2003 IFA Congress: Montreal, Canada

Normal Rates and Disfluencies in French And English

Patricia M. Roberts1 and Ann Meltzer2
1Audiology and Speech-Language Pathology, University of Ottawa, 451 Smyth Road, Ottawa, ON KII-I 8M5 Canada
2Stuttering Treatment Clinic, The Rehabilitation Centre, 505 Smyth Road, Ottawa, On. K1H 8M2,

SUMMARY

Most published studies of normal speech disfluencies and rate of speech have reported on English speakers. To adequately serve non-English speaking people who stutter, we -need data on other languages. This paper examines speaking rate and normal speech disfluencies in unilingual English-speaking adults and French-speaking adults during a monologue task. The rate of speech for English and French speakers was similar but the mean number of disfluencies per 100 syllables in the French group was double that of the English group. English norms for normal disfluencies in non-stuttering adults should not be applied to French (Canadian) speech.

  1. Introduction
Clinicians (and researchers) routinely measure the rate of speech and the frequency and type of disfluencies as part of assessing/studying stuttering (Guitar, 1998). The findings. are compared to rates seen in the normal speaking population. The degree of deviation from the norm is one way to estimate stuttering severity. Intervention strategies and measures of treatment outcome may also be based on the average rate of speech and frequency and type of disfluency in the general population (Andrews & Tanner, 1982; Blood, 1995; Franken et al., 1992; Langevin & Boberg, 1993; Silverman, 1996). To do this, one needs to know the normal range for speaking rate and for normal disfluencies in non-stuttering speakers. When clients in therapy for stuttering express concern that the â_˜new way of speaking’ sounds 'unnatural /different’ it is important for both client and clinician to know that the recommendations are based on sound studies of normal speech.

Normal speaking rates for spontaneous speech in non-stuttering adults stutter are in the range of 170-230 (sprn) (Andrews & Ingham,l97l; Walker 1988). The Texas university students in Lutz and Mallard’s study (1986) spoke more quickly with a median of 217 spin and showed greater variability (118 to 295 spm). Duchin & Mysak (1987) reported a mean rate of 183 words per minute (roughly 256 spm using 1.4 syllables per minute as the conversion factor).

Studies on frequency and type of disfluency in the speech of normal speakers are limited (Conture, 2001). For English-speaking adults, a range of 6 to 9 disfluencies per 100 syllables is found across several studies. Johnson (1961) recorded 6.5 disfluencies per 100/syllables, Duchin & Mysak ( 1987) 7.6 disfluencies /100 words = 5.8 /100 syllables, Meltzer & Mackay (1995) 7.4/100 syllables. However, using these figures as benchmarks for normal speech is problematic due to a number of shortcomings: 1) Limited range of ages, occupations and educational levels; 2) Variation in topics and sample lengths across studies; 3) Different procedures for identifying and counting disfluencies; 4) Failure to take into consideration the effects of bilingualism and regional dialects, although it is known that they can affect many aspects of speech; 5) The studies are all on speakers of English, primarily American English.

There are few reports of rate and disfluencies in the speech of non-stutterers in languages other than English. Some authors have found that the range of spm is wide but not dissimilar from the published ranges for English. European French (Grosjean & Deschamps, 1975), German, Spanish, Dutch, Hindi, and Japanese (reviewed in O’Connell & Kowal, 1972).

Disfluencies identified in these same studies are similar to the patterns of English in many but not all aspects. For example Spanish-speaking adolescents from Mexico and English-speaking adolescents from the United States have a different rate and different types of revisions, repetitions and interjections (de Johnson, O’Connell, & Sabin, 1979). In European French (Grosjean & Deschamps, 1975), and in Igbo (Nigeria) (Nwokah, 1988) speakers may prolong final syllables of words where an English speaker would insert an interjection.

Yairi and colleagues have found that preschool children who do not stutter rarely repeat 1 syllable words, parts of words, and rarely produce prolongations or blocks. Therefore he has dubbed these disfluencies “stuttering-like”, because they are typical of stuttered speech, but not of normal speech. Ambrose and Yairi found that preschool children produce 1 to 2 SLDs per 100 syllables (Ambrose & Yairi, 1999). Therefore, speech with more than 2 % SLDs may indicate stuttering. Although not everyone accepts this view, it has proved quite useful in studies of stuttering in young children. We do not yet know to what extent it may be relevant for adults’ speech.

