Conférence donnée par le professeur invité Derek Turner, organisée par le département de Géosciences et l’IHPST.
One popular view among philosophers of science is that prediction does not have much of a role to play in historical science, and in the geosciences in particular. Cleland (2011, 2013) defends this view explicitly. This non-predictivism is also implicit in Frodeman (1995) view that geology is an interpretive, hermeneutic science. Another important source for this view is Gould (1989) discussion of the contingency of history. Gould argues that contingency means that the outcomes of historical processes are unpredictable. In this paper, I aim to challenge the view that prediction does not matter in the geosciences. I argue that occasional novel predictive successes have played an important role in historical science, and that non-predictivism depends on taking an overly restrictive sense of prediction (in Cleland case), or else on conflating the unpredictability of contingent historical outcomes with the unpredictability of evidential traces (in Gould case). Finally, I introduce some considerations that help explain why prediction might seem less important in the geosciences than it really is. One problem, for example, has to do with temporal scale. Some of the predictions that geological theories do generate concern processes that unfold at large spatio-temporal scales, and we may not have the luxury of waiting around to see if those predictions are true.