Evaluating finger-temperature data.

Data are presented to support the hypothesis that finger temperature is distributed in a bimodal manner. It is empirically demonstrated that using conventional parametric inferential statistics to evaluate absolute changes in finger temperature is at least misleading and may result in invalid conclusions. Alternative statistical procedures to assess the finger-temperature response are suggested. Also, using a smaller homogeneous sample of subjects, it is shown that lengthy, presession baseline recordings (a standard procedure in temperature research designed to allow subjects to stabilize hand temperature) may serve only to increase the distance between the two modes or distributions. Finally, a suggestion as to why finger temperature is distributed in a bimodal manner is offered and some ideas about how these findings might affect clinical practice are discussed.