The paper addresses the issue of how the processes of concept formation and change in science can be brought to bear on the problem of incommensurability. It argues that the problem arises out of a methodological approach that identifies the conceptual structure of a science with a language and transfers what is thought to be known about languages to science. Employing a cognitive-historical method that shifts the focus to the representational and reasoning practices of scientists in constructing new concepts provides a way of uncovering the nature of the commensurability relations between successive representations of a domain.