Sex differences in distortion-product and transient-evoked otoacoustic emissions compared

Abstract

Although several studies have documented the existence of sex differences in spontaneous otoacoustic emissions (SOAEs) and transient-evoked OAEs (TEOAEs) in humans, less has been published about sex differences in distortion-product OAEs (DPOAEs). Estimates of sex and ear differences were extracted from a data set of OAE measurements previously collected for other purposes. In accord with past findings, the sex differences for TEOAEs were substantial for both narrowband and wideband measures. By contrast, the sex differences for DPOAEs were about half the size of those for TEOAEs. In this sample, the ear differences were small for TEOAEs in both sexes and absent for DPOAEs. One implication is that the cochlear mechanisms underlying DPOAEs appear to be less susceptible to whatever influences are responsible for producing sex differences in TEOAEs and SOAEs in humans. We discuss the possibility that differences in the effective level of the stimuli may contribute to these outcomes.

INTRODUCTION

In humans, otoacoustic emissions (OAEs) exhibit marked sex differences. Spontaneous OAEs (SOAEs) are more numerous and stronger in females than in males, and click-evoked OAEs (CEOAEs) are stronger in females than in males (e.g., Bilger et al., 1990; Talmadge et al., 1993; McFadden, 1993b, 1998; McFadden and Loehlin, 1995; McFadden et al., 1996; McFadden and Pasanen, 1998, 1999; McFadden and Shubel, 2003). There appears to be an ear difference overlaid on the sex difference. The number of SOAEs and the strength of CEOAEs are greater in right ears than in left ears (e.g., Bilger et al., 1990; Talmadge et al., 1993; McFadden et al., 1996; McFadden and Pasanen, 1998, 1999; Khalfa et al., 2001), although the magnitudes of these ear differences are smaller than those of the sex differences. Because these sex and ear differences exist in newborns as well as in adults (Strickland et al., 1985; Burns et al., 1992, 1994; Morlet et al., 1995, 1996; Thornton et al., 2003; Berninger, 2007), and because OAEs appear to be reasonably stable through life as long as there is no damage to the cochlea; e.g., Harris et al., 1991; Franklin et al., 1992; Burns et al., 1993, 1994; Engdahl et al., 1994; Marshall and Heller, 1996; McFadden et al., 1996), the implication is that the sex and ear differences are produced by mechanisms operating during prenatal development. Thus, the degree of exposure to androgens is a likely candidate mechanism for these differences (especially the sex differences, see McFadden, 2002, 2008) simply because degree of prenatal exposure to androgens is known to be responsible for so many other sex differences in body, brain, and behavior (e.g., Nelson, 2005).

The evidence available suggests that the sex and ear differences in distortion-product OAEs (DPOAEs) may be smaller than those for SOAEs and CEOAEs (Bonfils et al., 1988; Gaskill and Brown, 1990; Lonsbury-Martin et al., 1991; Moulin et al., 1993; Cacace et al., 1996; Dhar et al., 1998; Bowman et al., 2000; O’Rourke et al., 2002; Dunckley and Dreisbach, 2004; Keefe et al., 2008). If confirmed, these dissimilarities would have implications for theories about the mechanisms underlying OAEs. Shera and Guinan (1999, 2003) have argued that one of the cochlear mechanisms underlying DPOAEs is fundamentally different from that underlying SOAEs and CEOAEs. Namely, SOAEs and CEOAEs are thought to be primarily the result of a linear, reflection-based mechanism, whereas DPOAEs are thought to be the result of both the linear reflection mechanism operating from a location on the low-frequency side of the primary tones, plus a nonlinear cochlear mechanism operating near the location of the higher primary tone. An absence of, or a diminution in, sex differences in DPOAEs relative to SOAEs and CEOAEs would suggest that the linear, reflection-based mechanism is subject to modulation by some agent(s) such as androgens, but that the nonlinear mechanism is not (or is less) subject to such modulation.

In an attempt to increase the available information about the relative sizes of the sex and ear differences in DPOAEs and other OAEs, we re-examined a set of data acquired by Martin and his colleagues for other purposes. A small subset of those data had been reported previously [Lonsbury-Martin et al., 1991, Fig. 5(b)], but that report did not include all of the subjects eventually tested. Because this data set contained both DPOAEs and TEOAEs obtained on the same subjects, we were able to compare the individual differences in those measures as well as to estimate sex and ear differences.

