This paper examines the gap between the learning potentials that computers provide and the actual impact of computer use on learning and development. It is argued that the computer's unique potential is derived in part from four basic attributes: information, symbol systems, user activities, and relations with the user. It is hypothesized that these attributes may affect four corresponding cognitions: knowledge structures, internal modes of representation, mental operations, and attitudes and perceptions. These effects may be obtained through "low road" learning (practice-intensive, leading to near automatic responses), or through "high road" learning (thinking-intensive, i.e., nonautomatic operations are mindfully employed). High road learning is seen as the more feasible and promising means to facilitate conceptual learning with computers. The extent to which high road learning occurs depends on the learners' volitional mindfulness, which in turn is partly determined by the materials encountered and by personal, perceptual, and attitudinal factors. It is argued that the opportunity for mapping the computer's attributes on corresponding cognitions often does not take place because learners do not become mindful on their own. The partner-like relationship which the computer can establish with the students can promote mindfulness, but it is the student's choice as to how mindful they will be while interacting with the computer. (Author/THC)