We have searched the literature and found no current study of disfluencies in a monologue task (used in clinical assessments) for French speakers (or indeed for other languages 1). Yet, in Canada, 6.8 million people list French as their first language (2001 Census), approximately 23 % of the population of Canada. Clinicians who see French-speaking adults who stutter have no data to guide them.

Research questions
  1. What is a normal range for speaking rate in both languages?
  2. What is the range of normal disfluencies in speakers of Canadian French?
  3. Does the 2% SLD guideline proposed by Yairi and colleagues apply to French Canadian adults (or to English speaking adults, for that matter)?
The answers to these questions will be of interest to many clinicians in Canada, and will also demonstrate the need to study speakers of other dialects of French, and other languages.

  1. Method
Participants:

25 English speaking and 24 French-speaking men volunteered for this study.

All were between 18 and 51 years old, completed at least high school, and all denied any history of speech, language .or hearing problems. All had either English or French as their native language with limited or no knowledge of a third language.

Bilingualism is not a dichotomy (bilingual - unilingual), it is a continuum. (Roberts 2001). It is impossible to find “pure” unilingual speakers in Canada, and in many other countries. Almost everyone in Canada takes either French or English as a second language in school. Advertisements and many product labels are printed in both languages. So, we sought English speakers who were as far as possible towards the unilingual end of the continuum of language knowledge. None of them could understand the radio news in French. None of them used French more than 5% of the time.

They rated their ability to speak French as less than or equal to 5 on a scale of 1 to 10. Only 1 person rated his verbal abilities as “5â_ , 3 were “4s” and the other 21 participants rated their French verbal expression as 3, 2, or 1.

Twelve men in the French group said they spoke English very well, rating their verbal ability at 8, 9 or 10 / 10, while 12 rated their English as much more limited: with self-ratings ranging from 1 to 6/10.

CLFP_t1.png

Table 1. Participants

Task

The instructions were: “Tell me all about your job. Tell me all about it- a typical day, what you like, what you don’t like, how you got into the type of work you do. Tell me everything you can about your job.” If the participant stopped, they were asked additional open-ended questions like “tell me more”. Students described their studies.

Sample: the first 500 intended syllables. Repetitions and interjections were not counted in these 500 syllables.

Analysis

Disfluencies: We counted disfluencies 2 ways: First, we counted total disfluencies per 100 intended syllables. For this, all disfluencies were counted: interjections, revisions, repetitions (of more than 1 word, multi-syllable words, 1 syllable words, part words), and prolongations and blocks. Second, the so-called “stuttering-like disfluencies” per 100 intended syllables were counted: repetitions of 1 syllable words, parts of words, prolongations and blocks. When a sound in an interjection was stretched out, we did not count this as a prolongation. It seems to be a normal feature of many interjections.

Rate of speech: All intended syllables in 3 20 second segments were counted. During pauses of more than 1 second (these were rare), the stopwatch was turned off.

Reliability

The first author checked all transcripts, syllable counts and disfluency scoring. A colleague analysed 5 transcripts in each language. Inter-rater agreement for the transcription (word by word agreement) and for total number of disfluencies was 99 and 97% respectively. Agreement for rate of speech was within 4 spm in all but 4 samples.

  1. Results
Rate of speech The rate of speech was 213 spm in French and 229 in English. This difference was not significant (two tailed t test (22) -1.44, p = .81). The French minimum rate was lower and the maximum rate was higher than the English rates of speech (see Table 2).

Total disfluencies

The French speakers were twice as disfluent as the English speakers (see Table 2). This difference was significant (t (36.3) = 6.15, adjusted for unequal variances, p< .001). The French minimum and maximum were both higher than those of the English speakers. The standard deviation of the French group was nearly double that of the English group, reflecting greater variability. There were no outliers in the French group. There were six speakers with rates of 16 or more influences per 100 syllables.

