frequency counts for consonant clusters in English

[Benjamin Munson]

Dear Michèle and fellow info-childes list-mates:

Unless I am misunderstanding the question, I think that there potentially 
huge differences between frequency counts from CELEX, which would tell you 
type frequencies of clusters (i.e., word-initial /pw/ occurs in exactly one 
monomorphemic word, pueblo; word-initial /tw/ occurs in 28 words, including 
twixt, twelve, etc) and corpora of spoken language, which could indicate 
either type of token frequency counts (i.e, word-initial /pw/ occurs in one 
word in the Switchboard corpus, and this one word appears 3 times).  CELEX 
can only tell you the type frequency of a cluster.  Corpora like 
Switchboard or Childes or the like could tell you either the type or the 
token frequency for a cluster.

In my dissertation and subsequent work I have argued that type and token 
frequency should have predictable, dissociable effects on phonological 
acquisition and processing: high token frequency should make a sequence 
fluent, but not necessarily flexible.  Sequences with high type frequencies 
should be flexible and fluent.  I conceive of flexible as generalizable to 
novel forms and to novel speaking tasks.  See my paper in the August 2001 
issue of the Journal of Speech, Language, and Hearing Research for a 
fleshed-out version of this argument.

Cordially,
Ben Munson
Department of Speech-Language-Hearing Sciences
University of Minnesota, Minneapolis



At 05:40 AM 5/20/04, you wrote:
>Dear colleagues,
>I am a Phd student working on the production and perception of consonant 
>clusters in people with Down Syndrome.
>
>I need to find the frequency with which certain clusters appear in spoken 
>English, for example word-initial 'pl', 'str' or word-final 'lp', 'fs' etc.
>
>Does anybody know if this data is readily available out there, or would I 
>have to perform a frequency count myself?
>
>I have been advised to use CELEX for this, but I was wondering whether it 
>would also be possible to use CHILDES & how one would go about doing this.
>
>Any ideas & comments welcome!
>
>Michèle Pettinato
>
>
>
>
>