The response-window regression method - some problematic assumptions: Comment on Draine and Greenwald (1998)

A new response-window regression method for evaluating unconscious semantic priming (S. C. Draine & A. G. Greenwald, 1998) is considered. Four assumptions of the method are discussed: (a) the assumption of standard (ordinary least squares) regression; (b) the assumption of comparable direct and indirect measures; (c) the assumption of linear relationship; and (d) the assumption of forced responding. Situations with strong relationships in the data between indirect- and direct-task measures, coupled with nonstandard regression techniques, provide the strongest evidence from this method. Situations in which the indirect- and direct-task measures show essentially no relationship in the data demand closely reasoned arguments and careful statistical treatment.