4.5 Visual word
recognition
word superiority effect
letters in words
(.e.g., name the middle letter in
"cat") can be identified more quickly
than letters in nonwords (.e.g., name the middle
letter in "sek") .
McClelland & Rumelhart
- an interactive activation model of context
effects in letter perception: Part 1. An account
of basic findings
- word
superiority effect
- role of
stimulus quality
- extension to pronounceable
nonwords; e.g. "MAVE"
(contrast w/ random letter strings -
"QLMW")
McClelland and
Rumelhart (1981) sought to develop a
comprehensive account of the word superiority
effect that could be simulated using a computer.
The model provides a number of insights about
what is happening during the recognition process.
The interactive model consists of three levels
corresponding to letter features, letters, and
words. Activation of features from sensory input
activate corresponding letters that contain those
features. At the same time, the letters begin to
activate words that are consistent with the
letters. Feedback from the word level to the
letter level enables decisions to be made faster
about letters in words than in nonwords.
Conversely, nonwords by definition cannot provide
any top-down information to facilitate letter
identification.
The type of model
that McClelland and Rumelhart describe was one of
the first in cognitive psychology to incorporate
a neural metaphor, the idea that information
could be represented as a network of simple
nodes. As well, the idea that the activation of
these units could occur in parallel at different
levels was made concrete in the interactive
model.
model accounts for
basic findings in visual word recognition
- word
superiority effect
- effect of
degraded stimuli
- effect of
pronounceable nonwords
- rich-get-richher
effect=effect of similar words
& frequency (MAVE - HAVE,
GAVE, SAVE)
- gang
effect = neighbourhood size
(e.g., MAVE - SAVE, MALE, MOVE)
|