This study reports on the legibility of digits, symbols, and letters from 20 onscreen typefaces. Participants were asked to identify individual characters after a brief exposure and performance was recorded. Classification tree analysis was used to examine the influence of specific character features on legibility. Features included attributes such as height, width, weight, stroke variation, contrast, and complexity. Results reveal the most influential features for six commonly confused characters (zero, one, "e", "l"., and $). The use of classification tree analysis appears to be promising for typographers in determining the limits of typeface features and for practitioners who must choose an optimal typeface for a particular use.