Introduction

Cochlear frequency tuning provides the foundation for tonotopic processing by the auditory system (Bourk et al., 1981; Schreiner and Langner 1997; Song et al., 2024). Otoacoustic emissions (Kemp, 1978) and behavioral testing (Moore, 1978) are two methods currently available to evaluate cochlear tuning in humans, and in other cases for which direct measurement is not feasible. Which methods accurately predict cochlear tuning remains controversial due to limited studies comparing otoacoustic, behavioral, and cochlear tuning in the same species.

Stimulus-frequency otoacoustic emissions (SFOAEs) are low-level acoustic signals recorded in response to tones that depend on active processes within the cochlea (Kemp, 1978; Shera and Guinan, 1999). SFOAE delay is expected to vary with cochlear tuning sharpness based on filter theory, with longer delays implicating sharper tuning. Previous studies found long SFOAE delays in humans, consistent with exceptional tuning sharpness (Shera et al., 2002). However, otoacoustic tuning estimation remains an indirect approach requiring validation. Behavioral experiments infer tuning from masking patterns, typically based on variation in tone thresholds across maskers differing in spectral content (Moore, 1978). However, behavioral tuning estimates depend on central processing of noisy signals in addition to cochlear tuning, and results vary across stimulus paradigms with active debate as to which methods better approximate actual tuning (Oxenham and Shera, 2003; Ruggero and Temchin, 2005; Leschke et al., 2022). Whether otoacoustic or behavioral methods provide closer estimates of actual cochlear frequency tuning remains unknown.

Several reports in nonhuman species have tested the extent to which SFOAE delays predict tuning of auditory-nerve responses at the output of the cochlea (Shera et al., 2010; Joris et al., 2011; Bergevin et al., 2015). These studies found that auditory-nerve tuning quality exceeds N_SFOAE (SFOAE delay in stimulus cycles) by a “tuning ratio” that decreases from ~3 to 1 from apex to base across several species (Shera et al., 2010). To our knowledge, only a single study has compared SFOAE and behavioral tuning estimates to auditory-nerve recordings in the same species, ferret (Sumner et al., 2018). SFOAE-based, behavioral, and auditory-nerve tuning quality in ferret all increased monotonically for higher frequencies, in broad accordance with one another (Sumner et al., 2018). A stronger test of SFOAE vs. behavioral tuning estimates could be obtained by studying species with specialized behavioral and/or physiological frequency-resolving abilities.

The budgerigar (Melopsittacus undulatus) is a parakeet species with well-known auditory processing abilities from behavioral studies (Dooling et al., 1989, 2000; Ryals et al., 2013; Henry et al., 2016, 2017; Wong et al., 2019; Henry and Abrams, 2024). Intriguingly, many studies show non-monotonic frequency dependence of tuning sharpness in budgerigars, with sharpest tuning at mid frequencies from 3.5-4 (Dooling et al., 2000; Fig. 1). Suggestive of an “auditory fovea” (Köppl et al., 1993), the pattern contrasts with the typical monotonic growth of tuning sharpness frequency in other species. In budgerigars, directly measured critical bands and psychophysical tuning curves based on multiple paradigms (Fig. 1) all indicate peak tuning sharpness from 3-4 kHz near the midpoint of this species’ audible frequency range (see also budgerigar critical ratios: Dooling and Saunders, 1975; Okanoya and Dooling, 1987; Farabaugh et al., 1988). Because auditory-nerve tuning has not been measured, it remains unknown whether budgerigars’ unusual behavioral tuning reflects specialized cochlear tuning, and if so, whether the pattern is detectable using SFOAEs.

Figure 1.

The present study compared the extent to which SFOAE and behavioral tuning measures predict actual cochlear tuning in budgerigars, a species that does not show typical monotonic growth of behavioral tuning sharpness with increasing frequency (Fig. 1). SFOAE-based tuning, estimated with low-level swept-tone stimuli, was compared to measures of auditory-nerve tuning and to previously reported behavioral measures of frequency tuning. In budgerigars, SFOAE and auditory-nerve tuning sharpness both increased monotonically with frequency, as in most other species, and in contrast to previously reported behavioral results. Thus, SFOAE delay predicted auditory-nerve tuning sharpness with greater accuracy than behavioral masking approaches.

