ABSTRACT
This chapter focuses on the processing of letters when they are presented in isolation. It applies a state-of-the-art Bayesian modeling technique, the Structural Forms algorithm, to study the mental representations of Roman letters. The seminal work by Townsend overcame some of the caveats by producing what are possibly the most comprehensive alphabetic confusion matrices to date. Multi-Dimensional Scaling is a well-developed analysis technique and a mainstay in uncovering structure in data. The Structural Forms algorithm uses the principles of organization present in the data set to produce a visual model that most likely represents the relationship between the data points. Conrad's acoustic confusions provide an interesting perspective on letter identification. First, perhaps another model governs the processing of Roman letters that only incidentally shares the misses-but-no-false-alarms assumption tested earlier. Finally, despite the advantages of the Structural Forms algorithm, there is a potential gap between the requirements of the algorithm and the hierarchy hypothesis.
