Killer whales (Orcinus orca) rely heavily on their auditory sense for critical life functions such as communication and echolocation. They live in highly social groups that inhabit marine regions across the globe (Baird, 2000; Hoelzel et al., 2002). As the largest delphinid odontocete, killer whales are the best hearing surrogate for larger toothed whales such as sperm whales (Physeter macrocephalus) and beaked whales (family Ziphiidae), which are more difficult to study. Killer whales also have more sensitive hearing at lower frequencies than other model odontocete species [e.g., bottlenose dolphins (Tursiops truncatus)] in correlation with their larger functional interaural distance (Branstetter et al., 2017; Heffner and Heffner, 2008, 2018). Sources of low frequency anthropogenic noise (e.g., boat engines) are common in many of the regions that killer whales inhabit (Holt, 2008; NOAA, Fisheries, 2022).

The broad geographic distribution of killer whales, along with their sensitive hearing, heightens the risk of negative impacts from exposure to anthropogenic noise (Erbe et al., 2018). Auditory masking, hearing loss, and behavioral reactions could occur in response to anthropogenic noise (Nowacek et al., 2007; Weilgart, 2007; Williams et al., 2014b; Erbe et al., 2016), which may result in loss of foraging opportunities (Holt et al., 2011) and diminished communication space (Holt et al., 2009; Houghton et al., 2015). Audiograms, which plot a listener’s hearing sensitivity as a function of frequency, have helped establish noise safety criteria for underwater anthropogenic sound (Southall et al., 2019). However, these standardized audiograms have limited applicability as they only measure the lowest audible sound pressure level (SPL) of the listeners over their frequency range of hearing with no established relationship to perceived loudness. Loudness growth patterns change with sound frequency and other characteristics, including contextual factors (Suzuki and Takeshima, 2004; Finneran and Schlundt, 2011; Wensveen et al., 2014; Mulsow et al., 2015). Measuring subjective loudness is ideal for developing predictions of the likelihood that anthropogenic noise will result in behavioral responses (Finneran and Schlundt, 2011; Suzuki and Takeshima, 2004).

Changes in loudness perception with certain sound characteristics such as frequency, onset, and duration are poorly understood in non-human animals—in part because they are more difficult to study than humans. Equal loudness can be measured using subjective loudness judgments (i.e., the louder of two choices is indicated by the listener, or one sound is adjusted to match the loudness of the other sound) to determine the SPL of tones that are perceived to be equally as loud as a function of the target sound. This is easily tested in human participants but is challenging in odontocetes (Finneran and Schlundt, 2011; Pfingst et al., 1975a,b). Reaction time (RT), the time between the onset of a stimulus and the initiation of a response, has been used as a simple alternative to approximate perceived loudness in humans (Humes and Ahlstrom, 1984; Wagner et al., 2004) and marine mammals such as the harbor porpoise (Phocoena phocoena) (Wensveen et al., 2014), bottlenose dolphin (Mulsow et al., 2015), harbor seal (Phoca vitulina) (Kastelein et al., 2011), and California sea lion (Zalophus californianus) (Mulsow et al., 2015). The results of those studies suggest that, in many cases, reaction time is a suitable proxy for direct loudness measurements.

Reaction time data as a function of sound frequency can be used to generate equal-latency contours, which are functions that describe the sound levels at which reaction times are equal across the frequencies tested. Equal-latency contours from the bottlenose dolphin, harbor porpoise, harbor seal, and California sea lion resemble the shape of the audiogram at near-threshold sound pressure levels (Mulsow et al., 2015; Wensveen et al., 2014; Kastelein et al., 2011). This supports the use of the audiogram in predicting loudness perception for near threshold SPLs in marine mammals.

In the present study, conditioned killer whale vocal responses to sounds of different durations, sensation levels and frequencies were measured, and reaction time was evaluated as a potential indicator of loudness perception. Vocal reaction time was expected to decrease as sensation level increased, and resulting equal latency contours at lower sensation levels were expected to resemble the audiogram (Stebbins, 1966; Green, 1975; Pfingst et al., 1975a,b; Moody et al., 1980; Ridgway and Carder, 2000; May et al., 2009; Mulsow et al., 2015; Wensveen et al., 2014; Kastelein et al., 2011).

The data for the present study was collected opportunistically from Branstetter et al. (2023) as a part of a psychoacoustic study where the objective was to measure behavioral hearing thresholds as a function of signal duration. This section includes both a summary of the methods from Branstetter et al. (2023) (Sections 2.1–2.3) and the data analysis new to the present study (Section 2.4). Below, the terms response latency and reaction time have been used interchangeably to refer to the same concept.