Stuttering-like influences

The English group produced approximately 1 SLD per 100 intended syllables, while the French group produced more than twice as many (2.6 versus 1.1). This difference was significant (t (35.4) = 4.62, adjusted for unequal Variances, p < .001)

CLFP_t2.png

Table 2. Disfluencies and rate of speech

The French speakers all seemed to be speaking quite normall”. None appeared to be stuttering. Why would their disfluency rate be so much higher than the English speakers? We explored several possible reasons for this startling difference. If some of the French speakers were more comfortable speaking English this could explain their high disfluency rates in French. This was not the case. The French speakers all rated their French to be as good or better than their English, for both comprehension and Verbal expression.

A second possible reason could be that the additional processing load for bilingual speakers makes them more disfluent. To check this possibility, we divided the French group in half. Those who rated their English expression as 8, 9 or 10 (out of 10), formed a bilingual group, and those who rated their English as 5 or lower formed the “unilingual” group. (Note: Quote marks here indicate that they were not literally unilingual. They all had some knowledge of English, but were much further towards the unilingual end of the continuum than the other 12 speakers.)

CLFP_t3.png

Table 3. Unilingual French, Bilingual French, and English-speakers compared

As Table 3 shows, the two groups of native French speakers were very similar. T tests found no significant difference between the 2 groups in disfluencies or rate. So bilingualism does not explain the much higher disfluency level in the French speakers.

  1. Discussion
Rate The native French- and English-speaking men spoke at roughly the same rate: 214 or 229 intended syllables per minute. Thus, these two groups of Canadian adults spoke at roughly the same rate as American and Australian English-speakers in the previously cited studies.

Normal speech disfluencies

Although similar in rate, the two groups were strikingly different in their normal speech disfluencies (NSDS), with the French producing double the number of the English speakers when all disfluencies are counted. This high rate is not due to one or two outliers raising the average, nor does it seem related to level of English ability. Two features of the French speech may be related to the high rate of NSDs

Part of the difference in disfluency rates is in the number of revisions: 9.9 in French compared to 6.3 in English. Some revisions seemed due to the need for agreement between articles and nouns or nouns and adjectives. In English one can say “the” before selecting the noun which will follow. In French, this is not possible, since the article must agree with the following noun. French speakers often said le, paused, and then had to revise this (to la or les) to finish the sentence. The same pattern occurred for “some” in English, compared to du, de la, and des in French. Thus there is a type of revision in French which does not occur in English.

Stuttering like disfluencies:

The unilingual English speakers produced a mean of 1% SLDs, while the French mean was more than double this: 2.6% SLDs per 100 intended syllables.

There were two components to this higher SLD figure. The French men repeated 1 syllable words much more than the English men did. Often these were articles or pronouns. Example: le le le bureau or c’est lui qui qui qui qui fait ca. A post-hoc check on the number of one syllable word repetitions found a mean of 9.6 per person for the French group and only 2.9 for the English group for the entire 500 syllable sample.

The other component of the SLDs in French was prolonged sounds. Where an English speaker would have used an interjection, the French speakers sometimes prolonged a sound. Twenty-two of the 24 French speakers produced prolongations, whereas only 11 of the English group did.

Clinical Implications

We need replications with larger samples, and different monologue topics to confirm these findings (see Roberts & Meltzer, forthcoming). However, these initial results suggest that:
  1. Speaking rate is remarkably similar for the two languages, so similar goals for “normal range” can be used in treatment
  2. Using English norms to assess French speakers in Canada is inappropriate, whether it is total number of disfluencies or number of SLDs
  3. The types of disfluencies are differently distributed in the two languages. Therefore, it seems inappropriate to use the NSD - SLD labels in French.
  4. The 1 to 2% SLD figure found in the speech of non-stuttering children also held for the English speaking adults.
  5. The range of variability is much bigger in French than in English. Reasons for this merit further investigation.
Acknowledgements
The authors thank the following graduate students who tested participants, transcribed and analyzed tapes: Lindsay White, Patrizzia Mazzocca, Caroline Tarte, Nathalie Bertrand and Christine Kelm. Joanne Wilding performed the reliability check for English speakers. This study was funded by a grant from the Institute for Rehabilitation Research and Development, The Rehabilitation Centre, Ottawa.

