Neural activity during visual attention task is a combination of high and low level visual processing. Low level processing consists of oculomotor activity and processing of physical image properties such as contrast, luminosity and frequency. High level processing on the other hand is guided by memory, image content and intention. This experiment investigates the effects of local changes in physical image properties on neural activity in a scene perception task. The overall aim of this study is to gain more insight into what guides visual attention, that is, how the oculomotor, low level physical image properties and high level factors related to image semantics interact in scene perception. In addition, we hope to improve the co-registration of eye movements with EEG/MEG that allow for more ecologically valid experimental setups.
The experiment is a free-viewing scene perception task where subjects will be presented with three kinds of visual stimuli: natural scenes, scrambled scenes and blank scenes. The natural scenes consist of photographs. The scrambled stimuli consist of same images as the first ones but meaningful content is randomly scrambled inside a 4x4 grid. Scrambled scenes still retain the same local image properties as the natural scenes. In addition to the two stimuli a control condition of a blank screen with same dimensions and grid is used. In each trial, the subject is presented with two images. First image is a small patch equal to the size of one grid element in the 4x4 grid. The second image is a full scene image and the subject is asked if the patch can be found in this image.
Preliminary study of the validity of the experimental setup was conducted at the Lund Humanities Laboratory with 10 subjects. In the pilot experiment, subjects' eye movements were recorded using a high-speed eye-tracker. The experimental setup was as described above, but the blank condition was omitted. The results showed that median fixation duration and the median dwell time on the areas-of-interest (AOIs) visited were longer for the natural scenes as compared to the scrambled scenes.Conversely the number of AOIs visited was larger in the scrambled scenes. These results coincided with the results reported by Foulsham et al. (Atten. Percept. Psychophys, 7, 2011, pp.2008-2025).
Currently the final experiment with EEG/MEG and eye tracking is underway. In this experiment the neural activity is recorded with EEG/MEG concurrently with eye movements. Simultaneous co-registration of eye movements allows us to bypass common obstacles related to free-viewing paradigms with EEG/MEG. The information about where the subject is looking at during the experiment (provided by the eye-tracker) enables us to synchronize neural activity related to local properties of the stimulus images.