Sex differences in the length of the organ of Corti in humans

The length of the organ of Corti (OC) in relation to range of hearing has been strongly implicated as a predictor of the frequency resolving power of the ear (Békésy,¹ Békésy and Rosenblith,² and see Fay³ for a review). This position is strengthened by the fact that in human beings the density of outer hair cells does not change with cochlear length, and the density of inner cells also does not change with cochlear length, except for the most apical few mm (Wright et al.).⁴ Bohne and Carr⁵ found a similar result for the chinchilla. This means that longer cochleas have more inner and outer hair cells than do shorter cochleas. Bohne et al.⁶ conclude their study of myelinated nerve fibers in the chinchilla cochlea with the statement, “In view of the present results, it is reasonable to expect longer cochleas (which contain more sensory cells) to have more spiral ganglion cells.” Nadol,⁷ a leading student of the human eighth nerve, concluded that for humans, longer organs of Corti with more inner and outer hair cells may have more eighth-nerve afferent fibers. For these reasons, a possible sex difference in the length of the cochlea is important as it might imply important sex differences in the numbers of sensory cells, numbers of primary afferent fibers, and in auditory function. One study of the human cochlea (Sato et al.⁸) reported a large sex difference in length of the organ of Corti. Electrophysiological measurement of response delay times combined with assumptions about cochlear mechanics led Don et al.⁹ to a similar conclusion. Casual examination of other sets of anatomical measurements did not indicate to the present author that such a large sex difference existed, which led to the following review and summary of studies that measured the anatomical lengths of the cochlea for both women and men.

Four anatomical methods have been used to measure the length of the OC. (1) The surface preparation method. In the middle 1800’s, Retzius¹⁰ found that the organ of Corti and the basilar membrane were of the same length and that the cochlear duct (scala media) was another $1.5–1.8mm$ longer. Retzius used a dissection method similar to the surface preparation method that was later used by Bredberg,¹¹ Ulehlova et al.,¹² Wright et al.,⁴ Leake,¹³ and Leake et al.¹⁴ By this method, one looks down on the organ of Corti and measures its length along the clearly defined junction of the outer pillar cell with the first outer hair cell. Individual pieces of the OC are cut and laid flat, each piece measured, and the total length taken as the sum of the lengths of the pieces. Takagi and Sando¹⁵ state that this method is accurate, the only problem being the possible loss of tissue during the process of cutting the pieces to be measured. (2) The serial section method. This method is based on serial sections and a projection of the loci of the junction between the heads of the pillar cells onto a plane (a 2D representation). The earliest version of this method was described by Guild¹⁶ and is called the Guild method. According to Bredberg,¹¹ the Guild method ignored about $1mm$ of the basal hook of the cochlear duct. The Guild method was modified by Schuknecht¹⁷ to include complete measurement of the basal hook and is known as the Guild/Schuknecht method. Both of these two-dimensional methods result in shorter lengths than the surface preparation method because: (1) the shorter radius of the cochlear spiral (the junction of the heads of the pillars is more medial than the junction of the outer pillar with first outer hair cell), (2) the rise in the elevation of the cochlear spiral is not included in the 2D projection, and (3) the possibility that, if the sections are not exactly parallel to the mid-modiolar axis, the projections may be foreshortened (Takagi and Sando).¹⁵ It is found here that measures made by the Guild method can be brought into agreement with those made by the surface preparation by adding $1.0mm$ for the hook, as suggested by Bredberg,¹¹ and then multiplying by 1.039 to take into account the radius, elevation, and possible foreshortening. Measures made by the Guild/Schuknecht method only need to be multiplied by 1.039 to be brought into agreement with the measures made from surface preparations. These corrections were found by trial and error to bring the means of the Guild and Guild/Schuknecht methods close to those found by the surface preparation method. (3) The 3D reconstruction method. This method was introduced by Takagi and Sando.¹⁵ Their method uses serial sections of the cochlea, but the 3D coordinates of the junctions of the pillar heads in each serial section are entered into a computer in relation to reference points that do not require that the sections be exactly parallel to the axis of the modiolus. A three-dimensional representation of the line formed by the path of the junction of the pillar heads is created in the computer and its length calculated. In a study of a single cochlea, they found good agreement with the Guild/Schuknecht method when the serial sections were corrected by computer to be parallel to the mid-modiolar axis. This 3D method was applied by Sato et al.⁸ to study sex differences. However, unlike previous investigators, they measured along the inner and outer borders of the basilar membrane (BM) and took the average of the two lengths to represent cochlear length. Their 3D reconstruction of these measurements includes the elevation of the cochlear spiral and has a larger radius than the surface method, as halfway between the inner and outer edges of the BM may fall nearer to the second or third row of outer hair cells than to the junction of the outer pillar with the first outer hair cell. It will be shown that Sato et al.⁸ found the male average to be about $3mm$ longer and the female average to be about $1mm$ shorter than the same averages found by other methods. The reasons for these differences are unknown. (4) The CT method. This method is a 3D reconstruction of the cochlea based on in vivo CT scans of the temporal bone and was introduced by Ketten et al.¹⁸ and Skinner et al.¹⁹ With this method, none of the soft tissue of the cochlea is visible. The centroid of the bony cochlear canal is located in each section of the cochlea and a mathematical spiral is “fitted” the 3D array of centroids so generated.

