Temporal dynamics of eye movements and attentional modulation in perceptual judgments of structure-from-motion (SFM)

How attention plays a role in resolving ambiguous perceptual judgments is one of the age-old scientific questions. Understanding the processes of perceptual grouping, switching, processing speed, and awareness is a key step towards solving significant problems in applications such as computer vision and automatic driving involving three-dimensional space. The rotating three-dimensional (3D) structure from motion (SFM) is a well-known bistable ambiguous stimulus. Thus far, it is still an open question how attention, eye movements, and depth cues modulate perceptual judgments of rotating directions of a 3D SFM. As early as 1925, motion perception resulted from eye tracking signals was proposed by Helmholtz (Cavanagh, 1992). Pomerantz (1970) claimed that eye movement plays an important role in the occurrence of kinesthetic perception. Furthermore, neuroscience studies has supported a common neural basis for eye movement and attention transfer (Grosbras, Laird & Paus, 2005). The current study aimed to investigate the characteristics and the time course of eye movements during SFM perception by controlling exogenous visual cues and ascertain the effect of attention on SFM perception. Using advanced eyemovements analysis, we investigated how attention under both unambiguous and ambiguous depth cues modulate perceptual judgments of rotation directions in deepth. Twenty-two experiment. Their task was to indicate the rotation directions of 3D SFM by pressing the left (for clockwise CW percept from top view) or the right key of a mouse (for CCW percept) with their left or right thumb. A computer simulated structure-from-motion (a 3D rotating sphere) was created via 30 coherently moving dots with 0.2° diameter each along with an elliptical trajectory of different radii at a mean velocity of 5°/s. The luminant dots were randomly distributed in a spherical area extended 5°×5°. Under unambiguous depth conducted using methods from Bonneh, Adini & Polat (2015) and Hermens & Walker (2010). The statistical analysis revealed that perceptual judgments of rotation directions under unambiguous cues were faster and more confident than those under the ambiguous conditions. For the micro-saccade, peak velocity and amplitude were higher during perception of unambiguous 3D rotation than those during the ambiguous rotations. There was no significant difference in saccade duration. When participants judged the SFM as rotation of clockwise (left), their microsaccade rate towards left was significantly higher than that towards right and vise versa while the counter-clockwise judgment was made. Under the unambiguous condition, significant differences between CW-cw and CCW-ccw were found during time widows of 150~400 ms and 500~970 ms. In contrast, ambiguous conditions (AMB-cw and AMB-ccw) differed most during 700~950 ms, which indicated extra time of attentional processes. Our findings of temporal dynamics of the ambiguous and unambiguous perceptual judgments of 3D rotations indicated two stages of processing. First, local speed calculation in three-dimensional structure construction during initial period of 150~200 ms after stimulus onset. Second, visual processing binds local motion vector flows to the overall perceptual judgment of rotation directions. The ambiguous conditions took longer time. When rotations were unambiguous, attentional facilitates during perceptual judgment of 3D rotation of SFM speed up in the higher-level processing.