hole, for the same purpose.

Recorded sound files were analyzed using matlab 2007 R (Mathworks, Natick, MA). Clicks were extracted from each trial using a custom click detector algorithm. Peak amplitudes (for both channels) were calculated within a 350-point sliding window. If the peak amplitude exceeded a threshold of 159 dB re 1 μPa, the time location of the peak was determined, and a waveform was extracted by taking 30 samples before and 200 samples after the peak (230 total samples, 460-μs duration). The 159 dB re 1 μPa threshold was chosen to exclude ambient noise (i.e., snapping shrimp) while including most of the dolphins biosonar signals. Extracted clicks were plotted and visually inspected to ensure the algorithm functioned properly. For each click, the peak-to-peak sound pressure level (SPLp-p) and peak frequency from a 128-point fast Fourier transform were calculated. To compare received signals at the two targets, differences in acoustic level and frequency were computed. The mean level difference (ΔL) between the S+ and S− targets can be defined as

where T_i and D_i were the SPLp-p at the S+ and S− hydrophones, respectively, for the ith click in a click train with N total clicks. Positive values for ΔL indicate higher sound pressure at the S+ target, while negative values indicate higher sound pressure at the S− target. ΔL can serve as a proxy to which target the dolphin attended to most. To quantify differences in SPLp-p magnitude between the targets, regardless of sign, the mean absolute differences in SPLp-p were also computed,

where L is the mean absolute level difference between the S+ and S− hydrophones for a click train, and |x| is the absolute value of x. Larger values of L indicate the two targets are acoustically separable. Differences between the peak frequencies of S+ and S− clicks were also calculated,

where $Tî$ and $Dî$ were the frequencies corresponding to maximum value from the fast Fourier transform of T_i and D_i, respectively (i.e., peak frequencies). Higher values of F indicate larger spectral differences between peak frequencies for the S+ and S− hydrophones.

SHA and WHP completed a total of 36 and 54 trials per angular separation, respectively. Psychometric functions were fit to the proportion correct data for the last 36 trials for both dolphins. Proportion correct as a function of angular separation (Table I) and the resulting psychometric functions are displayed in Fig. 4 for both dolphins. Discrimination thresholds at the 0.75 proportion correct level for SHA and WHP were 3.9° and 1.5°, respectively. Figure 4(C) displays both dolphins' proportion correct thresholds with 95% CIs. Individual differences between the two animals were likely due to different levels of expertise in the task rather than differences in perceived spatial acuity, reflected by the much smaller 95% CI with WHP. Because of the smaller CI, data from WHP were used for comparison with previous studies. Data from Branstetter et al. (2003) and Renaud and Popper (1975) were fit with psychometric functions using the same procedure from the current study [Fig. 4(D)]. Psychometric functions from the three studies are remarkably similar in both slope of the curve and resulting thresholds, despite three different methods with three different dolphins. The new threshold fits for the current study, Branstetter et al. (2003), and Renaud and Popper (1975) were 1.5°, 1.6°, and 1.1°, respectively.

Only the dolphin WHP participated in this study. Spatial separation between the targets is reported in centimeters rather than degrees, since the angle between the targets varied with the dolphin's range to target. Table II displays each spatial separation and resulting total correct trials, total trials, and proportion correct. WHP's 0.75 proportion correct discrimination threshold was 6.69 cm [Fig. 5(A)]. WHP's performance data and a comparison with the results from the two harbor porpoises (Freja and Sif) from Malinka et al. (2021) are displayed in Fig. 5. WHP's data are similar to the data for the harbor porpoises, except more conditions were completed for smaller spatial separations, allowing the data to extend from ceiling to chance level performance.

Acoustic data were collected during a single session. Because the procedure for collecting acoustic data was slightly different (i.e., there was a single hydrophone attached to each target), those data were not included in the psychometric function fits. However, performance with the acoustic trials was almost identical (Table III). Examples of recorded click trains are displayed in Fig. 6. The hypothesis that the dolphin attends more to the S+ target than the S− target was tested by comparing the relative level difference between each corresponding click at the S+ and S− targets. If the dolphin attended more to the S+ target, there should be a greater average SPL at the S+ stimulus [ΔL from Eq. (1)]. This difference should also increase as the target separation increases. Target separation was a significant predictor of ΔL (F_1, 4445 = 1725, p < 0.001) and is plotted in Fig. 7(A). To determine if the targets were acoustically separable, the absolute differences in SPLp-p between the S+ and S− hydrophones were compared [Eq. (2)]. Target separation was a significant predictor of absolute level differences (L) between the two targets (F_1, 4445 = 3422, p < 0.001) and differences in the peak frequencies (F) of clicks (F_1, 4445 = 239, p < 0.001). Figures 7(B) and 7(C) display noticeable differences in SPL and peak frequencies between the targets, as the spatial separation increased. Figure 7(D) compares the dolphin's proportion correct performance with the dB difference between the targets. Performance decreases as the absolute level difference between the two targets decreases.

