Temporal stability of psychophysiological stress profiles: a re-analysis using intraclass correlation coefficients

This is a re-analysis of data from a previous study which examined the temporal stability of three psychophysiological responses [frontal electromyographic activity (EMG), hand surface temperature, and heart rate]. Each response was recorded on 64 subjects over four sessions, each of which consisted of a 20-min, adaptation period, a baseline condition, and two stressors (one cognitive, the other physical). Rather than using Pearson product-moment correlations, as nearly all psychophysiological test-retest reliability studies have, we have now analyzed the data using intraclass correlation coefficients. This type of correlation allows one to incorporate more than two test-retest values on the same subjects. Analysis indicated that, with the exception of EMG during the physical stressor, the absolute values of the responses had quite significant reliability (.70 or greater). Treating the responses as relative measures (percent change from baseline or simple change scores from baseline) produced smaller and frequently less stable coefficients. It is concluded that statistical estimates of psychophysiological response reliability are functions of the design and particular reliability analysis employed.