Evaluation is central to Labov and Waletzky’s (1997) narrative model; without it stories lack any ‘point’. But in narratives elicited from witnesses and suspects at the start of police interviews, evaluation is markedly sparse or absent. This paper examines the presence of three kinds of evaluation in the interview’s questioning phase: interviewer evaluation, elicited evaluation and interviewee evaluation, focusing on discourse markers and evaluative patterns and frameworks that reveal how evaluation is carried out in a range of question and response speech acts (Stenström 1984) and looking at the marking of stance by interviewers in relation to the evidential value of the elicited detail. It shows how interviewers and interviewees change their stance and footing, moving from interrogator to therapist and from cooperation to non-cooperation. Drawing on pragmatic principles from conversational analysis of institutional interaction (Drew and Heritage 1992) and from appraisal theory (Martin 2000), analysis reveals features of contested and collaborative evaluation, marked in turns that reveal concessive and adversative positions. Conclusions point to the function of evaluative frames as important features of interviewer activity, suggesting that these function to achieve a change of state in suspect and witness knowledge and in the evidential value of information.