Age-related hearing loss in humans is characterized by progressive loss of threshold sensitivity, especially at high frequencies. A multivariable regression of histopathological metrics from normal-aging human cochleae (Wu et al., 2020) showed that hair cell loss better predicts the audiometric shifts than either neural loss or strial atrophy, however considerable variability in age-related threshold elevation remained unexplained. Here, we develop and apply an algorithm to quantify stereocilia pathology in high-power confocal images of inner and outer hair cells in normal aging human cochleae, aged 21 – 71 yrs. Microdissected epithelial wholemounts of the cochleae were immunostained for myosin VIIa and espin to show cuticular plates and stereocilia, respectively, and each cochlea was imaged at 10 log-spaced locations along the cochlear spiral. An approach based on Fourier transforms was used to quantify the regularity of each stereocilia bundle, and the outcome was compared to a parallel analysis by a human observer. Results show a significant age-related decline in stereocilia regularity and increase in stereocilia loss and fusion. Stereocilia pathology was especially severe on the outer hair cells and in the basal half of the cochlea, and may represent a key contributor to age-related threshold elevations. For the one case with an associated pre-mortem audiogram, the threshold shifts are better predicted from the pattern of stereocilia damage than from the pattern of hair cell loss alone.