Variations in Maximum Amplitude of Facial Expressions between and within Normal Subjects

Abstract

Definitive proof of efficacy of preventions and therapeutic interventions, and of risk factors in lower motor neuron facial paralyses continue to be confounded by the lack of repeatable quantitative measures of outcome. Clinical and research experience with human facial expression repeatedly demonstrates wide variations between subjects. To our knowledge, little information is available to isolate and describe the differences in dynamic facial expression between and within normal subjects. The purpose of this study is to use a statistical model to analyze the components of the observed variations of maximum amplitude measurement of image change during normal human subject facial expressions. Seventeen consecutive normal adult human subjects with no current or past evidence of facial nerve or ear disease were studied. Videotapes of command facial expressions were taken using specific and standardized conditions. The tapes were analyzed using a new computer-assisted image-change analysis program capable of generating dimensional data for the maximum amplitude of expression. These data were statistically analyzed using a General Linear Model with Nested variables to isolate and define component variations and errors. The General Linear Model predicted 88% of the observed total variation ( p < 0.05).* A model performance this high suggests that most of the important independent variables were being studied. The major component of the variations was the difference among (between) subjects. Seventy-seven percent of the predicted variation was due to this difference ( p < 0.05). Little of the variation (1%) seemed to be within-subjects. Test-retest agreement was acceptable. Most of the data were tightly clustered about the mean and there was no stochasticly significant difference between test-retest ( p = 0.1187). We are encouraged by these results. They suggest the potential value of this and other dimensional techniques applied to facial expression. This study demonstrates that a computer-assisted image-change analysis program is capable of generating dimensional data that can be statistically analyzed in order to isolate and define component variations and errors.