Recent empirical studies suggest a need for a ?exible patent regime responding to industry characteristics. In practice, sector-speci?c modi?cations of patent strength already exist but lack theoretical foundation. This paper intends to make up for this neglect by scrutinizing in what direction industry characteristics in?uence optimal patent strength. It is found that patents ought to be weaker, the more intense competition, the higher R&D productivity, and the more intricate reverse engineering are. Unlike similar step-by-step innovation models of economic growth, the model assumes Cournot competition and introduces an empirically substantiated measure of sector di?erences in the ability to catch up with the technological leader. It is found that for most empirically plausible cases the familiar inverted-U between patent length and growth carries over to the Cournot set-up.