A total of 225 (145 for Whale C, n = 80 for Whale E) hearing test sessions were analyzed with a total of 1,348 (n = 838 for Whale C and n = 510 for Whale E) individual response latency measurements. Because of the study design and methods, it was not possible to obtain equal sample sizes for all signal conditions (i.e., frequency, SPL and duration combinations), sample sizes were small in some cases (see Table 1), and the data were not normally distributed. Therefore, a non-parametric Mann–Whitney U test was used to assess statistical significance of the effect of the individual whale on overall response latency to determine whether data from both subjects could be combined. Whale C had a median reaction time (RT) of 521 ms, which was significantly faster (z = 5.892, p < 0.001) than Whale E which had a median response latency of 658 ms, so analysis of the effect of sensation level, frequency, and duration was conducted separately for each whale. False alarm rates for all sessions ranged from zero to 33%.

Different RT trends were observed with increasing frequency for both whales. Median RT for Whale C decreased with increasing frequency while median RT for Whale E increased (see Figure 3). A significant relationship between median RT and frequency was observed for Whale C (F (1, 4) = 7.748, p = 0.0462) but not for Whale E (F (1, 4) = 9.081, p = 0.0581).

Here, we measured latencies of conditioned vocal responses in killer whales for the first time. The median RTs at each frequency measured for both whales in the present study ranged from 491 to 705 ms (see Figure 3 in results). In general, other types of non-vocal motor response latencies, such as from a switch release, are more commonly used in equal-latency studies in humans (Wagner et al., 2004; Marshall and Brandt, 1980; Ulrich et al., 1998; Pfingst et al., 1975a,b) and other mammals (Kastelein et al., 2010; May et al., 2009; Green, 1975; Wensveen et al., 2014). Table 3 illustrates the variability of RT measurements among other mammals compared to the present study. The relatively longer RTs in killer whales (see Table 3) are not surprising because of their large size. While nerve conduction occurs at the same maximum velocity, the sensorimotor pathway distance increases with body size (More et al., 2010; More and Donelan, 2018). For the odontocete species in Table 3, RT measurements increase with body size: the harbor porpoise showed a range of non-vocal RTs from 110 to 770 ms (Wensveen et al., 2014), the bottlenose dolphin whistle RTs ranged from 230 to 845 ms (Mulsow et al., 2015), and the killer whale raspberry RTs ranged from 306 to 1,561 ms.

The vocal response itself, a trained “raspberry,” likely influenced the reaction times measured in the present study. This is because vocal response evocation—particularly for marine mammals—is probably delayed by the time required to generate sufficient air pressure (Ridgway, 2011; Mulsow et al., 2015). The lowest RT measurement from the present study (306 ms) was longer than the lowest vocal RT measurements in bottlenose dolphins (230 ms) (Mulsow and Finneran, 2013; Mulsow et al., 2015). This could be due to the considerable size difference between the two species, or the type of vocal response (e.g., whistles). Although the type of response measured in the present study is different from previous work in odontocetes, the observed decrease in median RT with increasing sensation level in the present study (see Figures 4, 5) is consistent with the previous research in bottlenose dolphins (Mulsow et al., 2015; Finneran and Schlundt, 2011; Mulsow and Finneran, 2013) and the harbor porpoise (Wensveen et al., 2014).

The individual session data from the present study demonstrated that an increase in RT occurred as SPL decreased within a hearing test session (see Figure 6). Reaction time for both vocal responses and non-vocal motor responses has been shown to decrease with increasing SPL until it plateaus (see Table 3). The present study supports these findings at most of the frequencies measured with both whales. The gradient of the median RT curves depicted in Figures 4, 5 exhibited a noticeable reduction in steepness within the 10–20 dB range for several of the measured frequencies. A similar trend was observed in bottlenose dolphins (Mulsow et al., 2015). The RT curves in the present study were not smooth at all frequencies. For example, Whale C had a higher median RT than expected at 20 dB SL for 10 kHz and 45 dB SL for 80 kHz. Whale E had a lower median RT than expected at 15 dB SL for 40 kHz. The unequal sample sizes among the SLs at each of the frequencies for both whales likely contributed to this variation in median RT.