METHODS

Each subject provided informed consent under a human-research protocol approved by the University of Miami School of Medicine’s Institutional Review Board. Each subject also received monetary compensation for participation in the study.

Subjects

Because one of the interests of the original study was the effects of aging on OAEs, the initial investigators (Lonsbury-Martin et al., 1991) expended considerable effort to obtain subjects with normal hearing (<20 dB hearing level or HL) over the standard set of audiometric test frequencies. However, by necessity, some of the 50 year olds eventually included in their data set did have hearing worse than this target (<30 dB HL), but only at one frequency in one or both ears. Both SOAEs and DPOAEs have been shown to decline in strength with increasing age in a similarly select sample (Lonsbury-Martin et al., 1991; Whitehead et al., 1995), suggesting that even though hearing sensitivity remains nominally normal, there apparently can be some age-related degradation of the cochlear mechanisms responsible for OAEs. (Note that age-related declines in both SOAEs and DPOAEs suggest that both the linear and nonlinear cochlear mechanisms are being affected—see Shera and Guinan, 1999, 2003.) Because here we wished to know about the relative sizes of the sex differences in OAEs prior to age-related declines, we restricted our subject pool to people aged 15–35. A total of 51 females and 57 males satisfied the age criteria. All subjects had normal tympanograms, normal acoustic-reflex thresholds at 1.0 kHz both contralaterally and ipsilaterally, hearing sensitivity of 20 dB HL or better in both ears at the octave frequencies between 0.25 and 8.0 kHz, and no history of hearing disorders.

Whitehead et al. (1993) demonstrated that SOAEs are more numerous and stronger in Blacks and Asians than in Whites, but little is known about ethnic or racial effects on the sex difference in DPOAEs. Unfortunately, the question could not be examined here because the number (N) of Blacks and Asians was so small. To be cautious, we excluded the 10 females and the 11 males who identified themselves as Black or Asian, leaving in the data pool 41 females and 46 males who identified themselves either as White or Hispanic (Whitehead et al. (1993) excluded Hispanics as well from their study). Also, subjects having OAEs 3.0 or more standard deviations from the mean of their group were excluded from the data analyses for that specific condition, and when ear differences were being assessed, only subjects having acceptable data for both ears were included. Thus, the Ns for individual conditions and comparisons varied but are shown in the figures and tables.

General

The general data-collection procedures were described in Lonsbury-Martin et al., 1991 and Whitehead et al., 1995. Two characteristics of the stimuli used to produce DPOAEs were atypical of current practice. The two primary tones were equal in level, and that level (75 dB sound-pressure level or SPL) was higher than typically used in recent years. The possible effects of stimulus level are discussed in Sec. 4. The primary tones used to collect DPOAEs ranged between 0.8 and 8.0 kHz in steps of approximately 0.1 octave. Each frequency step was characterized using the geometric mean of the primary tones. The strength of the DPOAE was measured during presentations of the primary tones that lasted approximately 90 ms. Also, a wideband TEOAE was obtained using the nonlinear procedure and the default settings provided by the Otodynamics ILO88 device. The default click value produced click stimuli averaging about 81 dB peak-equivalent SPL (peSPL) in the ear canals. That wideband response then was filtered to obtain narrowband TEOAE responses of about 0.1 octave in width.1

To simplify the analyses, we targeted four frequency regions for analysis: 1.5, 2.0, 3.0, and 4.0 kHz. In order to obtain measures as stable as possible, a mean level of the distortion product was calculated across three conditions for each target frequency for each subject. Those three conditions were the ones having primary tones whose geometric means were closest to each of the four target frequencies (e.g., 1.409, 1.516, and 1.623 kHz for the nominal value of 1.5 kHz; and 3.729, 4.003, and 4.287 kHz for the nominal value of 4.0 kHz). A parallel procedure was used for the TEOAE data. Specifically, means were obtained for each of the four target frequency regions by averaging the measures extracted with the filter set to the three frequencies closest to the four target values. Prior to the calculation of these three-frequency means, all obtained measures that exceeded the mean for an individual frequency condition by 3.0 standard deviations or more were deleted.

Note that the TEOAE frequency regions chosen for examination corresponded to the frequencies of the primary tones used to collect the DPOAEs, not to the frequencies of the 2f₁-f₂ distortion products themselves. The reason for this choice was that the DPOAE data were collected with relatively intense primary levels (75 dB each), meaning that the predominant source of the 2f₁-f₂ component was the one residing near the f₂ location along the cochlear partition (Shera and Guinan, 1999, 2003; Kalluri and Shera, 2001).

