Otoacoustic emissions but not behavioral measurements predict cochlear-nerve frequency tuning in an avian vocal-communication specialist

Reviewer #1 (Public review):

Summary:

In their manuscript, the authors provide compelling evidence that stimulus-frequency otoacoustic emission (SFOAE) phase-gradient delays predict the sharpness (quality factors) of auditory-nerve-fiber (ANF) frequency tuning curves in budgerigars. In contrast with mammals, neither SFOAE- nor ANF-based measures of cochlear tuning match the frequency dependence of behavioral tuning in this species of parakeet. Although the reason for the discrepant behavioral results (taken from previous studies) remains unexplained, the present data provide significant and important support for the utility of otoacoustic estimates of cochlear tuning, a methodology previously explored only in mammals.

Strengths:

* The OAE and ANF data appear solid and believable. (The behavioral data are taken from previous studies.)

* No other study in birds (and only a single previous study in mammals) has combined behavioral, auditory-nerve, and otoacoustic estimates of cochlear tuning in a single species.

* SFOAE-based estimates of cochlear tuning now avoid possible circularity and were are obtained by assuming that the tuning ratio estimated in chicken applies also to the budgerigar.

Weaknesses:

* In mammals, accurate prediction of neural Q_ERB from otoacoustic N_SFOAE involves the application of species-invariance of the tuning ratio combined with an attempt to compensate for possible species differences in the location of the so-called apical-basal transition (for a review, see Shera & Charaziak, Cochlear frequency tuning and otoacoustic emissions. Cold Spring Harb Perspect Med 2019; 9:pii a033498. doi: 10.1101/cshperspect.a033498; in particular, the text near Eq. 2 and the value of CFa|b).

Despite this history, the manuscript makes no mention of the apical-basal transition, its possible role in birds, or why it was ignored in the present analysis. As but one result, the comparative discussion of the tuning ratio (paragraph beginning on lines 383) is incomplete and potentially misleading. Although the paragraph highlights differences in the tuning ratio across groups, perhaps these differences simply reflect differences in the value of CFa|b. For example, if the cochlea of the budgerigar is assumed to be entirely "apical" in character (so that CFa|b is around 7-8 kHz), then the budgerigar tuning ratios appear to align remarkably well with those previously obtained in mammals (see Shera et al 2010, Fig 9).

* For the most part, the authors take previous behavioral results in budgerigar at face value, attributing the discrepant behavioral results to hypothesized "central specializations for the processing of masked signals". But before going down this easy road, the manuscript would be stronger if the authors discussed potential issues that might affect the reliability of the previous behavioral literature. For example, the ANF data show that thresholds rise rapidly above about 5 kHz. Might the apparent broadening of the behavioral filters arise as
a consequence of off-frequency listening due to the need to increase signal levels at these frequencies? Or perhaps there are other issues. Inquiring readers would appreciate an informed discussion.

Reviewer #2 (Public review):

Summary:

This manuscript describes two new sets of data involving budgerigar hearing: 1) auditory-nerve tuning curves (ANTCs), which are considered the 'gold standard' measure of cochlear tuning, and 2) stimulus-frequency otoacoustic emissions (SFOAEs), which are a more indirect measure (requiring some assumptions and transformations to infer cochlear tuning) but which are non-invasive, making them easier to obtain and suitable for use in all species, including humans. By using a tuning ratio (relating ANTC bandwidths and SFOAE delay) derived from another bird species (chicken), the authors show that the tuning estimates from the two methods are in reasonable agreement with each other over the range of hearing tested (280 Hz to 5.65 kHz for the ANTCs), and both show a slow monotonic increase in cochlear tuning quality over that range, as expected. These new results are then compared with (much) older existing behavioral estimates of frequency selectivity in the same species.

Strengths:

This topic is of interest, because there are some indications from the older behavioral literature that budgerigars have a region of best tuning, which the current authors refer to as an 'acoustic fovea', at around 4 kHz, but that beyond 5 kHz the tuning degrades. Earlier work has speculated that the source could be cochlear or higher (e.g., Okanoya and Dooling, 1987). The current study appears to rule out a cochlear source to this phenomenon.

Weaknesses:

The conclusions are rendered questionable by two major problems.

