We consider development of statistical models for analysis of high-dimensional questionnaire data. We seek to infer hierarchical structure between the questions and between the subjects. The answers to the questions are represented in terms of a small set of latent factors, generalizing low-rank formulations. To infer meaning of these latent factors, we also develop a topic model for the text of the questions, with the topic model analysis performed jointly with analysis of the question answers; this allows us to place words (meaning) on each of the inferred latent factors. Finally, we demonstrate how the statistical model may be employed to adaptively ask questions of individuals, so that one may infer understanding about a person through (adaptive) asking of a small subset of questions. Results are presented on real data, for a large number of questionnaires and subjects.