Statistical Parametric Mapping of Gaze Duration: A Novel Application of a Spatially Extended Statistical Approach to Eye Movement Data

Statistical parametric mapping (SPM) software was implemented in the early 1990s so that neuroscientists could test spatially extended hypotheses using functional imaging data, usually in 3D space and allowing for a mass univariate approach to hypothesis testing that is agnostic to where any significant effects may lie. Here, we apply the same approach to gaze duration data, i.e. visual fixations, collected using a virtual reality headset, which extends across a large 2D area of visual space, measuring 32{degrees} either side of central fixation and 24{degrees} above and below this point. In order to evaluate this novel method, we measured the locus of average gaze in a group of 17 patients with hemispatial inattention to the left, a neurological condition caused by damage to the right parieto-frontal brain networks, that induces a systematic bias in lateralised visual attention. This causes people to experience difficulty in paying attention to one side of space, both in their extrapersonal world and relative to their own bodies. We used a free visual exploration paradigm (viewing multiple naturalistic scenes for 7 seconds), which is sensitive to spatial biases encountered in this condition. 23 age-matched and neurologically healthy controls also took part. The visual stimuli were original and mirror flipped versions (Left to Right ie L-R) to correct for any lateralised informational biases inherent in the images. When compared with age-matched controls, the patients exhibited an average spatial shift of attention of 18{degrees} to the right of the midline. We demonstrated this approach using patients with hemispatial inattention, but it can be applied to any fixation-based or dwell time data. This is an advance on current methods that generated visual heatmaps or attentional maps, as our technique allows formal testing of spatially extended hypotheses on gaze duration data using a standard, frequentist statistical approach.