Materials and methods

Animals

Budgerigars were sourced from a breeder or our institutional breeding program. SFOAEs were recorded from 22 ears in 14 budgerigars (7 female). Birds were 6 to 32 months old with a median age of 8.5 months at the time of SFOAE experiments. Auditory-nerve responses were recorded from 6 ears in 6 budgerigars (4 female). Birds were 6 to 22 months old at the time of the auditory-nerve recordings (median: 7.5 months). All experiments were approved by the University of Rochester Committee on Animal Resources.

SFOAE recordings

Animals were anesthetized for SFOAE recordings by subcutaneous injection of 0.08–0.12 mg/kg dexmedetomidine and 3–5.8 mg/kg ketamine. Following anesthesia, birds were placed in a stereotaxic apparatus located inside a double-walled acoustic isolation booth (Industrial Acoustics; 2 × 2.1 × 2.2 m). Temperature was maintained at 40°C throughout the experiment using a feedback-controlled heating pad (Harvard Apparatus Model 50-753, Edenbridge, KY, USA). Careful control of temperature was found to be critical for robust emission measurements. Breathing rate was monitored using a thermistor-based sensor.

After birds were placed in the stereotactic apparatus, a probe assembly consisting of a low-noise microphone (Etymotic ER10-B+, Elk Grove Village, IL, USA) and two earphones (Etymotic ER2, Elk Grove Village, IL, USA) was sealed to the ear using ointment. Acoustic stimuli were generated in MATLAB (The MathWorks, Natick, MA, USA; 50-kHz sampling frequency) and processed with a digital pre-emphasis filter (5000-point FIR) that corrected for the frequency response (i.e., magnitude and phase) of the system. The frequency response of the system was measured with the probe microphone at 249 linearly spaced frequencies (1-V peak amplitude) from 0.05-15.1 kHz. Stimuli were converted to analog at maximum scale (10-V peak amplitude; National Instruments PCIe-6251, Austin, TX, USA), attenuated to the required levels (Tucker Davis Technologies PA5, Alachua, FL, USA), and amplified using a headphone buffer (Tucker Davis Technologies HB7).

Stimuli for measuring SFOAEs were swept tones, as in Kalluri and Shera (2013). Two distinct acoustic signals were presented: a probe and a suppressor. The probe consisted of a tone linearly swept at a constant rate from 500 Hz to 8 kHz over 2.05 s, presented at 40 dB SPL. The suppressor consisted of a tone swept at a constant rate from 540 Hz to 8.04 kHz over 2.05 s, presented at 55 dB SPL. Stimuli were presented in repeating sequences in which the probe was presented first, then the suppressor, and finally the probe and the suppressor simultaneously. All stimuli had 25-ms raised-cosine onset and offset ramps. Each experiment consisted of forty repetitions of the stimulus sequence. There was a half-second silent time gap between presentations of each stimulus sequence. Response waveforms were amplified by 40 dB (Etymotic ER10B+) prior to sampling at 50-kHz and storage on a computer hard drive.

After the completion of the experiments, birds were given a subcutaneous injection of 0.5 mg/kg atipamezole and placed in a heated recovery chamber until fully alert.

Auditory-nerve recordings

The surgical procedures for neurophysiological recordings in budgerigars have been described previously (Wang et al., 2021). Equipment is described above in the section on SFOAE recordings, except where noted. Briefly, animals were anesthetized with a weight-dependent subcutaneous injection of ketamine and dexmedetomidine, as indicated above, and a head-post was mounted to the dorsal surface of the skull with dental cement to facilitate head positioning. Animals were placed in a stereotaxic device inside a sound-attenuating chamber, with the beak projected downward. The right ear was aligned with an earphone/microphone assembly and sealed with silicone grease. Body temperature was maintained at 40°C using a homeothermic control unit, and breathing rate was monitored throughout experiments. Animals were maintained in an areflexic state through continuous subcutaneous infusion of additional ketamine (3-10 mg/kg/h) and dexmedetomidine (0.08-0.2 mg/kg/h), along with physiological saline, using a syringe pump. DPOAEs were recorded periodically throughout experiments to test for possible deterioration of inner-ear function (Wong et al., 2019). Animals were euthanized at the conclusion of recordings, typically after 10-12 hours.