The first problem is that the study does not provide new behavioral data, but instead relies on decades-old estimates that used techniques dating back to the 1970s, which have been found to be flawed in various ways. The behavioral techniques that have been developed more recently in the human psychophysical literature have avoided these well-documented confounds, such as nonlinear suppression effects (e.g., Houtgast, https://doi.org/10.1121/1.1913048; Shannon, https://doi.org/10.1121/1.381007; Moore, https://doi.org/10.1121/1.381752), perceptual confusion between pure-tone maskers and targets (e.g., Neff, https://doi.org/10.1121/1.393678), beats and distortion products produced by interactions between simultaneous maskers and targets (e.g., Patterson, https://doi.org/10.1121/1.380914), unjustified assumptions and empirical difficulties associated with critical band and critical ratio measures (Patterson, https://doi.org/10.1121/1.380914), and 'off-frequency listening' phenomena (O'Loughlin and Moore, https://doi.org/10.1121/1.385691). More recent studies, tailored to mimic to the extent possible the techniques used in ANTCs, have provided reasonably accurate estimates of cochlear tuning, as measured with ANTCs and SFOAEs (Shera et al., 2003, 2010; Sumner et al., 2010). No such measures yet exist in budgerigars, and this study does not provide any. So the study fails to provide valid behavioral data to support the claims made.

The second, and more critical, problem can be observed by considering the frequencies at which the old behavioral data indicate a worsening of tuning. From the summary shown in the present Fig. 2, the conclusion that behavioral frequency selectivity worsens again at higher frequencies is based on four data points, all with probe frequencies between 5 and 6 kHz. Comparing this frequency range with the absolute thresholds shown in Fig. 3 (as well as from older budgerigar data) shows it to be on the steep upper edge of the hearing range. Thus, we are dealing not so much with a fovea as the point where hearing starts to end. The point that anomalous tuning measures are found at the edge of hearing in the budgerigar has been made before: Saunders et al. (1978) state in the last sentence of their paper that "the size of the CB rapidly increases above 4.0 kHz and this may be related to the fact that the behavioral audibility curve, above 4.0 kHz, loses sensitivity at the rate of 55 dB per octave."

Hearing abilities are hard to measure accurately on the upper frequency edge of the hearing range, in humans as well as in other species. The few attempts to measure human frequency selectivity at that upper edge have resulted in quite messy data and unclear conclusions (e.g., Buus et al., 1986, https://doi.org/10.1007/978-1-4613-2247-4_37). Indeed, the only study to my knowledge to have systematically tested human frequency selectivity in the extended high frequency range (> 12 kHz) seems to suggest a substantial broadening, relative to the earlier estimates at lower frequencies, by as much as a factor of 2 in some individuals (Yasin and Plack, 2005; https://doi.org/10.1121/1.2035594) - in other words by a similar amount as suggested by the budgerigar data. The possible divergence of different measures at the extreme end of hearing could be due to any number of factors that are hard to control and calibrate, given the steep rate of threshold change, leading to uncontrolled off-frequency listening potential, the higher sound levels needed to exceed threshold, as well as contributions from middle-ear filtering. As a side note, in the original ANTC data presented in this study, there are actually very few tuning curves at or above 5 kHz, which are the ones critical to the argument being forwarded here. To my eye, all the estimates above 5 kHz in Fig. 3 fall below the trend line, potentially also in line with poorer selectivity going along with poorer sensitivity as hearing disappears beyond 6 kHz.

The basic question posed in the current study title and abstract seems a little convoluted (why would you expect a behavioral measure to reflect cochlear mechanics more accurately than a cochlear-based emissions measure?). A more intuitive (and likely more interesting) way of framing the question would be "What is the neural/mechanical source of a behaviorally observed acoustic fovea?" Unfortunately, this question does not lend itself to being answered in the budgerigar, as that 'fovea' turns out to be just the turning point at the end of the hearing range. There is probably a reason why no other study has referred to this as an acoustic fovea in the budgerigar.

Overall, a safe interpretation of the data is that hearing starts to change (and becomes harder to measure) at the very upper frequency edge, and not just in budgerigars. Thus, it is difficult to draw any clear conclusions from the current work, other than that the relations between ANTC and SFOAEs estimates of tuning are consistent in budgerigar, as they are in most (all?) other species that have been tested so far.

Author response:

We genuinely appreciate the reviewer critiques of our submitted paper, “Otoacoustic emissions but not behavioral measurements predict cochlear-nerve frequency tuning in an avian vocal-communication specialist.” We are planning a number of changes based on the reviewers’ helpful comments that we feel will substantially improve the manuscript and clarify its implications.

