Response variability as a predictor of skin temperature biofeedback performance

The purpose of the present study was to compare the power of three subject variables as predictors of performance in a skin temperature biofeedback task. Data from three related experiments (N = 52) designed to train digital skin temperature increases in four sessions were pooled. Three measures (mean skin temperature, standard deviation and standard error of estimate), derived from a prefeedback instructional control session, were correlated with three criteria derived from training sessions (temperature change in the first, last and mean of training sessions). There were significant correlations with standard error of estimate, the largest with the mean change measure, accounting for 14% of temperature variance. The findings are consistent with an operant conditioning model of biofeedback learning, and have implications for more optimal training methods and experimental designs based on assessment of the standard error of estimate.