Access to the auditory nerve was accomplished by aspirating the right third of the cerebellum to reveal stereotaxic landmarks associated with the anterior semicircular canal. Recordings were made using pulled glass microelectrodes with 10-80 MΩ impedance. The tip of the electrode was positioned just anterior to a bony spur located on the medial surface of the ampulla of the anterior canal, and advanced into the brainstem using a hydraulic oil-filled remote microdrive (Drum drive; FHC, Bowdoin, ME, USA). Neurophysiological activity was amplified (variable gain; Grass P511 AC amplifier and high-impedance headstage), filtered from 0.03-3 kHz, and broadcast to the experimenters (outside the booth) over a monitor speaker. Action potentials were detected with a custom spike-discriminator circuit, consisting of a Schmitt trigger followed by a peak detector, and timed with National Instruments hardware. A wideband-noise search stimulus was used to isolate auditory-nerve fibers as the electrode was slowly advanced.

Recordings were excluded as likely originating from nucleus angularis based on bipolar action potential morphology, non-primary-like post-stimulus time histograms in response to tones (i.e., onset or chopping responses), or weak phase locking to tones. Tuning curves were determined through an automated algorithm that tracked the threshold sound level for rate excitation across frequencies. The algorithm estimated thresholds as the minimum tone level that repeatedly produced an increase in driven discharge rate compared to the most recent silent interval (Chintanpalli and Heinz, 2007). Tones were 50 ms in duration with 5-ms raised-cosine onset and offset ramps, presented once every 100 ms. Thresholds were sampled with a resolution of 28 frequencies per octave near characteristic frequency (CF), starting at 6-8 kHz and proceeding downward in frequency (typical range: 6 kHz to 200 Hz).

SFOAE processing

Analyses were performed in MATLAB R2023a and R version 4.3.2. Occasional microphone-signal artifacts were rejected by thresholding. Surrounding each signal transient artifact, 20 ms of signal was removed. Responses to the probe, suppressor, and combined probe and suppressor were averaged across repeated presentations. P_SFOAE was calculated by vector subtraction (Shera et al., 2002), as indicated in Eqn. 1, where P is the response waveform in Pa and subscripts denote the stimulus class.

The noise floor was estimated by the spectrum of the difference between odd and even trials of SFOAEs. Magnitude and phase spectra were extracted from the mean data for each component of the signal using the generalized least-squares fit described in Long and Talmadge (1997) and Long et al. (2008). Phase gradients were extracted after phase unwrapping and conversion to cycles.

Two exclusion criteria were implemented prior to phase-gradient extraction, for N_SFOAE estimation. Data points less than 10 dB in magnitude above the median of the estimated noise floor (Fig. 2a, gray shaded region) were rejected. Data points at the base of magnitude troughs were also rejected due to noted phase irregularities. Similar algorithms have previously been applied (e.g., Shera and Bergevin, 2012) to emphasize phase-gradient delays near magnitude maxima.

Figure 2.

Auditory-nerve response processing

Auditory-nerve tuning curves were first smoothed with a 5-point triangular window. CF was calculated as the frequency of the lowest threshold, Q₁₀ was calculated as CF divided by the tuning curve bandwidth 10 dB above the lowest threshold, and Q_ERB (i.e., the tuning quality of the idealized rectangular filter with equivalent bandwidth to the measured tuning curve) was calculated based on the area under the inverted tuning curve, as in Bergevin et al. (2015).

Statistical analysis

Linear mixed-effects models were implemented to determine the effect of frequency on SFOAE magnitude and phase-gradient responses. Magnitude and N_SFOAE values were grouped into five log-spaced frequency bands, with center frequencies ranging from 1 to 5 kHz. For each ear, first-degree robust polynomials were fit to the data in each frequency band. Magnitude and phase-gradient values were defined as the value of the polynomial fit at the center frequency in each band. Frequency was treated as a categorical variable. Ear intercepts were modeled as a random effect. The Satterthwaite approximation was used to calculate degrees of freedom for F tests. Statistical analyses were conducted with a significance level of α = 0.01.