There was a performance gap between the two dolphins that was unlikely related to individual differences in sensory capabilities. Both dolphins have good broadband hearing with SHA having a slightly higher high-frequency cutoff (Fig. 2). SHA's large lapse rate [Fig. 4(A)] and large CI [Fig. 4(C)] suggest that she had not mastered the discrimination task in the allotted time. WHP, however, displayed relatively superior performance with a discrimination threshold of 1.5°. This is likely the true measure of bottlenose dolphin biosonar angular resolution. Measures of the MAA (Renaud and Popper, 1975) and angular difference discrimination (Branstetter et al., 2003; Branstetter et al., 2007) have psychometric functions remarkably similar to the current study, suggesting the four experiments were measuring the same sensory capability [Fig. 4(D)]. Interestingly, the MAA (Renaud and Popper, 1975) and angular difference discrimination (Matrai et al., 2015) in the vertical plane are as good as in the horizontal plane.

The horizontal angular resolution psychometric functions for the bottlenose dolphin, big brown bat, and human are remarkably similar in shape, slope, and threshold (Fig. 8). One hypothesis is that a common mammalian mechanism governs auditory spatial acuity. The alternative is that all three species independently converged on similar capabilities. The current theory on the origins of toothed whale and bat biosonar systems is that both have undergone considerable evolutionary adaptations in head and ear morphology, signal production, and auditory signal processing to support biosonar capabilities (Alcuri, 1980; Jones and Teeling, 2006; Ketten, 1992; Moss and Sinha, 2003). However, blind humans can apparently learn to echolocate out of necessity, without specialized adaptations, with spatial resolution, at least in the horizontal plane, as good as effectively as bats and dolphins (Teng et al., 2012). This suggests the biosonar-related adaptations found in toothed whales and bats are for capabilities other than horizontal angular resolution.

The hypothetical location for the biosonar spatial filter includes the acoustic projector (Au et al., 1986; Malinka et al., 2021), the acoustic receiver (Branstetter and Mercado, 2006; Moore et al., 1995), or a combination of the two. The acoustic projector hypothesis predicts that (1) acoustic differences between two point source targets are required for targets to be resolvable; (2) those differences are achieved by a sonar signal with a narrow beam width that will differentially ensonify the point source targets, and angular resolution will be a direct function of the beam width of the sonar signal; and (3) passive and active sound localization rely on different mechanisms and can result in performance differences. Experiment II of the current study tested the first prediction and failed to reject the projector spatial filter hypothesis. Indeed, there appears to be a correlation between the dolphin's performance and the SPL differences between the targets [Fig. 7(D)]. Of course, this is not necessarily a causal relationship because SPL differences between the targets was not a predictor variable, but an outcome variable, and a covariate with performance. The second and third predictions can be tested post hoc by reviewing results from previous experiments. First, bats, dolphins, and humans have emission beams that vary considerably from one another. The bottlenose dolphin produces broadband clicks, with typical peak frequencies around 120 kHz and durations around 40 μs (Au, 1980), with a 3-dB beam width of approximately 10° (Au et al., 1986; Branstetter et al., 2012; Finneran et al., 2014). The big brown bat produces descending, frequency-modulated chirps that are milliseconds in duration with multiple harmonics (Jones and Holderied, 2007; Shimozawa et al., 1974; Simmons et al., 1975) with a dynamic 3-dB beam width of approximately 45° (Ghose and Moss, 2003; Hartley and Suthers, 1989; Simmons, 1969). Human echolocation signals have not been characterized with the same level of detail as dolphin and bat signals. However, human biosonar emissions vary considerably, having durations typically between 3 and 20 ms with peak frequencies typically below 5 kHz (Teng, 2013; Thaler et al., 2011; Thaler et al., 2017). These three species share remarkably similar angular resolution psychometric functions, with similar slopes and thresholds (Fig. 8), despite having radically different emission beam properties. These results are inconsistent with the second prediction of the projector spatial filter hypothesis. The third prediction from the projector spatial filter hypothesis predicts spatial resolution will likely be different for passive and active localization tasks. However, the dolphin's psychometric functions for passive and active listening are almost identical [Fig. 4(D)], providing further evidence against the projector spatial filter hypothesis.

