<html>
Dear Info-Childes Members:<br>
<br>
Thanks for your very, very useful (and prompt) responses to my query
about the regularity of sound-spelling correspondence. <br>
<br>
Below, please find the summary of the responses that were sent to me
privately. I have not replicated the responses that were sent to
the entire list. It looks like Jay McClelland will be helping me to
calculate a course-grained measure of spelling-sound
correspondence. With his permission, I will share whatever we come
up with.<br>
<br>
Again, thanks for your responses. They were most useful.
<br>
<br>
Brian MacWhinney wrote:<br>
<br>
I don't think one could even begin to compute such an index without
making <br>
some simplifying assumptions. What one can do is compute the number of
near <br>
neighbors for a given word. Usually this is done for the words in a
<br>
particular experiment. I agree that computing this for all words would be
<br>
of some value. I think the CELEX folks might have done something like
this.<br>
<br>
Ngoni Chipere wrote:<br>
<br>
A researcher in England did create a measure of sound-spelling <br>
correspondences along the lines you describe, but I can't seem to
remember <br>
his name. His articles were published in education journals (he used the
<br>
measure to predict which words had a greater probability of being
misspelt <br>
by school children), so maybe the ERIC database would be a better place
to <br>
look. I'll look up my mailbox when I get to the office next week, as I
<br>
corresponded once with him, and I may be able find his name & email
address.<br>
<br>
Jay McClelland wrote:<br>
<br>
The frequency-consistency equation discussed on pp. 72ff of Plaut,
McClelland, Seidenberg, and Patterson (1996)<br>
<font color="#0000FF"><u><a href="http://www.cnbc.cmu.edu/~jlm/papers/PlautMcCSeidenbergPatterson96.pdf" eudora="autourl">http://www.cnbc.cmu.edu/~jlm/papers/PlautMcCSeidenbergPatterson96.</a><a href="http://www.cnbc.cmu.edu/~jlm/papers/PlautMcCSeidenbergPatterson96.pdf" eudora="autourl">pdf<br>
</a></u></font>May be of some use to you. You would first need to encode
items and their pronunciations according to the input and output
representations used in Plaut et al (Table 2, p 66). You could then
calculate, for each phoneme j in the correct pronunciation of test word
t, the value of s_j^t, based on equation 12. These s_j^t values could
then be averaged to give an average value for each word, I suppose, our
combined in some other way. Also, it might be useful to take into account
the strength of activation of phonemes not in the pronunciation of the
word. I'd be happy to work with you to compare the adequacy of variants
of this measure and/or to help with some of the calculations.<br>
<br>
Alan Cruttenden wrote:<br>
<br>
Try E.Carney, A Survey of English Spelling London~: Routledge 1994. Some
of <br>
the information there is included in my own book: A.Cruttenden, Gimson's
<br>
Pronunciation of English. Sixth edition London: Hodder<br>
<br>
Yours,<br>
Benjamin Munson<br>
<br>
<br>
<x-sigsep><p></x-sigsep>
Benjamin Munson, Ph.D.<br>
Assistant Professor<br>
<br>
Department of Communication Disorders<br>
University of Minnesota<br>
115 Shevlin Hall<br>
164 Pillsbury Drive, S.E.<br>
Minneapolis, MN 55455<br>
<br>
Department: <x-tab> </x-tab>612-624-3322<br>
Office:
<x-tab> </x-tab>612-624-0304<br>
Fax:
<x-tab> </x-tab><x-tab> </x-tab>612-624-7586</html>