Your Vestibular Thresholds May Be Lower Than You Think: Cognitive Biases in Vestibular Psychophysics

Abstract

Purpose: Recently, there has been a surge of interest in measuring vestibular perceptual thresholds, which quantify the smallest motion that a subject can reliably perceive, to study physiology and pathophysiology. These thresholds are sensitive to age, pathology, and postural performance. Threshold tasks require decisions to be made in the presence of uncertainty. Since humans often rely on past information when making decisions in the presence of uncertainty, we hypothesized that (a) perceptual responses are affected by their preceding trial; (b) perceptual responses tend to be biased opposite of the "preceding response" because of cognitive biases but are not biased by the "preceding stimulus"; and (c) when fits do not account for this cognitive bias, thresholds are overestimated. To our knowledge, these hypotheses are unaddressed in vestibular and direction-recognition tasks.

Conclusions: Results in normal subjects supported each hypothesis. Subjects tended to respond opposite of their preceding response (not the preceding stimulus), indicating a cognitive bias, and this caused an overestimation of thresholds. Using an enhanced model (MATLAB code provided) that considered these effects, average thresholds were lower (5.5% for yaw, 7.1% for interaural). Since the results indicate that the magnitude of cognitive bias varies across subjects, this enhanced model can reduce measurement variability and potentially improve the efficiency of data collection.

Figure 1.

Schematic of the perceptual tasks. (A) This schematic shows three trials, with each trial having a movement followed by a response. The movement was a yaw rotation either to the left or right. The subject then responded to report whether they perceived a leftward or rightward movement (“L” or “R”). The dashed line labeled “effect of the preceding response” illustrates the hypothesized influence of the preceding response on the current trial. (B) This schematic differs only in that the motions are interaural (lateral) translations rather than rotations.

Figure 2.

(A) Example data and psychometric curve fits from one subject. The fraction of positive responses at each stimulus amplitude is shown for all trials (open circles), trials with a preceding negative response (gray − sign), and trials with a preceding positive response (gray + sign). The solid line is a Gaussian function fitted to the data collected during a bidirectional recognition task. The dashed line is the psychometric curve fit to only trials for which the preceding responses were negative (rightward yaw). The dotted line is the psychometric curve fit only to trials for which the preceding responses were positive (leftward yaw). (B) Results of the enhanced model in which fits were done based on current stimulus and prior response, showing predictions for which the preceding response were negative (dashed line) and positive (dotted line). The enhanced model produces two curves with the same slope determined by a common threshold σ. This contrasts with the fits in A, in which the data were “separately” fit to trials for which the preceding response was negative or positive, resulting in two separate threshold values. A larger separation between the curves indicates a larger cognitive bias.

Figure 3.

This figure portrays the values of the threshold (normalized to traditional threshold) according to the two different models developed. Each data point corresponds to a different subject, with open circles representing yaw rotation and filled circles representing interaural translation. Negative slopes indicate that thresholds are lower for the enhanced model versus those for the traditional model.

Figure 4.

Threshold (normalized) versus the coefficient for the effect of the preceding response on the current response for yaw (open circles) and interaural (filled circles) motions. This graph supports the hypothesis that larger effects of the preceding response result in larger overestimations of threshold.