In the majority of cases, data were available for both ears of a subject, but in some cases the data for one ear were excluded. Because only subjects having data for both ears were included in the ear-difference analyses, the Ns available for calculating ear differences (a within-subject difference) were smaller than those available for calculating sex differences (a group difference).

The sex and ear differences will be expressed in terms of effect size, which was calculated here as the difference between the means of the two groups divided by the square root of the weighted mean of their variances. Thus, effect size is a measure like d^′ (Green and Swets, 1966), which is commonly used in psychoacoustical studies. Cohen (1992) suggested that effect sizes of 0.2, 0.5, and 0.8 could be considered small, medium, and large, respectively. A measure like effect size is preferable to the outcomes of statistical tests when a large number of comparisons are to be made. Here, a two-tailed t-test was calculated whenever an effect size greater than 0.4 was obtained.

RESULTS

Both the sex differences and ear differences were smaller for DPOAEs than for TEOAEs. The basic data are shown in Figs. 12, and Table 1 shows the effect sizes for the various sex and ear differences. The figures and the effect sizes shown for the sex differences were based on the data for all possible subjects; the effect sizes shown for the ear differences were calculated using only those subjects having data for both ears.

Figure 1.

Level of the DPOAE produced by equal-level primary tones of 75 dB SPL located in the frequency regions indicated. For each designated frequency region, the results for three pairs of primary tones were averaged; see text. Only data from Caucasian and Hispanic subjects included. Flags designate the standard errors of the mean.

Figure 2.

Strength of the TEOAE responses evoked by a click of about 81 dB SPL. The averaged TEOAE waveform was analyzed in several narrowbands as well as wideband. For the narrowband conditions, the levels shown are for 0.1-octave bands centered approximately on the frequency values shown.

Table 1.

Effect sizes for the sex and ear differences shown in Figs. 12. Positive effect sizes correspond to the mean OAE being stronger in females than males, or stronger in right ears than left ears; Ns for sex differences are shown in Figs. 12; Ns for ear differences=28∕22 for female∕male, except=28∕21 at 4000 Hz.

Condition	Effect sizes for
	Sex difference		Ear difference
	Left ear	Right ear	Female	Male
Distortion-product OAEs
1500 Hz	0.326	0.218	0.173	0.190
2000 Hz	0.274	0.296	0.000	0.007
3000 Hz	−0.064	0.117	0.060	−0.041
4000 Hz	0.514a	0.271	−0.009	0.204
Mean=	0.263	0.226	0.056	0.090
Transient-evoked OAEs (nonlinear procedure)
1500 Hz	0.492a	0.668b	0.354	0.121
2000 Hz	0.432	0.405	0.358	0.186
3000 Hz	0.535a	0.402	0.175	0.277
4000 Hz	0.741b	0.804b	−0.013	−0.115
Mean=	0.550	0.570	0.218	0.117
Wideband	0.621a	0.756a	0.413	0.087

^aTwo-tailed t-test: p<0.05.

^bTwo-tailed t-test: p<0.01.

As can be seen by comparing the top and bottom halves of Table 1, the effect sizes for sex difference in DPOAEs averaged about 0.25 while those for TEOAEs averaged about 0.56. For DPOAEs, only one of the ten sex differences examined achieved statistical significance, while for TEOAEs seven of ten did. The most significant sex-difference comparison was for the TEOAE in the right ear at 4000 Hz (p=0.0017) and when that value was corrected for the 11 t-tests calculated (Darlington, 1990, pp. 249–251), p=0.019. The bottommost entries in Table 1 are for the wideband TEOAE data.

Few other published studies provide the information necessary to calculate effect sizes for comparison, but McFadden and Pasanen (1999) and McFadden and Shubel (2003) reported effect sizes of 0.76 and 0.84, respectively, for the sex difference in a wideband TEOAE (not using a nonlinear procedure like that used here; see footnote ¹). Also, Thornton et al. (2003) obtained effect sizes of only about 0.20 for the sex difference in the overall TEOAEs obtained from neonates using the ILO88 system, but the sex differences were substantially larger at high frequencies than at low (their Fig. 5), meaning that calculating the overall TEOAE underestimates the sex differences at higher frequencies. Unfortunately for current purposes, neither McFadden and Pasanen (1999) nor Thornton et al. (2003) obtained measures of DPOAEs from their subjects for comparison with the TEOAEs.