References
Ambrose, N.G., & Yairi, E. (1999). Normative disfluency data for early childhood stuttering. Journal of Speech, Language, and Hearing Research, 42, 895-909.

Andrews, G., & Ingham, RI. (1971) Stuttering: considerations in the evaluation of treatment. British Journal of Disorders of Communication, 6, 129-138.

Andrews, G., & Tanner, S. (1982). Stuttering: the results of five days treatment with an airflow technique. Journal of Speech and Hearing Disorders, 47, 427-428.

Blood, G.W. (1995). A behavioural-cognitive therapy program for adults who stutter: computers and counselling. Journal of Communication Disorders, 28, 165-180. -

Conture, E.G. (2001). Stuttering: Its nature, diagnosis and treatment. Boston: Allyn and Bacon.

Duchin, S.W., & Mysak, E.D. (1987). Disfluency and rate characteristics of young adult, middle aged, and older males. Journal of Communication Disorders, 20, 245-257.-

De Johnson, T.H., O’Connel1, D.C., & Sabin, EJ. (1979). Temporal analysis of English and Spanish narratives. Bulletin of the Psychonomic Society, 13, 347-350. Section 7. Multi-cultural and Multi-lingual Aspects of Stuttering 395

Franken, M.C., Boves, L., Peters, H.F.M., & Webster, R.L. (1992). Perceptual evaluation of the speech before and after fluency shaping stuttering therapy. Journal of Fluency Disorders, 17, 223-241.

Grosjean, F., & Deschamps, A. (1975). Analyse constrative des variables temporelles de l’anglais et du francais: vitesse de parole et variables composantes, phénomenes d’hésitation. Phonetica, 31, 144-184.

Guitar, B. (1998). Stuttering: An integrated approach to its nature and treatment. Baltimore: Williams & Wilkins.

Johnson, W. (1961). Measurements of oral reading and speaking rate and disfluency of adult male and female stutterers and non-stutterers. Journal of Speech and Hearing Research Monograph Supplement. 7, 1-20.

Langevin, M., & Boberg, E. (1993). Results of an intensive stuttering therapy program. Journal of Speech-Language Pathology and Audiology, 17, 158-166.

Lutz, K.C., &, A.R. Mallard (1986). Disfluencies and rate of speech in young adult stutterers. Journal of Fluency Disorders, 11, 307-316.

Meltzer, A., & Mackay, I.R.A. (1995). Normal disfluency in stutterers. In C.W. Starkweather & H.F.M. Peters (Eds.). Proceedings of the First World Congress on Fluency Disorders. Munich, Germany: International Fluency Association.

Nwokah, E.E. (1988) The imbalance of stuttering behaviour in bilingual speakers. Journal of Fluency Disorders, 13, 357-373.

O’Connell, D.C., & Kowal, S. (1972). Cross-linguistic pause and rate phenomena in adults and adolescents. Journal of Psycholinguistic Research, 1, 155-164.

Roberts, P.M. (2001). Aphasia assessment and treatment for bilingual and culturally diverse patients. In R. Chapey (Ed.). Language intervention strategies in aphasia and related neurogenic communication disorders. (4"â_˜ ed.) (pp. 208-232). Baltimore: Lippincott, Williams & Wilkins.

Roberts, P.M., & Meltzer, A. (forthcoming). Rate of speech, number and distribution of normal speech disfluencies in English- and French-speaking adults.

Silverman, F.H. (1996) Stuttering and other fluency disorders. (2’â  ed. ). Neeham Heights, MA: Allyn & Bacon.

Walker, V.G. (1988). Durational characteristics of young adults during speaking and reading tasks. Folia Phoniatrica, 40, 12-20. 396 Theory, research and therapy in fluency disorders

Translation

The IFA implemented Japanese translations of some pages on the site for the 2018 Joint World Congress. Choosing Japanese below to see these translations.

Not all pages are translated, but you can use Google translate to see a machine translation using the switch below

Google Translate