The article describes several studies in which respondents were asked to write autobiographical stories and were given personality questionnaires. The authors developed scales of content analysis maximally close to Cloninger’s TCI-questionnaire. They show that correlation between test results and content-analysis data does exist, yet it is mediated by a large number of parameters such as the subject’s bias, his or her self-image and readiness to share information with the psychologist. These factors go far beyond the usual social expectations. They authors argue for reforming the practice of questionnaire making and including new «provocation-scales» and «trap-scales» in the tests.