Dunckley and Dreisbach (2004) measured DPOAEs over a wide frequency range, and they did provide the information necessary to calculate effect sizes. Unfortunately for comparisons here, however, Dunckley and Dreisbach (2004) used primary tones of unequal level, and they used such strict exclusion criteria that about one-third of females and two-thirds of males presenting with normal hearing were excluded from their study. Such a strong selection bias was bound to have considerable effect on the outcomes. That said, the effect sizes over the frequency range 1.0–5 kHz averaged about 0.11, favoring the females; the effect sizes over the range 6.0–10 kHz averaged about −0.37, favoring the males; and the effect sizes over the range 11.0–15.0 kHz averaged about 0.30, again mostly favoring the females. Because the exclusion criteria focused on the highest-frequency regions, it is those data that were most likely to be affected by the selection bias. All factors considered, the Dunckley and Dreisbach (2004) data appear to confirm the conclusion that sex differences for DPOAEs are small in humans.

The entries at the far right in Table 1 reveal that the ear differences for TEOAEs were quite small (effect sizes averaging about 0.17) and that the ear differences for DPOAEs were essentially zero (effect sizes averaging about 0.07). None of these comparisons achieved statistical significance, which is generally in accord with what is known about ear differences in OAEs, although very little has been published previously about ear differences in DPOAEs. In the past, ear differences in TEOAEs and SOAEs have yielded effect sizes of about 0.26 and 0.17, respectively (McFadden and Pasanen, 1999). Thornton et al. (2003) also reported small effect sizes for the ear difference, especially when the left ear was tested first. Keefe et al. (2008) reported significantly stronger TEOAEs in the right ear than the left for newborns, and no ear difference for DPOAEs, but no analyses of sex differences were reported. Note that, for TEOAEs in Table 1, the effect sizes for the three lowest target frequencies and the wideband condition were generally larger than those for the 4.0 kHz region. We have no explanation for this pattern.

Partial dissociations between DPOAEs and other types of OAE have been observed following administrations of ototoxic drugs (e.g., Wier et al., 1988; McFadden and Pasanen, 1994; Whitehead et al., 1996), and these dissociations are commonly interpreted as evidence in support of DPOAEs originating from a different mechanism in the cochlea from the one(s) responsible for TEOAEs and SOAEs. [The presence of a seasonal effect for the CEOAEs of rhesus monkeys in the absence of such an effect for DPOAEs (McFadden et al., 2006a) also can be interpreted as a dissociation between types of OAE.] If DPOAEs and TEOAEs do originate from different underlying mechanisms, then one would expect them to correlate less highly than do TEOAEs and SOAEs. To test this inference, correlations were calculated between DPOAEs and TEOAEs. The results are shown in Table 2, where the correlations are shown within frequency region, within sex, and within ear. As can be seen, these correlations averaged around 0.5, which is considerably smaller than the correlation of about 0.76 previously reported between TEOAEs and SOAEs (McFadden and Pasanen, 1999). For the Ns involved here, correlations of about 0.32 and 0.42 would be required to be judged significantly different from 0.0 with probability values of 0.05 and 0.01, respectively (two-tailed). Smurzynski and Kim (1992), Moulin et al. (1993), and Gorga et al. (1993) also reported correlations of about 0.5–0.7 between TEOAEs and DPOAEs. In spotted hyenas, the correlations between DPOAE strength in the 3.5–kHz region and wideband TEOAE strength were about 0.16 for females, about 0.01 for males, and about 0.10 when the data were pooled across the 13 females and 14 males in the various treatment groups (McFadden et al., 2006b). For rhesus monkeys, the previously unpublished correlations between DPOAEs and CEOAEs were about 0.41 and 0.28 for all females combined and all males combined, respectively; these values are for the fall, when the male OAEs were weakest (see McFadden et al., 2006a).

Table 2.

Correlations between TEOAEs and DPOAEs, shown separately for the various frequency regions, the two ears, and the two sexes. Ns for females=35∕34 for Lt vs Lt and Rt vs Rt, except=33∕33 at 4000 Hz; Ns for males=33∕32 for Lt vs Lt and Rt vs Rt, except=32∕32 at 4000 Hz.

