Do separable dimensions matter? The effect of graph complexity on errors of interpretation

The purpose of this thesis was to better understand some reasons why errors occur when graph perceptual complexity varies. This question is interesting because perceived graph complexity may be more affected by factors other than the amount of information or the pattern of the data. The problem of line graph complexity was addressed here by investigating the extent to which graphical information can be efficiently processed with brief exposure times, when distracter line size, color and orientation varied. The results from four experiments showed that line graphs can be interpreted accurately in 50 ms. The results also showed that orientation, color and size are separable dimensions (Garner, 1974) in a line graph context as evidenced by a lack of interference and also by the adoption of specific processing strategies that are only available when dimensions are separable. Errors of interpretation could be understood in terms of state- and process limitations as well as in terms of parallel and serial processing mechanisms. Parallel processing was assumed to occur when target and distracter lines were consistently oriented due to 'configural superiority effects' (Pomerantz, 1981). Serial processing was expected in inconsistent distracter line configurations due to low discriminability (process limitation) between the two lines (Gamer, 1974). Differences in processing strategies appear to be related to the number and nature of the stimulus set permutations as well as the nature of the distracting information. Exploring separable dimension combinations, namely orientation-color, orientation-size and color-size, helped to explain performance efficiency by identifying situations under which state- and process limitations could account for errors and by identifying processing strategies that were most likely adopted to enhance performance.