Background and purpose: Nightmare distress (ND) is associated with a broad spectrum of psychopathological conditions such as anxiety, depression and bipolar disorder. Some studies have indicated that dream beliefs play an important role in the occurrence and treatment of nightmare distress. However, existing instruments used to assess dream beliefs either fail to satisfy the requirements of the psychometrics or fail to capture the essence of dream beliefs. This research pursued two objectives: (1) to develop a questionnaire, called the Beliefs About Dreams Questionnaire (BADQ), to measure beliefs people hold about their dreams and (2) to describe the dream beliefs of Chinese college students. Methods: The structure and items on the BADQ were based on the previous literature and were the result of an open questionnaire. Some items were deleted through expert review and the result of predict test. To evaluate its validity and reliability, a sample of 1408 Chinese college students from two universities answered the BADQ, the Chinese version of Van Dream Anxiety Scale (CVDAS), the Dream Survey Questionnaire (DSQ), the Patient Health Questionnaire-9 (PHQ-9) and the Generalized Anxiety Disorder-7(GAD-7) Questionnaire. After two weeks, 95 of them answered the BADQ again. Exploratory factor analysis (n=704) and confirmatory factor analysis (n=704) were conducted to explore and verify the structure of BADQ. The correlation between the CVDAS and the BADQ was calculated to evaluate the divergent validity. Results: The BADQ contains 26 items. Exploratory factor analysis indicated a five-factor structure: Dream omen and health, Dream superstitions, Dream meaninglessness, Dream reality, and Dream attitude. The result of the confirmatory factor analysis also supported the five-factors structure. Acceptable internal consistency (Cronbach's α of all subscales exceeded 0.80) and ordinary to moderate test-retest reliability (the intraclass correlation coefficient of all subscales ranged from 0.467 to 0.713) of the BADQ were presented. Low degree correlation between the CVDAS and the BADQ (ranged from to −0.052 to 0.219) showed that they were two different variables, indicating a good divergent validity. Conclusion: The present investigation revealed moderate to high construct validity and reliability of the BADQ.