Condition	Females		Males
	Left vs left	Right vs right	Left vs left	Right vs right
1500 Hz	0.452	0.356	0.483	0.441
2000 Hz	0.527	0.442	0.451	0.465
3000 Hz	0.713	0.552	0.357	0.518
4000 Hz	0.246	0.413	0.404	0.634
Mean=	0.484	0.441	0.424	0.514

For completeness, we show in Table 3 the correlations between left and right ears separately for DPOAEs and TEOAEs in these same subjects. The correlations between ears were generally larger than those between DPOAEs and TEOAEs in the same ear. We note that the within-subject correlations were generally similar in magnitude for DPOAEs and TEOAEs. Were this outcome to be confirmed, it would imply that the linear and nonlinear cochlear mechanisms are about equally similar interaurally. For comparison, Thornton et al. (2003) reported a correlation of about 0.56 between the TEOAEs measured in the two ears, also obtained using the ILO88 system used here, and Berninger (2007) obtained interaural correlations in the TEOAEs of infants ranging from about 0.3 to 0.7 across his half-octave analysis bands.

Table 3.

Correlations between the two ears for DPOAEs and for TEOAEs measured at various frequencies. Ns for DPOAEs=28∕23 for females∕males, except=26∕23 at 4000 Hz; Ns for TEOAEs=28∕22 for females∕males, except=28∕21 at 4000 Hz.

Left ear Condition	Females Right ear		Males Right ear
	DPOAEs	TEOAEs	DPOAEs	TEOAEs
DPOAEs
1500 Hz	0.812	⋯	0.771	⋯
2000 Hz	0.764	⋯	0.708	⋯
3000 Hz	0.679	⋯	0.785	⋯
4000 Hz	0.680	⋯	0.776	⋯
Mean=	0.734	⋯	0.760	⋯
TEOAEs
1500 Hz	⋯	0.641	⋯	0.751
2000 Hz	⋯	0.308	⋯	0.683
3000 Hz	⋯	0.447	⋯	0.647
4000 Hz	⋯	0.711	⋯	0.646
Wideband	⋯	0.624	⋯	0.877
Mean=	⋯	0.546	⋯	0.721

Note that the correlations shown in Tables II and III may underestimate the values that would have been obtained had the subjects originally not been so highly selected for hearing sensitivity. That selection surely introduced a restriction of the range of OAE strength.

In passing we note that, for this highly selected group of relatively young subjects, hearing sensitivity was greater in females than in males (compare McFadden, 1993a, 1998; McFadden and Mishra, 1993). Averaged across the four test frequencies of interest here, the effect sizes for the sex difference were 0.45 and 0.59 for the left and right ears, respectively. By comparison, the averaged effect sizes for the ear difference in hearing sensitivity were only 0.05 and 0.14 for females and males, respectively.

DISCUSSION

Although some other investigators have analyzed for sex and∕or ear differences in DPOAEs (Bonfils et al., 1988; Gaskill and Brown, 1990; Lonsbury-Martin et al., 1991; Moulin et al., 1993; Cacace et al., 1996; Dhar et al., 1998; Bowman et al., 2000; O’Rourke et al., 2002; Dunckley and Dreisbach, 2004; Keefe et al., 2008), this study is among the first to compare the sizes of the sex and ear differences in DPOAEs with those for TEOAEs in the same subjects. In accordance with previous reports, both the sex and ear differences for DPOAEs were smaller than those for TEOAEs.

When interpreting our various outcomes, the reader needs to remember that the DPOAE data all were collected with both primary tones at 75 dB SPL, and that may not be the optimal level for seeing sex differences, ear differences, etc. Also, the duration of the primary tones here was approximately 90 ms, not the longer presentations used by some investigators.

Two anonymous reviewers of this paper noted that the observed difference in the size of the sex difference for CEOAEs and DPOAEs may be partly or wholly attributable to differences in the effective level of the stimuli used to produce these two types of OAE. Specifically, CEOAEs typically are elicited by click stimuli that range in level from about 80 dB peSPL and below. Because the typical CEOAE click is about 100 μs in duration or shorter, the bandwidth of the click extends to 10.0 kHz or higher, meaning that even with relatively strong clicks, the spectrum level of the typical CEOAE click is only about 40 dB SPL or lower. To the extent that the overall strength of the echo depends on the local magnitude of displacement at various locations along the length of the basilar membrane, then the CEOAEs obtained with clicks of this sort are based on effective stimulation that is weak locally. In contrast, the DPOAEs reported here were measured with primary tones that were strong by current standards: 75 dB SPL each. Thus, if the sex difference in OAE strength declined with increasing stimulus level, then the smaller sex difference for DPOAEs than CEOAEs in humans might be just a natural consequence of the relative difference in effective strength of the stimuli typically used to elicit the two types of OAE. That is, the cochlear amplifier might be contributing comparatively more to the typical CEOAE response than to the typical DPOAE response.

