167 engineering students evaluated through a questionnaire the usefulness of a vocal and of a manual input modality for computerized systems. Five factors, grouped in triplets, were systematically manipulated in a set of scenarios: the type of task, the mode of presentation of information to the users, the task difficulty, the task duration, and the task frequency. Each triplet included the first two factors with one of the last three factors. The main results show that in "benign" conditions where the task is either short, not frequent, or easy, the users perceive either input modality as equally useful. In "stressful" conditions where the task is either long, frequent, or difficult, two general reactions are observed: the users prefer a manual input modality in spatial tasks where the information is presented visually, and slightly prefer a vocal input modality for all the other situations.