The law of initial value: a rare exception.

The well-known negative correlation between initial value and difference score, referred to as the law of initial value (LIV), is analyzed in the present study. It is shown that whenever the correlation between the initial and final values is less than 1.00, the negative correlation is influenced by the a(a - b) effect. By relating values to the first principal component axis, this spurious effect can be eliminated, thus allowing analysis of any real dependency on initial values. Data collected in a psychophysiological experiment were used to test this hypothesis. A subject sample of 125 male students experienced various challenges (Cold Pressor Test, breath holding, reaction time measurement, digit-series test) while several physiological variables (e.g. blood pressure, stroke volume, electrodermal activity) were monitored. Results fail to support the LIV as originally advanced by Wilder. On the contrary, after eliminating the a(a - b) effect, a positive dependency as indicated by the slope of the first principal component axis (anti-LIV) was observed. The reciprocal relationship between the LIV and anti-LIV is discussed with respect to the measures employed. Advantages and disadvantages of various methods to correct for the initial value dependency are presented.