We review the current literature on attribute non-attendance in stated choice and use data from beef and chicken choice experiments using both inference and the respondents' own statements. Inference is based on panel data analysis by mixed logit models of choice with both discrete and continuous mixtures of coefficients, and is conditional on the observed pattern of choice. Information from respondent statements on non-attendance is directly embedded in the specification of the indirect utility function. Results show no clear winner between the inferential approaches, but the inference based on constrained latent class panel models better matches the observed data. © 2012 The Author (2012). Published by Oxford University Press on behalf of the Agricultural and Applied Economics Association. All rights reserved.