The Structure of Scientific Revolutions

by Thomas S. Kuhn

the successful new theory must somewhere permit predictions that are different from those derived from its predecessor. That difference could not occur if the two were logically compatible.