Comparison of predictions with observations commonly is thought to be central to the scientific method. Construing prediction deductively, Carl G. Hempel developed this notion into influential analyses of testing, confirmation, theories, and explanation. Late in his career he read and published papers arguing that predictive application of theories to observations was not deductively valid since such inferences make essential recourse to unstatable ceteris paribus "provisos". Hempel's provisos papers widely are regarded as repudiating his earlier influential work.

I argue that Hempel's provisos views (a) are modest extensions of ideas found in his 1934 dissertation and papers written during his 1935-39 Belgium exile and (b) were fully articulated by 1945. Thus they constitute no fundamental late-career recantation. In his provisos problem, Hempelhad his finger on a fundamental philosophical problem how scientific theories connect with observational or experimental results, but his basic positivistic/logical-empiricistic commitments precluded solution of the problem. I show this by comparing Hempel's failure with a solution using the now widely-held Semantic Conception of Theories.

I suggest that Hempel's early work in Germany and Belgium is crucial to interpreting Hempel's later career, that his most famous and influential work may be somewhat plodding development of his juvenilia, and that his intellectual debts to Paul Oppenheim may be underestimated.

Frederick Suppe is Professor of Philosophy, Distinguished Scholar-Teacher, and chair of CHPS at the University of Maryland, College Park. He is a leading expert on the nature of scientific theory and models and on the history of the philosophy of science. His books include The Structure of Scientific Theories and The Semantic Conception of Theories and Scientific Realism. He has published over 100 articles.

Fall 1999