Generating hypotheses after 14 years of marital followup; or, how should one speculate? A reply to DeKay, Greeno, and Houck.

Gottman and Levenson (2002) for the purpose of post hoc speculation, developed a number of ordinary least squares regressions to model the length of marriage of divorcing couples in a 14-year longitudinal study. We believe that our analyses are appropriate for our purpose. We do not agree with DeKay, Greeno, and Houck (2002) that a duration-model approach would have been more appropriate, and instead argue that the analyses used are more powerful and generate more interesting speculations. When speculating, one makes an important contribution just by being interesting, not necessarily by being right. The purpose of post hoc speculation is to generate discussion, and we are pleased that even at the outset we have accomplished this goal. In this reply to DeKay et al., we argue that the two-process model for earlier versus later divorcing that we propose is both interesting and clinically useful.