Results

Budgerigar SFOAEs

SFOAEs were recorded from 22 ears in 14 animals using swept-frequency tones presented at 40 dB SPL. SFOAE level (Fig. 2a) typically ranged from 5-15 dB at stimulus frequencies below 5-6 kHz and descended into the noise floor at higher frequencies. Within the frequency of measurable emissions, SFOAEs in individual ears (Fig. 2a, colored lines) typically showed three or more deep spectral notches in emission level. Frequency regions near spectral notches, or for which SFOAE level was within 10-dB of the estimated noise floor, were excluded from subsequent analyses as described above (see II. METHODS. D. SFOAE processing). SFOAE phase decreased monotonically with increasing frequency and was relatively consistent across individual ears (Fig. 2b). Finally, SFOAE delay was estimated from the phase gradient of the response in stimulus cycles, i.e., as N_SFOAE (Fig. 2c). So calculated, N_SFOAE increased monotonically with increasing frequency, consistent with gradually increasing tuning quality along the length of the cochlea from apex to base. The local mean trendline for N_SFOAE (Fig. 2c, thick red line) was estimated using a Gaussian weighting with sigma of 0.25 octaves, fit to group delay (i.e., N_SFOAE × frequency) because this quantity was approximately normally distributed. 95% confidence intervals in Fig. 2c and throughout are based on 1,000 bootstrap repetitions.

Linear mixed-effect modeling revealed a significant effect of frequency on SFOAE level (F_4,84 = 18.91, p < 0.0001), associated with slightly lower SFOAE level in the higher frequency bands. Moreover, a linear mixed-model analysis of N_SFOAE revealed a significant effect of frequency (F_4,84 = 44.05, p < 0.0001) due to greater N_SFOAE in higher frequency bands. In summary, SFOAE results pooled across ears were consistent with monotonically increasing tuning sharpness from the cochlear apex to base.

Auditory-nerve tuning curves

Auditory-nerve tuning curves (n=127) in budgerigars had CFs ranging from 0.28-5.65 kHz (median: 2.97 kHz; interquartile range: 1.13-3.88 kHz) and associated thresholds at CF as low as 5 dB SPL (Fig. 3a, crosses). Individual tuning curves (Fig. 3a, colored lines) appeared symmetric around CF on a log-frequency axis, with no clear indication of a secondary tail region of low-frequency sensitivity for stimulus levels up to 80 dB SPL (i.e., the highest level tested). Q₁₀, the measure of tuning-curve sharpness, increased monotonically and significantly for higher CFs (Fig. 3b; t₁₂₅=6.542, p<0.0001; t-test of linear regression slope between log-transformed variables) as in other animal species, without an apparent peak at 3.5-4 kHz. The local mean trendline for Q₁₀ (Fig. 3b, thick black line) was calculated using a Gaussian weighting function with sigma of 0.5 octaves. The rate of increase for Q₁₀ with increasing frequency was similar to but slightly less than the rate observed for N_SFOAE (Fig. 3b; red line). Q_ERB also increased monotonically for higher CFs (see below) and exceeded Q₁₀ by a median factor of 1.76 (interquartile range: 1.65-1.89).

Figure 3.

Defining the tuning ratio, r, for SFOAE-based Q_ERB prediction

Previous studies (e.g., Shera et al., 2002, 2010) established the concept of a tuning ratio, r, assumed to vary slowly with frequency and fit to the empirically derived quotient of auditory-nerve Q_ERB and N_SFOAE, as indicated in Eqn. 2.