The literature was searched for measured sex-identified cochleas. If both right and left cochleas were measured for the same individual, the average of two lengths was used. This was deemed appropriate as Bohne et al.²⁰ found for 151 chinchilla cochleas that the correlation between right and left lengths was 0.96, and it was found here for 46 human cochleas that the correlation between right and left lengths was 0.74. In cases where only one cochlea was measured for an individual, possible laterality effects were ignored as the proportion of right and left cochleas did not differ substantially between the sexes (60% right for males and 55% right for females). In addition, when all 46 cases for which both right and left cochleas were measured are considered, the average difference was $0.5mm$ in favor of the right ear. This difference was not statistically significant and only amounted to 0.20 standard deviation units.

Measurements from eleven sources are summarized in Table I. Note that the lengths from Hardy²¹ were converted by adding $1mm$ and then multiplying by 1.039. Lengths from Walby,²² Hinojosa et al.,²³ and Pollak et al.²⁴ were multiplied by 1.039. As described in the discussion of the serial section methods above, these conversions make the measurements using the Guild and Guild/Schuknecht methods comparable to those made by the surface methods. Also, note that no data from Ulehlova et al.¹² and from Wright et al.⁴ were included, as the former only reported on male cochleas and the latter did not identify the sex of the cochleas.

As can be seen in Table I, the mean length of the cochlea for each of the eleven sets of measurements is longer for males than for females. However, by $t$ test none of these differences are statistically significant $(p=0.05)$ except for the results of Sato et al.⁸ They find a substantial difference of 15% and 2.78 standard deviation units. Sato et al.⁸ find the average length of the male cochlea to be $37.1mm$⁠, whereas the average length for the ten other studies is $33.9mm$⁠. Sato et al.⁸ find the average length of the female cochlea to be $32.4mm$⁠, whereas the average length for the other ten studies is $33.1mm$⁠. There are three possible explanations of this discrepancy. (1) The Sato et al.⁸ results simply represent a Type I sampling error, in which case they should be averaged with the others. (2) The Sato et al.⁸ data may contain some unknown factor or error that leads to an overestimation the of lengths of the male cochleas. (3) The Sato et al.⁸ data are, in fact, the most accurate data and represent the true state of sexual dimorphism in the lengths of the human cochlea. It seems prudent to assume that explanation 1 is correct, as it represents all of the available data. Explanation 3 can only be verified by careful, comparative studies of two or more of the methods as applied to many cochleas. Until such studies are conducted and prove otherwise, it appears that there may be a sex difference in cochlear length of $1.11mm$⁠, which amounts to about 3.36% and represents 0.49 standard deviation units. However, whether the observed difference is statistically significant is questionable. One method of analysis was to pool the measurements, as if they had been collected in one study of 109 male and 57 female cochleas, and conduct a $t$ test with 164 degrees of freedom. This method resulted in a statistically significant difference with $p=0.003$⁠. If the data of Sato et al.⁸ are dropped from the pool, the $t$ value failed to reach significance with $p=0.064$⁠. Also, a statistical meta-analysis was conducted using a method described by Hedges and Olkin.²⁵ An unbiased estimate of the effect (⁠$d$ defined on p. 81 of Ref. 25) was calculated for each of the eleven sets of data, and the weighted average (⁠$d+$ defined on p. 111 of Ref. 25) was found. The effect size, so defined, was found to be 0.37 units with a 95% confidence interval of $±$ 1.50 units (defined on p. 112–113 of Ref. 25). By this method, the null hypothesis could not be rejected.

In summary, the average observed length of the male cochlea is about $34mm$⁠, and the average observed length of the female cochlea is about $33mm$⁠. The population standard deviation is $2.28mm$⁠. The range of the 166 cochlear lengths that comprise the data base studied here is $13.78mm$⁠, which is consistent with calculated 6 SD range of $13.68mm$⁠. Thus, the observed sex difference in cochlear length is small in comparison to the observed variability of the lengths of normal cochleas. It is tentatively concluded that there may be a small difference in the lengths of male and female human cochleas even though statistical analyses of the data are not decisive.

The author wishes to thank P. A. Leake of the UCSF School of Medicine for providing cochlear measurements. Also, the following people provided thoughtful reviews of the manuscript: Barbara A. Bohne and Gary W. Harding of Washington University School of Medicine, Dennis McFadden of the University of Texas, and J. C. Saunders of the University of Pennsylvania Medical School.