This paper reflects on the kinds of evidence able to confirm that letter and word identification in reading can be supported by encoding the underlying visual structure of the text, and specifically by deriving structural descriptions for letters. It is proposed that structure-driven processes are intimately linked to the implementation of font-specific rules for translating visual features into elements of a letter’s structural description. Evidence for such font tuning comes from studies exploring the impact of font-mixing on reading fluency, and from studies showing how the benefits of experience with a novel typeface can generalise to letters that have yet to be seen in the typeface. After reviewing this evidence, three new experiments are reported which explore font tuning in the context of the lexical decision task. The time course of font tuning, which is monitored by changing the time interval between successive test stimuli, is shown to be sensitive to the overall probability with which successive stimuli appear in the same typeface. In addition, font tuning is shown to reflect item-by-item fluctuations in this probability. Finally, the effects of font-switching are shown to generalise beyond the particular letters present in the text, and to be confined to 1-back transitions. It is concluded that font tuning reflects the implementation of a set of font-specific translation rules held in working memory, and is moderated by the reader’s implicit knowledge of the constraints present in the sequencing of successive portions of text.