Abstract
The ways in which scientists account for and justify their own scientific views are analyzed by examining in detail transcripts of interviews with biochemists working on oxidative phosphorylation. It is shown that scientists use two repertoires, the `empiricist' and the `contingent', to account for their beliefs. The empiricist repertoire derives from and reinforces the traditional conception of scientific rationality according to which data obtained from impersonal, standardized routines are used to establish the validity of hypotheses and to decide between competing theories. However, when the contingent repertoire is adopted, `facts' are seen as depending on fallible interpretative work. Both repertoires are used in informal interaction, scientists moving flexibly between the two as they construct accounts of theory-choice. In view of this variability of accounts, it is concluded that it is impossible to obtain definitive evidence of how theories are actually chosen and that a new form of sociological analysis is required. An attempt is made to illustrate such an analysis.