We compared the performance of a computerized symptom assessment tool called Choice in terms of number of symptoms and time needed for 92 cancer patients with two other standard methods: interview and questionnaire. Choice outperformed interviews in number and precision of reported symptoms, and did better than questionnaires in focusing on bothersome symptoms. There were high correlations between Choice and questionnaires. Interface and symptom structure play an important role for adequately eliciting patients' symptoms.