What may be most astonishing is that general sound localization capabilities between dolphins, bats, and humans are remarkably similar despite large differences in head size, auditory morphology and function, and sound speed in each animal's respective acoustic medium. For passive listening, the horizontal MAA for broadband stimuli near the mid-sagittal plane is 1.0° for humans and 1.1° for dolphins [new fit using data from Renaud and Popper (1975)], respectively (Mills, 1958; Perrott and Saberi, 1990; Renaud and Popper, 1975). Bats can track targets with about 1.6° of accuracy (Masters et al., 1985). The angular resolution of bats, dolphins, and humans during echolocation are all approximately 1.5°. All of these experiments from both passive and active listening tasks with three different species reveal psychometric functions that are remarkably similar in shape and threshold, suggesting a common mammalian mechanism or common limitation in how auditory spatial information is processed. Then why have echolocating animals evolved broadband signals and narrow emission beams if not for spatial resolution? The flashlight analogy may shed light on the answer. For example, an omnidirectional light will illuminate a space in all directions. If a parabolic reflector is placed near the light source, a beam is created increasing light intensity and illumination in one direction. However, visual acuity is not defined by the beam width of light, but by the refractive properties of the eye's cornea and lens, as well as the cone density at the fovea. Likewise, if the biosonar acoustic emission is directed forward by the animal's head structures, the SPL of the emission will increase in the forward direction, increasing target detection range, reducing peripheral clutter, and informing the listener of the general direction of the target (i.e., the direction the listener is pointing its head). However, the angular resolution of the biosonar system is not tethered to the beam width of the sonar emission but is the product of the binaural auditory system and its ability to compute loudness, time, and phase differences between the ears as well as analyze spectral differences caused by the animal's HRTF. The high frequencies (small wavelengths) allow echo returns from small targets (e.g., insects for bats, and fish swim bladders for dolphins), and broadband signals are easier to localize because interaural loudness differences can be compared across multiple auditory filters. But the auditory system is responsible for localizing these point sources and segregating them from other point sources in close spatial proximity and appears to perform this function equally well via passive or active listening. The facts that passive and active angular resolution are almost identical and that widely varying acoustic beams of bats, dolphins, and humans result in almost identical angular resolution measurement support a model where auditory angular resolution resides on the receiving end of biosonar systems.

Several lines of evidence suggest dolphins form images through biosonar (Harley et al., 2003; Pack et al., 2002; Pack and Herman, 1995). An image can be defined as any representation (e.g., visual, acoustic, haptic) that preserves the spatial structure of an object. By defining the spatial acuity of the dolphin biosonar system, insight can be gained into the clarity and detail of the dolphin's internal representation though echolocation. In the horizontal plane, the angular resolution is approximately 1.5°. In the vertical plane, angular resolution appears to be equivalent but requires validation. Attempts to measure range resolution have been made using range discrimination (Murchison, 1980) and jittered-echo delay discrimination paradigms (Finneran et al., 2020). However, true range resolution, defined as the ability to resolve two points on the range axis, has yet to be measured with two physical targets on the same line of acoustic propagation. A jittered-echo delay threshold of 1.3 μs (Finneran et al., 2020) suggests that range resolution is likely superior to horizontal and vertical resolution.

The bottlenose dolphin's horizontal angular resolution during echolocation is approximately 1.5°. The value is consistent with previous measurements of spatial acuity with both passive and active hearing dolphins. The value is also consistent with spatial acuity measurement with both bat and human echolocators. The location of the hypothetical spatial filter that governs spatial acuity during biosonar is likely seated in the listener's auditory system, but the current study cannot rule out the contribution of the emission beam pattern in echoic spatial acuity.

This project could not have been completed without the aid of the animal care staff and interns of the National Marine Mammal Foundation and United States Navy Marine Mammal Program. Darren Schreher helped design and fabricate the experimental apparatus used to present the targets. We thank Randall Dear, Dorian Houser, and Jason Mulsow for logistical support and helpful comments related to dolphin training. Many past discussions with Louis Herman, Adam Pack, and Whitlow Au helped develop the conceptual framework for this study. This work was supported by the Office of Naval Research. This is scientific contribution #331 from the National Marine Mammal Foundation.