Reliability of Online Surveys in Investigating Perceptions and Impressions of Faces

We analyzed the reliability of online surveys in investigating the perceptions and impressions of facial photographs that demonstrate various skin conditions using computer graphics (CG). Online surveys allow us to obtain a large amount of data, although the quality of visual stimuli is more difficult to control remotely. Meanwhile, the statistical analysis of a larger volume of data may reduce errors despite the unavoidable lesser control of experimental conditions. In this study, we conducted online and laboratory experiments with less controlled (Online) and well-controlled conditions (Control), respectively. For each experiment, 100 participants rated the same questionnaire regarding their impressions of the same 28 CG facial photographs. The results showed significant correlations between the two experiments for all 19 items in the questionnaire. Regarding the errors, standard deviations from the stimuli and individual differences were 56–84% and 88–104% in the Online and Control groups, respectively. Moreover, the rates of mismatching perceptual evaluations to the physical features demonstrated in the photographs were 4.9–9.7% on average in an additional online survey of another 2,000 participants. These results suggest that online surveys can be applied to experiments to investigate impressions from CG facial photographs by obtaining an appropriate number of participants to offset larger errors.