The value of r in mammals decreases from ~3 at the apical end of the cochlea to 1 for frequencies processed more basally (Shera et al., 2010). r appears largely conserved across related species, such that multiplying N_SFOAE by r from a different species or species group accurately predicts auditory-nerve Q_ERB in many cases (Shera et al., 2010; Joris et al., 2011; Sumner et al., 2018). Given substantial differences in cochlear morphology between birds and mammals (Gleich & Manley, 2000), we calculated r from chicken (Gallus gallus domesticus) to predict Q_ERB from N_SFOAE results in the budgerigar. Chicken data were selected rather than barn owl (Tyto alba; Bergevin et al., 2015) due to greater similarity of the chicken’s cochlear morphology (Gleich & Manley, 2000) and frequency limits of hearing (chicken: 9.1-7,200 Hz [Hill et al., 2014]; budgerigar: 77-7,600 Hz [Heffner et al., 2016]; frequency ranges for which behavioral audiometric thresholds are less than 60 dB SPL).

Auditory-nerve Q₁₀ from two previous chicken studies (Manley et al., 1991; Saunders et al., 1996) increases monotonically for higher frequencies (Fig. 4a). N_SFOAE in chicken also increases monotonically for higher frequencies (Bergevin et al., 2008; Fig. 4a, red line), similar to the pattern observed for auditory-nerve Q₁₀. Finally, the Q₁₀ trendline was fit to pooled data pooled across the two auditory-nerve studies and multiplied by 1.76 (approximating auditory-nerve Q_ERB; see above) and divided by N_SFOAE to yield chicken r (Fig. 4b). Chicken r decreased from a maximum value of ~7 at 500 Hz to 3-4 at frequencies from 1-3.35 kHz.

Figure 4.

Note that empirical evaluation of chicken r was limited to frequencies less than 3.35 kHz due to the scala tympani approach of the two auditory-nerve studies, which limited sampling from high-CF fibers (Manley et al., 1991; Saunders et al., 1996). We therefore extended r to 4.6 kHz by first extrapolating chicken Q₁₀ to 4.6 kHz. Extrapolation was performed with a linear regression model fit to Q₁₀ data from fibers with CFs greater than 1.7 kHz (Fig. 4a). Extrapolated Q₁₀ (predicted means ±1SE; Fig. 4a) trended slightly upward in frequency in accordance with the regression model slope. Extended-frequency r (Fig. 4b, magenta lines from 3-4.6 kHz), calculated as the quotient of extrapolated Q₁₀ over N_SFOAE, trended downward with increasing frequency from 3.35-4.6 kHz.

Finally, budgerigar r was calculated using the same empirical method, but without extrapolation due to broad sampling of CFs with our auditory-nerve approach. Budgerigar r was marginally lower than that of chickens and followed the same trajectory with increasing frequency, decreasing from a maximum of ~4 at 700 Hz to ~2 at 5.5 kHz (Fig. 4b, thick black line).

SFOAEs but not behavioral measurements accurately predict auditory-nerve frequency tuning in the budgerigar

SFOAE-based predictions of cochlear tuning in budgerigars were made by multiplying N_SFOAE by chicken r as a function of frequency (Fig. 5; magenta). Budgerigar SFOAE predictions, plotted together with auditory-nerve Q_ERB (thick black line) and average behavioral tuning sharpness in this species (thick blue line), highlight the similarities and differences across these measures. Behavioral Q₁₀ values from the literature (Fig. 1) were scaled by a factor of 1.76 to approximate Q_ERB. Whereas N_SFOAE predictions made using both the empirical and extended-frequency forms of chicken r reasonably approximated budgerigar auditory-nerve Q_ERB measurements, falling parallel to and slightly above the mean trendline, the behavioral results show a fundamentally different pattern characterized by a distinct peak from 3.5-4 kHz. Thus, SFOAE results provided closer estimates of actual auditory-nerve Q_ERB than did behaviorally estimated tuning measures in the budgerigar.

Figure 5.

