A mixture modeling approach was used to assess the existence of latent classes in terms of the perceptions of individuals toward Computer anxiety and subsequently predictors of the identified latent classes were examined. The perceptions of individuals were measured using the Computer Anxiety Scale. Mixture models are ideally suited to represent subpopulations or classes of respondents with common patterns of responses. Using data from a sample of Turkish college students, two classes of respondents were identified and designated as occasionally uncomfortable users and as anxious computerphobic users. Results indicated that the best predictors of the identified classes were variables dealing with past computer experiences.