To address this suggestion, we examined some DPOAE data from an ongoing study and some CEOAE data from a previous study. The DPOAE data were collected with a wide range of levels for the primary tones and over three frequency ranges. We obtained estimates of the strength of the DPOAE for each subject in each frequency region for primary tones of both 71 and 50 dB SPL, and then we calculated effect sizes for the sex difference between the 36 males and 13 females in that sample. The effect sizes, averaged across frequency regions, were 0.50 and 0.58 for the 71 and 50 dB primaries, respectively. That is, the magnitude of the sex difference was larger for the weaker primary tones, although the increment was not large. In a previous study (McFadden and Pasanen, 1998), we collected CEOAE data from 57 males and 57 females using four click levels. For an 18 dB decrease in click level, the effect size for sex difference increased from 0.70 to 0.73. So, for both CEOAEs and DPOAEs, we have evidence that the locally effective level of the stimuli can affect the magnitude of the sex difference, and in the same direction. Thus, the smaller sex differences seen for DPOAEs than CEOAEs might be attributable in part to the stimuli routinely used to elicit them.

While this insight is unquestionably important when it comes to evaluating explanations for the difference between DPOAEs and CEOAEs, it is less important practically. Regardless of how much the difference in stimulus level may contribute to the difference between the sexes in DPOAEs and CEOAEs, DPOAEs will continue to be a poor measure for investigators interested in sex differences in OAEs and hormone effects on OAEs simply because measurements of DPOAEs with truly weak primary tones typically are impractical, and the sex difference is invariably small when moderate and strong primaries are used.

One way of thinking about the smaller sex difference for DPOAEs than for CEOAEs is that the cochlear mechanisms underlying the production of CEOAEs are sensitive to some agent or agents operating differently in the two sexes at some point early in development, and that the mechanisms underlying DPOAEs are less sensitive to those agents. Because DPOAEs are thought to be the result of both the linear, reflection-based mechanism that underlies CEOAE production plus a nonlinear distortion mechanism operating near the location of the f₂ primary tone (Shera and Guinan, 1999, 2003), a second implication of the present findings is that the greater part of the DPOAE response in the ear canal originates from the nonlinear distortion mechanism, at least for the primary levels used here (75 dB SPL each). Otherwise, the sex difference in DPOAEs should have been more similar to that seen for CEOAEs. One possible agent for the production of the sex difference in the reflection-based mechanism is exposure to androgens prenatally (McFadden, 2002, 2008).

When evaluating explanations for the sex and ear differences in OAEs, it is necessary to consider whether differences in the acoustics of the outer or middle ears are playing a major role. Perhaps the most compelling counterargument is that many investigators adjust the levels of their stimuli in the ear canal prior to collecting both TEOAEs and DPOAEs. This procedure ought to greatly reduce, if not eliminate, any group or individual differences in such dimensions as ear-canal volume. Furthermore, Johansson and Arlinger (2003) observed the human-typical sex difference in TEOAEs even though there were no sex differences in middle-ear compliance or middle-ear pressure in those same subjects. Also, Keefe et al. (2008) reported that ear differences in vLo (the equivalent middle-ear volume averaged from 0.25 to 1.0 kHz) and rHi (energy reflectance averaged from 2.0 to 8.0 kHz) do exist in infants, but those ear differences did not explain the ear differences observed in the DPOAEs or (nonlinear) TEOAEs of those infants. Margolis et al. (1999) reported small, but significant, sex differences in measures of wideband reflectance; however, those differences were primarily at frequencies below 1.0 kHz [their Fig. 5(c)], and the sex differences in OAEs extend to frequencies well above that. Elsewhere, one of us has argued that the sex differences in OAEs may be attributable, at least in part, to differential exposure to androgens during prenatal development (McFadden, 2002, 2008), and that the ear differences may be attributable to asymmetries in the strength of the efferent system (McFadden, 1993a, 1998).

