Optical coherence tomography (OCT) has proven to be a major technological advance in auditory research for the measurement of sound-induced vibrations from within the organ of Corti (Gao et al., 2014, 2013, 2011; Olson and Strimbu, 2020; Wang and Nuttall, 2010). Compared to the previously used technique of laser Doppler vibrometry (Nuttall et al., 1991), OCT offers three fundamental advantages. First, intracochlear vibrations can be assessed non-invasively with OCT, through the intact otic capsule bone or round window membrane (Cooper et al., 2018; Dong et al., 2018; Lee et al., 2015; Recio-Spinoso and Oghalai, 2018, 2017; Strimbu et al., 2020) as opposed having to open the cochlea to visualize reflective regions or to drop a reflective bead onto the structure to be recorded from. This minimizes the risk of cochlear trauma. Second, rather than being limited to select measurement points, usually wherever the reflective bead lands (typically the underside of the basilar membrane), OCT allows the experimenter to image the tissue so that they can identify specific regions within 3D volume of tissue and then make measurements from any or all these locations (Cho et al., 2022; Cho and Puria, 2022; Dewey et al., 2021, 2019; Lee et al., 2016; Ramamoorthy et al., 2016; Vavakou et al., 2019). This has permitted recordings from throughout the organ of Corti and tectorial membrane at multiple locations throughout the cochlea. Third, because of the ease of making these measurements, the “hit rate” of recording high-quality data from an experimental animal has improved dramatically. Using laser Doppler vibrometry, experimental success rates of 10-15% were common. With OCT, this rate approaches 100%. These benefits have been so dramatic that, within the past five years, nearly every cochlear physiologist has converted to using OCT as their primary means for vibrometry experiments.

The experimental difficulties inherent to laser Doppler vibrometry experiments made it difficult to average data from multiple animals within each experimental cohort and compare the results statistically. However, the simplicity of OCT has now made this possible. This is exciting for the field because it means that proper statistical methodology can be used to test experimental hypotheses. Thus, there is no longer a reason to disregard more modern and rigorous statistical methods. Using better statistical methods is important because it will help to overcome the replication crisis that is common in neuroscience and frankly in all medical research (Aarts et al., 2015; Button et al., 2013; Munafò et al., 2017).

On the other hand, OCT rapidly provides vast amounts of data that can be difficult to manage and perform meaningful statistics. For example, a typical recording from one animal consists of the magnitude and phase of the vibratory response from one location in response to a range of sound stimulus frequencies and intensities. In our lab, we call this “recording a set of tuning curves”. We then record tuning curves from multiple mice in multiple experimental cohorts, with the goal of assessing for differences between cohorts. Given that there may be 20 different stimulus frequencies and 8 different stimulus levels for each experimental condition, this means that there are 160 different vibratory magnitudes and another 160 different vibratory phases for each cohort.

Here, I sought to use modern statistical approaches to compare tuning curves between experimental cohorts to determine if they are statistically different from each other. I tested a linear mixed-effect modeling approach to see if it could fit the magnitude and phase data across all stimulus frequencies and levels to compare the responses. I found that a third-order modeling procedure offers a reasonable compromise between complexity and high-quality fitting and provides a way to easily compare experimental cohorts.