Abstract
Differences between norm ratings collected when participants are asked to consider more than one picture characteristic are contrasted with the traditional methodological approaches of collecting ratings separately for image constructs. We present data that suggest that reporting normative data, based on methodological procedures that ask participants to consider multiple image constructs simultaneously, could potentially confounded norm data. We provide data for two new image constructs, beauty and the extent to which participants encountered the stimuli in their everyday lives. Analysis of this data suggests that familiarity and encounter are tapping different image constructs. The extent to which an observer encounters an object predicts human judgments of visual complexity. Encountering an image was also found to be an important predictor of beauty, but familiarity with that image was not. Taken together, these results suggest that continuing to collect complexity measures from human judgments is a pointless exercise. Automated measures are more reliable and valid measures, which are demonstrated here as predicting human preferences.