The two whale subjects’ RTs differed, with Whale C having substantially lower RTs than Whale E (see Table 1), on the order of approximately 100 ms. Ridgway (2011) observed similar magnitude (100–150 ms) between-subject differences in bottlenose dolphin vocal RTs, with the youngest subject having the lowest RTs. It is unlikely that age, body size or other physical differences between the two whales were the source of this individual variation, as the whales were similar in size and age (see Section 2.1). For both subjects, response latencies for long-duration signals (> 500 ms) often resulted in the vocal response overlapping with the test signal. In other words, neither whale waited for the tone to end before responding. One possible explanation for the observed subject-specific differences in RT was Whale C’s additional experience with behavioral hearing tests compared to Whale E. While both whales participated in tone detection (Branstetter et al., 2017) and auditory masking studies (Branstetter et al., 2021, 2024), Whale C began his initial training several months before Whale E and Whale C completed data collection for a temporal integration study where Whale E did not (Branstetter et al., 2023). Interestingly, the two subjects showed diverging trends in RT as a function of sound frequency (see Figure 3). Whale E showed an increase in RT with increasing frequency, and Whale C showed a decrease in RT with increasing frequency for all sensation levels. Data from Branstetter et al. (2023) does not demonstrate hearing loss at 80 kHz for Whale E (see Supplementary Table S2), suggesting that other factors may influence the difference between the two subjects’ relationship between sound frequency and RT.

The RT dataset had a low sample size at many of the short duration signals (< 20 ms) making it difficult to differentiate an effect on RT from signal duration (see Supplementary Figures S1, S2; Figure 2). Previous studies have analyzed the relationship between RT and signal duration in humans. Ulrich et al. (1998) observed a decrease in RT with increasing signal duration to a specific point before RT plateaued, which matches the relationship between hearing threshold and signal duration in Branstetter et al. (2023). Therefore, the effect of signal duration on RT was likely captured by using SL in the present study.

The equal-latency contours in the present study collapsed closer together at higher frequencies, indicating that sensation level affected response latency less than at lower frequencies (also see Figures 4, 5). Compression of the equal latency contours at the upper and lower frequencies has been observed in previous studies (Mulsow et al., 2015; Wensveen et al., 2014; Kastelein et al., 2011; May et al., 2009; Marshall and Brandt, 1980). It was surprising that compression was not observed at lower frequencies (1 kHz) where the hearing thresholds were highest for Whale C, but sample size was lowest for this frequency (see Table 1), which may have influenced this result. However, killer Whale communication calls often contain peak energy at lower frequencies (1–3.5 kHz) (Williams et al., 2014a), and this biological relevance could affect loudness perception. The contours were expected to flatten out at higher sensation levels, as higher intensity signals have been shown to reduce the effect of frequency on equal loudness and equal latency contours (Kohfeld et al., 1981). This phenomenon is likely dependent on neural rise time rates at the onset of the signal, which reach a maximum at higher intensities (Kohfeld et al., 1981). While the 25 and 30 dB sensation level contours did exhibit some flattening across the lower frequencies, the higher frequencies showed lower equal latency SLs than expected based on previous research.

At 80 and 100 kHz, the results of the equal latency contours for Whale C (see Figure 7) do not conform to the hypothesized audiogram-like shape. The high-frequency portion of the equal latency contours was compressed at lower sensation levels compared to frequencies within the best hearing range (5–20 kHz) for Whale C (see Figure 2). Equal latency contours from previous studies with marine mammals have shown the contours to follow similar trends as the behavioral audiogram, with lower levels required to elicit the same RTs at more sensitive hearing frequencies compared to frequencies outside the best range of hearing (Mulsow et al., 2015; Finneran and Schlundt, 2011). Overall, the equal latency contour data in the present study do not consistently support that lower SLs are required to elicit equal RTs in the best hearing range, or that equal-latency contours follow similar trends as the audiogram. The gradual onset of the raspberry vocal response and methods used in the present study likely introduced variability that influenced the equal latency contours. Further research on response latency in odontocetes would benefit from using a vocal response with a rapid onset (e.g., sonar clicks or burst pulse) or a non-vocal motor response—such that automated detection methods such as pressure or light sensors—could reduce response onset uncertainty.

Overall, the relationship between response latency and sensation level observed in the present study supports that RT can be used as a proxy for perceived loudness growth in killer whales. These first measurements of killer whale vocal response latencies provide insight into the relationship between auditory sensitivity and loudness perception in large odontocetes. Further study of loudness perception is important for understanding the vulnerability of these animals to acoustic disturbance, and for informing conservation efforts.