| |
Abstract:
Word frequency and spelling-to-sound regularity produce
robust behavioral effects in word reading. Low-frequency irregular
words are pronounced more slowly and less accurately than
low-frequency regular or high-frequency words. Can fMRI identify
regions sensitive to these features? In a prior study, sixteen
subjects read aloud lists of words, both mixed and blocked by
feature. Twenty-five words were presented in each of eight
event-related fMRI runs. Data for mixed and blocked lists were
analyzed separately. Across data sets, average images including all
word types revealed similar, but not identical, patterns of
activation. Regions typically observed during reading were evident
(e.g., motor cortex, midline frontal regions, lingual gyrus,
cerebellum, etc.). An ANOVA revealed effects of frequency and/or
regularity, but these were neither consistent nor robust. To
determine whether this resulted from relatively few observations
for each feature, eight additional subjects were run using only
mixed lists. A 2-factor (frequency, regularity) ANOVA was performed
on the 24 subjects. Statistically significant effects of frequency
and/or regularity (including interactions) were clearly present in
several regions. Timecourses were extracted from these regions.
Noteworthy were "pure" frequency effects in posterior regions that
were easily visualized in timecourses. These data indicate that
reliable effects of stimulus characteristics on reading can be
imaged using fMRI, though the number of subjects required may be
large. Supported by NIH-NS06833, NS32979, McDonnell Center for
Higher Brain Function.
|
