ABSTRACT
It is generally assumed that the understanding of the skill of reading should be based in
part on an understanding of the storage and retrieval of words. These processes are often
studied through the use of the lexical decision task, requiring participants to distinguish
words (e.g., CHAIR and FUME) from nonwords (e.g., GREACH and ANSU). In tasks in
which accuracy is near ceiling, three critical findings are seen in the response times: (1)
The word frequency effect. Words that occur regularly in natural language (high
frequency or HF words such as CHAIR) are classified correctly faster than words that
occur relatively rarely (low frequency or LF words such as FUME). (2) The repetition
priming effect. Prior exposure to a word leads to faster correct classifications for that
word on a second presentation. This increase in performance is particularly pronounced
for LF words (e.g., FUME benefits more from prior exposure than CHAIR). (3) The
nonword lexicality effect. Nonwords that look like words (e.g. GREACH) take longer to
be classified correctly than nonwords that are relatively dissimilar to words (e.g. ANSU).
In this article1, we use a new variant of a signal-to-respond procedure that produces
findings in the accuracy domain that mimic those listed above for response times. We will
fit a new Bayesian model, REM-LD, to the data. The advantage of the signal-to-respond