Discussion

The present study measured SFOAEs in budgerigars to predict cochlear frequency tuning. SFOAEs were recorded from 22 budgerigar ears. Tuning curves were also recorded from budgerigar single auditory-nerve fibers for direct comparison. SFOAE level remained above the noise floor for frequencies below 6 kHz and trended downward with increasing frequency. SFOAE delay expressed as N_SFOAE increased steadily for higher stimulus frequencies, consistent with increasing tuning quality along the length of the cochlea in the 650 Hz - 6 kHz range. Moreover, quantitative SFOAE predictions calculated as the product of budgerigar N_SFOAE and chicken r showed close agreement with measured auditory-nerve Q_ERB across the frequency range available for comparison. In contrast, previous behavioral studies of budgerigar psychophysical tuning curves, critical bands, and iso-level masking function highlight a prominent peak in tuning sharpness from 3.5-4 kHz suggestive of an auditory fovea (Dooling and Saunders, 1975; Saunders et al., 1978, 1979; Saunders and Pallone, 1980; Kuhn and Saunders, 1980). Overall, the results show that SFOAE delays in budgerigars better predict auditory-nerve tuning sharpness compared to previously reported behavioral frequency tuning estimates based on masking.

SFOAE delay expressed as N_SFOAE increases for higher frequencies in a wide range of taxa, in qualitative agreement with patterns of auditory-nerve tuning quality in these species (macaque monkeys: Joris et al., 2011; chinchilla: Shera et al., 2010; gecko: Bergevin and Shera, 2010; cat and guinea pig: Shera et al., 2002; ferret: Sumner et al., 2018; chicken: Bergevin et al., 2008; barn owl: Bergevin et al., 2015). More detailed quantitative prediction of Q_ERB requires scaling N_SFOAE by r, where r is the empirically defined quotient of Q_ERB over N_SFOAE. In the budgerigar, use of r from chicken predicted auditory-nerve Q_ERB with good quantitative precision across the full range of frequencies evaluated (Fig. 5). The precision of SFOAE predictions is attributable to similarity of r between the budgerigar and chicken (Fig. 4b). The value of r in budgerigars decreased from ~4 at 700 Hz to ~2 at 5.5 kHz and was marginally higher in the chicken with similar frequency dependence. While the selection of chicken r for our study was reasonable, given similar frequency limits of hearing and cochlear anatomy between chicken and budgerigar (Gleich & Manley, 2000; Heffner et al., 2016; Hill et al., 2014), note that other species show lower r that, if used for SFOAE predictions in the budgerigar, would considerably underestimate auditory-nerve Q_ERB. The value of r in mammals decreases from ~3 to 1 from apex to base (Shera et al., 2010), whereas r in the barn owl ranges from 1-2 (Bergevin et al., 2015). The reasons for species differences in r are unknown, though level dependence of r in previous studies suggests that differences in hearing sensitivity could be a factor (Bergevin et al., 2015). Because r is such a key parameter in otoacoustic estimation of cochlear frequency tuning, further studies of its taxonomic variation, level dependence, and physiological basis are highly warranted (Bergevin et al., 2008; Manley 2022).

Measurements of chicken auditory-nerve tuning were drawn from two previously published studies and were limited to CFs less than 3.35 kHz due to their recording methodology (Manley et al., 1991; Saunders et al., 1996). CFs likely exist up to at least 5 kHz based on the chicken’s behavioral audiogram (Hill et al., 2014), but were not sampled due to the scala tympani approach employed by these studies (Manley et al., 1991; Saunders et al., 1996). Consequently, computing r across the full range of the chicken N_SFOAE function (0.5-4.6 kHz) required moderate extrapolation of the auditory-nerve Q₁₀ trendline from 3.4-4.6 kHz based on lower CF (1.7-3.35 kHz) data. While probably reasonable given continued increase of Q₁₀ above ~3 kHz in many other avian species (Sachs et al., 1974; Gleich & Manley, 2000), reliance in part on extrapolated chicken Q₁₀ remains a limitation of this study. It should also be noted that Q₁₀ data from a third chicken study (Salvi et al., 1992) were considered but ultimately not included in our estimation of chicken r because CFs were less than 2 kHz and the reported Q₁₀ values appeared somewhat discrepant (i.e., some reported values as high as 24).

