Surveys are frequently used to collect information on people’s purchase intentions and nutrition attitudes towards food, and the literature is replete with examples of questions that can be used. Typically the data is analysed using a classical test theory approach: the raw scores of question responses are used in the analysis, e.g. counts of “correct” answers to related questions, or summation and/or averaging across Likert scale items. From a measurement theory approach, this type of analysis is only appropriate when all items are of equivalent difficulty in the questionnaire, because the items provide equivalent information. However, in reality the true/false items in a knowledge subscale will vary in difficulty, and the probabilities of finding a particular score on Likert scale items will also vary, suggesting that some type of item response theory method is more appropriate. While novel to many social scientists, item response theory, and Rasch analysis in particular, has a long pedigree in educational assessment. Rasch analysis uses the survey data to model the probability of a response as a logistic function of the difference between the person and item parameters. Respondents are classified from lower performers to higher performers, and items are classified from easiest to most difficult. In addition, as part of its output, Rasch analysis identifies items that do not appear to be performing as intended, for example where some response options to a question have no probability of being observed this may suggest issues with the number of response options or response option anchoring. The ability to delve into survey data into this level of depth means that Rasch analysis is useful both for questionnaire development, and also for normal survey data analysis. This presentation will demonstrate the usefulness of Rasch analysis in studying food choices by demonstrating the use of the technique on a set of Likert scale items (purchase intent) and on a set of true/false items (nutritional knowledge). The presentation is not mathematical, and will focus on the demonstration of the technique’s usefulness, and interpretation of relevant output will be provided.