Sex and ear differences have been studied only infrequently in nonhumans, but the existing evidence does suggest that stronger OAEs in females than males may be the basic mammalian pattern. Rhesus monkeys (McFadden et al., 2006a) and sheep (McFadden et al., 2008a, 2008b) show stronger TEOAEs in females than in males, and, just as for humans, the sex difference in rhesus monkeys and sheep was smaller for DPOAEs than for TEOAEs. Lonsbury-Martin and Martin (1988) observed humanlike sex and ear differences in the SOAEs of a colony of pigtail macaques, although the number of males tested and the number of SOAEs recorded both were small. The DPOAE data of Torre and Fowler (2000) also contained a small humanlike sex difference in the youngest group of their rhesus monkeys (all of whom were middle-aged). Valero et al. (2008) found substantial sex differences in the DPOAEs of a primitive New World monkey, the marmoset, but little is yet known about TEOAEs in that species. Guimaraes et al. (2004) reported stronger DPOAEs in female CBA mice than in males. However, this difference was evident only in middle-aged mice, not in young mice, suggesting that this sex difference actually might be attributable to differential age-related hearing loss rather than a developmental difference in the two sexes. McFadden et al. (1999) found no sex differences in the DPOAEs of chinchillas even though an evoked-potential measure of hearing sensitivity suggested that females were more sensitive than males at high frequencies. It is interesting that two amphibian species, leopard frogs and bullfrogs, showed substantially stronger DPOAEs in females than in males (Vassilakis et al., 2004). Establishing whether there is a common mammalian plan for sex and ear differences in OAEs is difficult because TEOAEs and SOAEs generally cannot be found in the small species commonly used for auditory research, and as we have seen, the sex differences exhibited by DPOAEs often are not representative of the sex differences exhibited by TEOAEs or SOAEs in the same ears.

Because the sex differences in TEOAEs and SOAEs exist in newborns as well as in adults (Strickland et al., 1985; Burns et al., 1992, 1994; Morlet et al., 1995, 1996; Thornton et al., 2003; Berninger, 2007), a likely contributing factor to those sex differences is the degree of exposure to androgens during prenatal development (e.g., McFadden, 2002, 2008). The existence of smaller sex differences for DPOAEs than for TEOAEs suggests that the nonlinear cochlear mechanism believed to be primarily responsible for DPOAEs is less susceptible to the effects of prenatal androgen exposure than is the linear, reflection-based mechanism believed to be primarily responsible for TEOAEs and SOAEs. This knowledge has implications for the ultimate understanding of the molecular mechanisms underlying the sex differences in OAEs.

Comparisons across species

Over the years, DPOAEs and CEOAEs have been measured in several mammalian species as well as in humans. One interesting outcome in the present context is that human DPOAEs are considerably weaker than those in all other species tested. For CBA mice, DPOAEs are quite strong, being about 15–40 dB weaker than primary tones that ranged from 55 to 75 dB SPL over the range of about 10–40 kHz (Jimenez et al., 1999). In rabbits, DPOAEs are about 25–30 dB weaker than primary tones ranging from 45 to 75 dB SPL in the vicinity of 7.0–10.0 kHz (e.g., Porter et al., 2006). For rhesus monkeys, spotted hyenas, sheep, and lemurs, the DPOAEs were about 45–55 dB weaker than the primary tones used to produce them (McFadden et al., 2006a, 2006b, 2008b). For marmosets, they were about 50–57 dB weaker (Valero et al., 2008). For the humans in the present study, the DPOAEs were about 65–70 dB weaker than the primaries, in accord with the human studies summarized by Probst et al. (1991). Why human cochlear mechanics produce such weak DPOAEs is not known.

Although the absolute strengths of the DPOAEs vary across species, the absence of a large sex difference was common to the DPOAEs of humans, rhesus monkeys, sheep, and spotted hyenas. By comparison, the effect sizes for the sex difference in marmoset DPOAEs were quite large (Valero et al., 2008); the sex difference in CEOAEs is currently being studied in marmosets.

Comparisons with non-mammals are complicated by the marked differences in the mechanics of stimulation; nevertheless, Meenderink and Narins (2007) have reported DPOAEs about 40–45 dB weaker than the primary tones in the leopard frog, and Kettembeil et al. (1995) have reported DPOAEs about 65–70 dB weaker than the primary tones in chickens and starlings. So when it comes to DPOAE magnitude, humans are more similar to birds than to other mammals. To our knowledge, no one yet has investigated sex differences in the OAEs of frogs or birds.