Vestibular perceptual thresholds for rotation about the yaw, roll, and pitch axes

Abstract

This effort seeks to further assess human perception of self-motion by quantifying and comparing earth-vertical rotational vestibular perceptual thresholds about the yaw, roll, and pitch axes. Early seminal works (Benson Aviat Space Environ Med 60:205–213, 1989) quantified thresholds for yaw, roll, and pitch rotations, using single-cycle sinusoids in angular acceleration with a frequency of 0.3 Hz (3.33 s motion duration) and found yaw thresholds to be significantly lower than roll and pitch thresholds (1.58–1.20 deg/s vs. 2.07 deg/s and 2.04 deg/s, respectively). Our current effort uses modern methods and definitions to reassess if rotational thresholds differ between these three axes of rotation in ten human subjects at 0.3 Hz and additionally across a range of frequencies: 0.1 Hz, 0.3 Hz, and 0.5 Hz. In contrast to the established findings of Benson et al., no statistically significant differences were found between the three rotational axes at 0.3 Hz. Further, no statistically significant differences were found at any of these frequencies. Instead, a consistent pattern was found for yaw, pitch, and roll of increasing thresholds with decreasing rotational frequency, consistent with the brain employing high-pass filter mechanisms for decision-making. We also fill a gap in the literature by extending the quantification of pitch rotation thresholds to 0.1 Hz. Finally, we assessed inter-individual trends between these three frequencies and across all three axes of rotation. In thoroughly considering methodological and other differences between the current and previous studies, we conclude yaw rotation thresholds do not differ from those in roll or pitch.

Appendices

Appendix

Examining sampling precision between studies

The testing methods of these two studies (ours and Benson et al.’s) vary in measurement precision when estimating underlying thresholds. This study collected responses using a semi-adaptive staircase with 100 trials whereas Benson et al.’s study collected responses using an adaptive staircase based on a Wald sequential probability test with 50 trials. In general, performing more trials yield a more precise threshold estimate; however, more responsively adapting the stimuli magnitudes can enhance efficiency (Karmali et al. 2016). To quantify this tradeoff, we again performed simulations: now with an underlying 1 – σ threshold ranging from between 1 and 4 deg/s to broadly capture the range of individuals’ rotational thresholds at 0.3 Hz. Figure 7 reveals that Benson et al.’s sampling procedure has relatively worse precision (in terms of a higher coefficient of variance in fitted threshold) than this study’s sampling method, across a broad range (1–4 deg/s) of underlying subject 1 – σ thresholds.

Fig. 7

Comparison of the coefficient of variance (standard deviation / mean) of the estimated threshold distribution found from 10,000 Monte Carlo simulations, for simulated subjects with underlying 1 – $\sigma$ thresholds ranging from 1 to 4 deg/s (representing the rang of thresholds observed at 0.3 Hz). Comparisons are made to an analytical formulation for the best coefficient of variance that is possible with optimal sampling, using 100 trials or 50 trials, as well as a 3-down 1-up (3D1U) staircase procedure with 50 trials (Karmali et al. 2016)

Comparison of head-on-axis and head-off-axis (200 mm) in roll and pitch

In a secondary study, we looked for potential differences in roll and pitch 1 – σ rotation thresholds between two head location conditions: with the rotation axis centered near the middle of head by the inner ear (what was done in the Results section of this study for all rotations) and rotations 200 mm below the head, towards the feet (roughly about the T2 vertebrae; what was done in Benson et al. for roll and pitch). This was assessed in separate subject groups for roll and pitch, at approximately 0.5 Hz. In roll, 14 subjects were tested head-on-axis and 11 head-off-axis; in pitch, 6 subjects were tested head-on-axis and 8 head-off axis. The demographic of subjects was similar to the demographic collected for the main study, consisting of college aged individuals.

No statistically significant differences were found using a 2-factor ANOVA (factors: rotational axis and head configuration) between rotational axes (F(1,36) = 0.01, p = 0.93) or between head configurations (F(1,36) = 0.09, p = 0.77) of the log-transformed threshold data. This was also the case for follow-up independent two-sample t-tests between the two groups (head on-axis or 200 mm off-axis) in roll (t(23) = 0.212, p = 0.83) and in pitch (t(12) =  – 0.65, p = 0.53). See Fig. 8a and b for comparisons in roll and pitch, respectively.

Fig. 8

a Comparison for roll at ~ 0.5 Hz between head-on-axis and head-off-axis groups b Comparison for pitch at ~ 0.5 Hz between head-on-axis and head-off-axis groups

Grouping subjects in roll, the geometric mean was 0.98 deg/s (10% difference from this study’s mean at 0.5 Hz) with a 95% CI of 0.82–1.16 deg/s; grouping subjects in pitch, the geometric mean is 1.00 deg/s ( – 6.5% difference from this study’s mean at 0.5 Hz) with a 95% CI of 0.73–1.36 deg/s.

Comparison of legs straight and legs bent (90 degrees) in roll and pitch

In another secondary study, in both roll and pitch, two groups of college aged subjects were tested in one of two conditions: legs straight (what was done in this study and Benson et al.) and legs bent (what was done in Wagner et al. for roll and pitch and Lim et al. in roll). Again, all tests were performed with motions at approximately 0.5 Hz, and we looked for differences in estimated 1 – σ rotation thresholds. In roll, 17 subjects were tested with legs straight and 14 legs bent; in pitch, 15 subjects were tested with legs straight and 6 with legs bent.

No statistically significant differences were found using a 2-factor ANOVA (factors: rotational axis and leg configuration) between rotational axes (F(1,49) = 0.02, p = 0.88) or between leg configurations (F(1,49) = 0.05, p = 0.82) of the log-transformed threshold data. This was also the case for follow-up independent two-sample t-tests between the two groups (legs straight or legs bent) in roll (t(29) = 0.20, p = 0.84) and in pitch (t(19) = 0.28, p = 0.78). See Fig. 9a and b for comparisons in roll and pitch, respectively.

Fig. 9

a Comparison for roll at ~ 0.5 Hz between legs straight and legs bent groups b Comparison for pitch at ~ 0.5 Hz between legs straight and legs bent groups

Grouping subjects in roll, the geometric mean is 0.98 deg/s (9.9% difference from this study’s mean at 0.5 Hz) with a 95% CI of 0.80–1.20 deg/s; grouping subjects in pitch, the geometric mean is 1.01 deg/s (-6.9% difference from this study’s mean at 0.5 Hz) with a 95% CI of 0.79–1.29 deg/s.