Investigating distortions in perceptual stability during different self-movements using virtual reality

Abstract

Using immersive virtual reality (the HTC Vive Head Mounted Display), we measured both bias and sensitivity when making judgements about the scene stability of a target object during both active (self-propelled) and passive (experimenter-propelled) observer movements. This was repeated in the same group of 16 participants for three different observer-target movement conditions in which the instability of a target was yoked to the movement of the observer. We found that in all movement conditions that the target needed to move with (in the same direction) as the participant to be perceived as scene-stable. Consistent with the presence of additional available information (efference copy) about self-movement during active conditions, biases were smaller and sensitivities to instability were higher in these relative to passive conditions. However, the presence of efference copy was clearly not sufficient to completely eliminate the bias and we suggest that the presence of additional visual information about self-movement is also critical. We found some (albeit limited) evidence for correlation between appropriate metrics across different movement conditions. These results extend previous findings, providing evidence for consistency of biases across different movement types, suggestive of common processing underpinning perceptual stability judgements.

Keywords: perception and action; perceptual stability; self-movement; virtual reality.

Figure 1.

Schematic illustration of the three observer movement/target movement combinations.

Figure 2.

Example psychometric functions for two (P15 and P16) of our 16 participants in the six conditions considered. In each panel, the dots correspond to local average estimates over the binary response variable. Curves correspond to fitted psychometric functions to 120 underlying binary responses using a cumulative Gaussian psychometric function form.

Figure 3.

PSS (left panel) and Gaussian s.d. (right panel) data for the three OTMCs and both active (A) and passive (P) movement generation types. Boxplots illustrate median (thick horizontal line), 25^th and 75^th percentiles (hinges) and data points less than 1.5 × IQR from the hinge (whiskers).

Figure 4.

Normalized perceived self-movement distance that would explain the bias observed in our six conditions. Horizontal dashed lines on the Active, (L + R):R condition represent approximate range of equivalent data from Tcheang et al. (2005). The two stars on the S:S condition data represent approximate mean values for the equivalent conditions from Wexler (2003).