Humans are incredibly efficient readers despite having no biological endowment for visual word processing. We will contend that this depends on our ability to track regularities in the co-occurrence of word parts, such as letters and morphemes. This would allow us to form a “lexical theory”—how words look like (and perhaps carry meaning) in a given language—rather than a collection of individual word memories. We also tested how general this mechanism is, that is, whether it also applies to visual material that is very different from letters and words, for example because it's not arranged horizontally, or because it refers to abstract visual features that are not even spatially segregated.