The goal of this dissertation was to better understand theoretical and empirical issues regarding how humans process the visual, orthographic forms of words. Central to this goal was a clarification of the concept of orthographic similarity which has been used by researchers for several decades for model building and theory testing across a variety of visual word processing contexts. Two general approaches were taken over the period of inquiry in examining orthographic similarity as it applies to visual word processing.

The first approach sought to expand models of wordform interaction in memory through the use of a graph theoretic, network-based approach. This approach sought to expand the conception of orthographic neighborhoods and explore how orthographic neighborhood structure compares to power law, scale free structures found in other language domains such as semantics and syntax. The results of the orthographic network analyses show that power law structure in orthographic word-forms arises in specific circumstances in which the degree of letter overlap is very high and occurs in contiguous letter positions among words.

The second approach used a form-priming, lexical decision task to test the priming effects of orthographic similarity during word processing in order to validate the orthographic similarity metrics used to define orthographic networks in the first approach. The experimentlal results found facilitory orthographic priming effects for primes and word targets of different lengths which supports the claim that the concept of orthographic neighborhood should be revised to include words that differ in length. The facilitory priming effects in the experiments were limited to conditions with 67 ms prime durations in which orthographic similarity was measured by the SOLAR metric (Davis, 2006). The facilitory priming effects found in this study supports the “sloppy”, relative-letter position encoding of the visual sytem (Davis & Bowers, 2006) during the early stages of word recognition.