The estimates of behavioral frequency selectivity referenced in the present study were originally obtained using three simultaneous masking paradigms: psychophysical tuning curves for probe-tone detection in tonal or narrowband maskers, critical bands for tone detection in bandlimited noise, and critical ratios for tone detection in wideband noise (Dooling and Saunders, 1975; Saunders et al., 1978, 1979; Saunders and Pallone, 1980; Kuhn and Saunders, 1980; Dooling et al., 2000). All three behavioral measures show that perceptual frequency selectivity in budgerigars is greatest from 3.5-4 kHz and declines markedly at higher and lower frequencies. Suggestive of a perceptual auditory fovea, this pattern differs substantially from behavioral findings in humans and other model species, which show monotonically increasing behavioral frequency selectivity for higher frequencies. Critically, the behavioral findings contrast with physiological tuning estimates of the current study, which show monotonically increasing SFOAE phase gradients with increasing frequency as well as monotonically increasing auditory-nerve Q₁₀/Q_ERB values for high CFs. Thus, neither SFOAEs nor auditory-nerve recordings suggest the existence of a peripheral auditory fovea that could explain prior behavioral findings in this species.

While the reason for disparate behavioral and SFOAE/auditory-nerve findings in budgerigars is unknown, the mismatch highlights the need for caution in inferring cochlear frequency selectivity using behavioral measures. Indeed, previous work in humans has shown that simultaneous masking paradigms can result in behavioral tuning functions that are qualitatively and quantitatively distinct from otoacoustic measures (Shera et al., 2002). Shera et al. (2002) and Sumner et al. (2018) demonstrated in humans and ferrets, respectively, that revised psychophysical paradigms designed to reduce the effects of nonlinear suppression and compression can yield behavioral results that more closely resemble otoacoustic measures. On the other hand, Ruggero and Temchin (2005) argued that forward-masking behavioral paradigms can substantially overestimate cochlear frequency selectivity. Moreover, psychophysical tuning curves obtained with forward masking in budgerigars (Kuhn and Saunders, 1980) have Q₁₀ values that are greater than twice simultaneous-masking values at the same frequencies (Fig. 1; red triangles), while also far exceeding the maximum auditory-nerve Q₁₀ values reported in Fig. 3. In summary, at least in budgerigars, it appears unlikely that any of the previously employed forward or simultaneous masking paradigms reasonably estimate cochlear frequency selectivity in this species.

A potential explanation for the disparity between behavioral and the other measures of frequency tuning lies in the nature of the measurements. Otoacoustic and auditory-nerve responses both reflect peripheral physiological processes originating in the cochlea. However, behavioral measures of frequency selectivity necessarily involve the entire auditory pathway and central nervous system, including neural mechanisms for processing masked signals, and are best explained by models that include both peripheral and central auditory system processes (Maxwell et al., 2020; Brennan et al., 2023). The unique pattern observed in the budgerigar’s behavioral Q₁₀ trendline may then reflect specializations for masked processing in the central auditory pathway of this highly vocal and gregarious species. While a previous study investigated neural tone-in-noise processing in the budgerigar’s inferior colliculus, the study did not explore the impact of noise bandwidth on neural detection thresholds (Wang et al., 2021; which would be needed to estimate neural critical bands). Note that the frequency selectivity of inferior colliculus response maps obtained with single tones of varying frequency and level showed no obvious peak in tuning quality for neural CFs from 3.5-4 kHz (Henry et al., 2016, 2017), similar to the present auditory-nerve results. Finally, several histological studies of budgerigar cochlear morphology also revealed no obvious specializations consistent with a peripheral auditory fovea (Manley et al., 1993; Wang et al., 2023)

In conclusion, SFOAE-based and auditory-nerve measures of tuning sharpness increased for higher frequencies in the budgerigars, and quantitative SFOAE predictions based on r from a different bird species (chicken) closely approximated actual budgerigar auditory-nerve tuning. The pattern is similar to findings in other avian and mammalian species, and contrasts with the unusual pattern of behavioral frequency tuning in budgerigars, which has maximum tuning quality from 3.5-4 kHz consistent with an auditory fovea. These results highlight the need for caution in interpreting behavioral measures of frequency tuning, which can show substantial deviation from cochlear tuning curves due to central processing mechanisms and other factors that necessarily influence behavioral detection of masked signals.

Acknowledgements

This research was supported by R01-DC017519 and R01-DC001641 from the National Institute on Deafness and Communication Disorders. Discussions with Natasha Mhatre are gratefully acknowledged.