Curious what other types of inferences we can make with the kind of data we’ve collected? Here, in no particular order, we offer a few:
Aesthetic Ratings:
- There is no significant difference in the average aesthetic ratings for seascapes and shipscapes, nor in the variability of those ratings: On a scale of 0 to 100, the mean aesthetic rating for seascapes is 67.9; the mean aesthetic rating for shipscapes is 68.5; the standard deviation of ratings for seascapes is 22.4; the standard deviation for shipscapes is 21.6.
- There is substantial variability in the ratings. One standard deviation is equivalent to a 1/5 of the ratings scale.
- There is a small, but significant positive correlation between both the time spent viewing the stimulus and the aesthetic rating attributed to that stimulus, such that observers spent more time looking at images they liked. (Trivial).
- A standard linear classifier can indicate with accuracy above chance whether a viewer gave ratings in the top or bottom 25% of all aggregate ratings based solely on fixation coordinates alone.
Scanpath Analysis
- There is significant variability across subjects in the ways they move their eyes around the canvas.
- Accelerometry from functional data analysis shows some subjects move their eyes rapidly and erratically from place to place; others move slowly, and in a more direct fashion.
- No single feature of the scanpath analysis (the position of the first and last fixation, the mean distance between fixations, the mean duration per fixation, the total extent of the path, or the number of fixations) is a significant or meaningful predictor of aesthetic response.
Region of Interest Analysis
- Many of the regions of interest demarcated from theory are borne out by the data: regions of interest on average attract a more than 150% greater density of fixations than the average for the canvas at large.
- The horizon is a particularly compelling region of interest across stimuli, attracting on average a more than 303% greater density of fixations than for the canvas at large.
- Even ROIs that include details tend to attract a disproportionately large quanity of fixations, with an average increase of 270% compared to the canvas at large.
- There is no systematic relationship (in terms of regression coefficients) between how frequently a viewer fixates a certain region of interest and how aesthetic that viewer rates the image to be.