Most recent research on error has concentrated on errors of planning, judgement, and action. This paper is concerned with errors in the acquisition of information which are caused by perceptual and attentional mechanisms. Failures to schedule attentional sampling of the environment will lead to accidents. Theoretical models exist for attention and for data acquisition. These models suggest how designers might approach human-machine systems so as to reduce error. A brief summary of the author's feelings about probabilistic risk assessment and related attempts to predict error is provided.