We will add more support for the claim that budgerigars show unusual patterns of behavioral frequency tuning compared to other species. The original manuscript relied on previously published studies of budgerigar critical bands and psychophysical tuning curve to make this point (e.g., Fig. 1). Critical bands and psychophysical tuning curves have unfortunately not been studied in many bird species. Consequently, it was somewhat unclear (based on the information originally presented) whether the “unusual” behavioral tuning results shown in Fig. 1 reflect a hearing specialization in budgerigars or perhaps simply a general avian pattern attributable to declining audibility above 3-4 kHz (a point raised by both reviewers). Fortunately, behavioral critical-ratio results are available from a broader range of species. Albeit a less direct correlate of tuning, the results clearly highlight the unique hearing abilities of budgerigars in relation to other bird species as elaborated upon below.

The critical ratio is the threshold signal-to-noise ratio for tone detection in wideband noise and partly depends on peripheral tuning bandwidth. Critical ratios have been studied in over a dozen bird species, the vast majority of which show similar thresholds to one another and monotonically increasing critical ratios for higher frequencies (by 2-3 dB/octave, similar to most mammals; reviewed by Dooling et al., 2000). By contrast, budgerigar critical ratios diverge markedly from other species at mid-to-high frequencies, with ~8 dB lower (more sensitive) thresholds from 3-4 kHz (Dooling & Saunders, 1975; Okanoya & Dooling, 1987; Farabaugh 1988; see Figs 5 & 6 in Okanoya & Dooling, 1987). The unusual critical-ratio function in budgerigars is not attributable to the audiogram and was hypothesized by Okanoya and Dooling (1987) to reflect specialized cochlear tuning or perhaps central processing mechanisms. A brief discussion of these studies will be added to the introduction, along with a new figure panel (for Fig. 1) illustrating these intriguing species differences in critical ratios.

Another question was raised as to whether the simultaneous-masking paradigms and classic methods used to estimate behavioral tuning in budgerigars should be considered as valid, given newer forward-masking and notched-noise alternatives. We will expand the discussion of this issue in the revised manuscript. First, many of the methods from the classic budgerigar studies remain widely used in animal behavioral research (e.g., critical bands and ratios: Yost & Shofner, 2009; King et al., 2015; simultaneous masking: Burton et al., 2018). We therefore believe that it remains highly relevant to test and report whether these methods can accurately predict cochlear tuning. While forward-masking behavioral results are hypothesized to more accurately predict cochlear tuning humans (Shera et al., 2002; Joris et al., 2011; Sumner et al., 2018), evidence from nonhumans is controversial, with one study showing a closer match of forward-masking results to auditory-nerve tuning (ferret: Sumner et al., 2018), but several others showing a close match for simultaneous masking results (e.g., guinea pig, chinchilla, macaque; reviewed by Ruggero & Temchin, 2005; see Joris et al., 2011 for macaque auditory-nerve tuning). Moreover, forward- and simultaneous-masking results can often be equated with a simple scaling factor (e.g., Sumner et al., 2018). Given no real consensus on an optimal behavioral method, and seemingly limited potential for the “wrong” method to fundamentally transform the shape of the behavioral tuning quality function, it seems reasonable to accept previously published behavioral tuning estimates as essentially valid while also discussing limitations and remaining open to alternative interpretations.

We will add clarification throughout the revision as to the specific behavioral measures used to quantify tuning in budgerigars (i.e., critical bands, psychophysical tuning curve, and critical ratios). This avoids potentially disparaging alternative behavioral methods that have not been tested. That the budgerigar behavioral data are “old” seems not particularly relevant considering that the methods are still used in animal behavioral research as noted previously. Rather, it seems important to clarify the specific behavioral techniques used to estimate budgerigar’s frequency tuning in the revised paper.

Finally, we plan to add discussion of the apical-basal transition from the mammalian otoacoustic-emission literature, as suggested by reviewer 1, including how this concept might apply in budgerigars and other birds.

References not already cited in the preprint:

Burton JA, Dylla ME, Ramachandran R. Frequency selectivity in macaque monkeys measured using a notched-noise method. Hear Res. 2018 Jan;357:73-80. doi: 10.1016/j.heares.2017.11.012.

King J, Insanally M, Jin M, Martins AR, D'amour JA, Froemke RC. Rodent auditory perception: Critical band limitations and plasticity. Neuroscience. 2015 Jun 18;296:55-65. doi: 10.1016/j.neuroscience.2015.03.053.

Yost WA, Shofner WP. Critical bands and critical ratios in animal psychoacoustics: an example using chinchilla data. J Acoust Soc Am. 2009 Jan;125(1):315-23. doi: 10.1121/1.3037232. PMID: 19173418; PMCID: PMC2719489.