LINGUIST List 12.1576

Fri Jun 15 2001

Sum: Measuring Standard Fluency Rates

Editor for this issue: Lydia Grebenyova <>


  1. Gina, Measuring Standard Fluency Rates

Message 1: Measuring Standard Fluency Rates

Date: Thu, 14 Jun 2001 13:13:48 +0100
From: Gina <>
Subject: Measuring Standard Fluency Rates

For Query: 12.1396


Awhile ago I posted a question of
what are all the existing different standards and procedures for measuring a
speaker's (dis)fluency index or (dis)fluency rate.

Many thanks to all who replied! These were the responses...

Dr. Lee Froninilono wrote:

You may want to attempt a cross index modulation of intonation rate reflex
over a an EEG model. This way, fluency will be measuted not over a stat
period but over the natural language curve of, say, a standard rate scheme.
Moreover, a non reduction is had, keeping full one to one relation between
both items!

Gedalyovich Chaim and Leah wrote:

> i measure speech rate as syllables per minute (as i was taught by susan
> bloch in melbourne) others do the same or measure words per minute (eg.
> ingham from sydney and at one stage janet costello who i think swapped
> afterwards to syllables per minute).
> you then count stutters per minute and work out the percentage of
> dysfluencies. criteria are usually 1% or 2% dysfluencies per minute are
> considered pathological. this type of counting results in both number and
> length of dysfluencies influencing the final 'score'. check van riper
> it's a bit old it's a classic.
> other things to take into account are types of dysfluencies (repetitions
> versus elongations and shwa insertion) as well as secondary symptoms (eg.
> facial grimacing)
> in clinical work i take a base rate of percent dysfluency at every
> this gives me a good measure of progress made.
> hope this helps!
> leah

Dr. Robin Lickley wrote:

 If you're talking about disfluency in normal speech, which, I assume you
are, the "standard" measure seems to be "disfluent event" per hundred words.
It is not always clear, though, whether the hundred words includes the
disfluent event itself.

I actually used a different measure in my thesis (1994) - words/disfls, but
in more recent work I and colleagues here in Edinburgh (*ref below) have
counted disfls per 100 "intended" words, like Shriberg (in her PhD, 1994,
c.f. in ) who counts disfls
per 100 "clean" words (i.e. removing reparandum words, and fillers etc).

Oviatt, (1995 in Computer Speech and Language, 9.) counts disfls per 100
words, but seems to include the disfluent event in the word count.

See also Bortfeld et al in a recent Language and Speech article.

There's nothing new in all this, of course. Maclay and Osgood (Hesitation
Phenomena in Spontaneous English Speech, in Word 15:19-44) also describe
some of their data in rates per hundred words.

If you want to see a disfluency annotation scheme that seems to work, go to:

If you're talking about "standard" counting of disfluency rates in stuttered
speech, I could do another email on that topic.

All the best,


(*) H. Branigan, R.J.Lickley and D. McKelvie. (1999) Non-linguistic
on rates of disfluency in spontaneous
speech. (in Proceedings of the ICPhS, San Francisco, pages 387-390).
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue