2003 IFA Congress: Montreal, Canada

The Influence of Syntactic Variables on the Development of Stuttering

Dieter Rommel
University of Ulm, Section of Phorziatrics, Schillerstr. 15, 89077 Ulm, Germany


The influence of selected psycholinguistic variables on the further course of childhood stuttering is presented and discussed by examining the study population from t0 (first contact, N=71) to t9 (4;6 years later). The stuttering children and their mothers were videotaped in 6 month-intervals in a play situation. The transcribed utterances of the children were analyzed with a sophisticated computer-based language profiling system. The results confirm the research data obtained up to date, stating that stuttering occurs more frequently with grammatically more complex and longer sentences. Results suggest that some syntactic variables are important with respect to the course of stuttering.

  1. Introduction
This research report is part of the 54-month multifactorial prospective longitudinal research project of the Ulm study group that started in 1992. This study, its course and its questions have been previously described (Johannsen et al., 1994). The main objective of the study is to find valid indicators to estimate the chance of a prospective chronic or recovered career in stuttering and to answer the still open question if and how it is possible to differentiate normal developmental disfluency from chronic disfluency. The results of the study are also considered to be very important with respect to therapeutic strategies in dealing with stuttering children (Rommel & Burgmaier, 1993). In the field of psycholinguistic variables first longitudinal data and results have already been published (Rommel, 1995, 1998, 2000, 2001; Rommel et al., 2000a, 2000b). Results obtained by our working group to date suggest that neither the stuttering symptomatology at first contact (quantitative and qualitative characteristics), the linguistic loci of stuttering (linguistic loci such as the type of sound transitions, word class, word and sentence length), the child’s linguistic and cognitive abilities (psychometric tests), the linguistic level of the child’s utterances (e.g. mean length of utterance, number of different and total words) nor so-called time pressure variables of mother and child (e. g. articulatory rate, tum-taking-velocity/behaviors) may significantly contribute to long-term prognosis of the course of stuttering. In contrast, the predictive power of the mother’s verbal behavior in the syntactic and semantic field is significant, which means, the more complex the sentences (mean length of utterance) and/or the more differentiated the vocabulary the mother uses (number of different words), the higher the prospects for a chronic course of the disease. The result is the same for the sum of total words. These findings correspond with the results of the longitudinal study with at risk-children (high family burden of stuttering) performed by Kloth et al. (1999), where also those children whose parents have a higher linguistic level remain stutterers. The outcomes of the longitudinal studies carried out by working groups with Yairi (e.g. Watkins et al, 1999; Yairi et al., 1996) and Ryan (2001) also show that the course of stuttering is independent from the linguistic abilities of the children. Such linguistic abilities are, as a trend, normal for the respective age-group or, contrary to all expectations, rather over-average when compared to normal and/or control children (see also Rommel et al., 2000 a, b). However, some partial results of our study and the findings of further working groups (see also the review of Bernstein Ratner, 1997) give reason to investigate the children’s utterances more thoroughly, beyond the range of analysis performed so far, using a linguistic profile analysis. So we find for example that foremost those stuttering children face a prospective chronic course of the condition who produce longer fluently spoken sentences with a distinctly higher articulatory rate and/or produce more frequently constative instead of expressive utterances (Rommel, 2000, 2001). Fortunately, first therapy studies have also been published; although tainted by methodical flaws (e.g. samples too small) (Bonelli et al., 2000, Onslow et al., 2001), those studies evidence that a reduction of stuttering by means of behavioral therapy methods may be accompanied by changes in linguistic variables (e.g. verbal output stagnation, N. Bernstein Ratner, personal communication, 13.12.2001). Therefore in the Ulm longitudinal study the question has arisen to which extent more detailed analyses of the children’s verbal behavior might still yield relevant findings with respect to the understanding of the course of stuttering in childhood (and might therefore also be important and noteworthy for diagnostic and therapeutic purposes).

  1. Design and sample description
In Figure 1 the study design is depicted. At the time of first contact (to), data concerning history, the actual stuttering and supposedly related physiological, psycholinguistic and psycho-social variables were collected. In the further course of the study, selected characteristics were reassessed in six-month-intervals. Children who had become permanently fluent could leave the study; however, they always had the possibility of rejoining in case their stuttering should reoccur. Some of the families concerned wanted to participate in further control examinations despite the fact that their child had become fluent; this was not refused. The following criteria to evaluate whether a child has become fluent were employed:

  • A distinct reduction of stuttering towards and below the limit of 3 % of words stuttered
  • The quality of disfluencies changed distinctly towards less stressed disfluencies, such as relaxed word-repetitions - No signs for associated behavior and/or avoidance behavior.
  • Agreement with the parents.
Initially 84 children participated in the study. Between t0 and t3 13 children and between t3 and t9 another 9 children dropped out, so that at t3 reliable data are available for 71 children and at t9 for 62 children. One and a half years after the first contact 36 children were still stuttering, four and a half years after the first contact at t9 the number had decreased to 14 still stuttering children.


Figure 1. Study design 408 Theory, research and therapy in fluency disorders

As can be seen in Table 1, the children were brought to our outpatient department at a mean age of 5 years because of stuttering. The children began to stutter at a mean age of 3 years and 1 month; the time span between the onset of stuttering and the point where professional help was sought was one year and 11 months. The sample group consisted of 53 boys and 18 girls. The children stuttered an average of 6.2% of the words at to. The differences in stuttering, age at onset ofstuttering, duration of stuttering, hereditary influence, language and cognitive abilities with respect to the childrenâ_˜s gender are not significant. For the assessment of cognitive abilities the German version of the Kaufmann Assessment Battery for Children (K-ABC, Kaufman & Kaufman, 1983) was applied. To test the language development the Heidelberger Sprachentwicklungstest (H-Sâ_ E-T, Grimm & Scholer, 1991) and the Aktiver Wortschatztest (AWST, Kiese & Kozielski, 1979) were employed. The H-S-E-T is a battery of 13 subtests, each of which is appropriate for a different linguistic area, assessing both receptive and expressive syntactic, semantic and pragmatic abilities. The AWST is a vocabulary test which requires the naming of non-colored drawings from every day objects and activities. Results showed no remarkable deficiencies in cognitive or linguistic performances of the whole group at entry in the longitudinal project. The mean results in K-ABC, H-S- E-T and AWST confirmed an appropriate state of the measured abilities. This is also true for the subtests in the K-ABC and in the H-S-E-T.


Table 1. Description of the sample at first Contact (t0)

  1. Methodology and results of the loci part of the study
The verbal behaviors of the stuttering children in a play situation with their mothers were videotaped in 6 month-intervals. The utterances were transcribed and analyzed according the rules of a very sophisticated computer-based language profiling system (COPROF, Clahsen & Hansen, 1991; assessed variables also described in Clahsen, 1991). The system is a German adaptation of the Language Assessment Remediation and Screening Procedure (LARSP) developed by Crystal et al. (1990). For the purpose of this study the fluent and stuttered utterances of the children were assessed separately, in such a way that for each child two separate profile analyses were performed. This method of linguistic analysis allows to do the following investigations: First, the assessment of the linguistic loci of stuttering provides an answer to the question to what extent certain syntactic structures can trigger stuttering or promote more fluent speech. Second, the same language profiling system can be used in order to check to what extent certain syntactic Variables can contribute some information on the prognosis of the further course of stuttering. The computer program COPROF ensures an objective and reliable analysis and allocation to the different linguistic categories. This program also offers detailed help functions and examples for every decision-step. In preliminary studies and during the training period of the raters, interobserver agreement > 90 % were achieved with respect to the classifications given by COPROF. Excluded from analysis were unintelligible, interrupted, ambiguous, formalized, simple, imitative and stereotype utterances. Furthermore the verb flexion was not analyzed, as our sample showed (due to the local Swabian dialect) various dialect-specific fiexion and conjugation variants. For ellipses no sentence structure analyses were performed, the grammatical description was limited to the internal structure of the phrases used. Coordinated subclauses were analyzed separately like main clauses. In table 2 the most important evaluation - and therefore linguistic - categories and their allocation to the supposed linguistic developmental stages are given. It should be mentioned that Clahsen et al. (o.c.) do not take the stages VI and VII in the sense of Crystal et al. (o.c.) into consideration.


Table 2. Description of the assessed syntactic variables and their allocation to linguistic development phases according to Clahsen et al. (o.c.)

In Table 3 it can be seen how many utterances were analyzed altogether at to, how many of them were not assessable and which types of utterances, sorted and summarized according to development stages, occurred at a certain frequency. In this table information on the most frequent and also most relevant categories in one developmental stage was added. All in all it becomes clear that the children can be allocated to two separate developmental stages (4 vs 5) with respect to fluent or disfluent speech according to MLU and that this is mostly so because children still often use stage I utterances especially in fluent speech (see table 2). Furthermore it becomes evident that stuttered utterances can be significantly more often analyzed, are not elliptic and longer (p <.00l). Results Show that all important linguistic structures are available to the children, they, however, do not employ them spontaneously to the expected extent. In this connection the many omissions are noteworthy that are rather typical for stage III.


Table 3. Absolute frequency of analyzed, not analyzed and elliptic utterances with distribution of analyzed utterances over the five stages of linguistic development

In order to be able to evaluate statistically to what extent certain linguistic structures coincide with stuttering, the absolute values (absolute frequency of occurrence of the variables presented in Table 2) were transformed into relative percentage values. The corresponding number of fluent or stutte-red utterances that had been analyzed were used as factor to calculate the relative percentage values. It is to be borne in mind that the sum of the relativized variables is > 100, since more syntactic criteria may apply per utterance. In order to display the results, the generic categories stages I - V can be used, as it became very quickly evident that the individual syntactic categories (see Table 2), if they occurred at a sufficient frequency (see Table 3), gave an identical picture. This means, the results presented for the individual linguistic developmental stages are generally also truefor the corresponding differentiated linguistic categories. The dominance of individual subcategories in the corresponding developmental stages also speaks well for the usage of the stages as generic categories.


Figure 2. Influence of syntactic variables, summarized according to linguistic developmental level, on stuttering at 10

It becomes evident, and this is in complete accordance with result obtained to date (Bernstein Ratner1997, Rommel, 1995, 1998, 2000, 2001), that linguistically more complex sentences more frequently trigger stuttering than syntactically simpler sentences. More precisely, it can be stated that syntactic structures from stage III on significantly increased the probability of the occurrence of stuttering. It can be ruled out that individual syntactic structures within one generic stage is particularly relevant with respect to triggering stuttering.

  1. Results of the prognostic part of the study
This part of the study investigated if certain general syntactic variables (summarized stage I-V variables) at to could be used for a prognosis of the disorder 4.5 years after to. To test the prognostic relevance of these variables, the 62 children were divided into two groups at to, one group consisting of children who would be fluent at t9 (N248), and one group with children who would still be stuttering at t9 (N=l4). The results obtained in this section are not nearly as consistent and obvious as in the loci part of the study. First it can be stated that the average length of utterance (MLU in words) of fluent and/or disfluent parts of the Children’s speech does not contribute to the prognosis. The same holds true for the usage 0" ellipses, the frequency of occurrence of missing elements, of utterances that cannot and/or can be analyzed. Also the syntactic variables of fluently spoken parts of speech do not contribute to the prognosis. However, it is noteworthy and very interesting that a linguistic analysis of stuttered utterances points to some variables relevant to the further course of stuttering. In Figure 3 it can be seen that above all those children remain stutterers that stutter at to more frequently with linguistically less complex utterances (stages 1 and 3) and, complementary to this, stutter less frequently with stage IV utterances. In other words, above all those children are more likely to become fluent that stutter at an increased rate with linguistically more complex utterances and that are, from a linguistic point of view, therefore disfluent at a higher language level. This interpretation is also confirmed by an additional loci analysis (see paragraph 3) of different groups (recovered vs persisting stuttering), the results ttiereof, however, cannot be discussed here for space reasons.


Figure 3. Influence of syntactic variables of stuttered parts of speech at to, summarized according to linguistic developmental levels, on the course of stuttering

  1. Summary and discussion
The results suggest that a more detailed linguistic analysis, above all of the stuttered utterances that, according to results obtained so far, are mostly linguistically more complex (see also Bernstein Ratner, 1997, Rommel, o.c.) than the fluent parts of speech (loci part of the study), appears to be promising regarding the prognostic power. It seems that foremost those children are at risk to remain stuttering whose stuttered utterances are at a syntactically lower level. With the help of a differentiated linguistic analysis of the stuttered utterances it appears to be possible to identify a subgroup of stuttering children that are particularly predetermined to develop chronic stuttering. It is also to be borne in mind that the results of our and other working groups (see Introduction) suggest that also with children where stuttering persisted there are no general linguistic deficits, so that the results may be interpreted mainly in three ways:
  • First, already at an early developmental stage (duration of stuttering < 2 years) some stuttering children begin to employ linguistic avoidance strategies (unconscious, in the sense of verbal conditioning) and/or avoid to broaden their linguistic repertoire.
  • Second, in the sense of Berstein Ratner (1997, pp. 103, 119), the results point to a sublime linguistic deficit of the stuttering children with respect to the generation of complex linguistic utterances; such deficit cannot be assessed with language tests available to date.
  • Third, in this connection, in particular when you are used to thinking in a multidimensional and case-oriented way (e. g. Rommel & Burgmaier, 1993) also more complex interactions and combinations of problems are possible, where apart from linguistic variables also physiological (e. g. speech motor abilities) and psycho-social variables (e. g. verbal and non-verba1parent-child interaction Variables, see for example Rommel, 2000, 2001) are to be taken into consideration.
As a therapeutic consequence it turns out that in particular with behavioral therapy approaches (fluency shaping, operant conditioning procedures) the fact of what is reinforced must be considered in a more differentiated way. Therefore not only linguistically simpler sentences are involved, but also linguistically more complex fluent and especially stuttered sentences, in order to timely counter-act a possible unfavorable development of stuttering. Moreover it might also be possible in the sense of the second interpretation to implement special language-therapeutic elements (e. g. supporting and forcing of linguistically challenging but of course age- and development-appropriate utterances) into the therapy plan.

Bernstein Ratner, N. (1997). Stuttering: A psycholinguistic perspective. In R.F.Curlee & G. M. Siegel (Eds.), Nature and treatment of stuttering: new directions (pp. 99-127). Boston: Allyn and Bacon.

Bonelli, P., Dixon, M., Bernstein Ratner, N., & Onslow, M. (2000). Child and parent speechand language following the Lidcombe Programme of early stuttering intervention. Clinical Linguistics and Phonetics, 14, 427-446.

Clahsen, H. (1991). Child language and developmental dysphasia. Linguistic studies of the acquisition of German [Normale und gestorte Kindersprache, tranlated by K. Richman]. Amsterdam: John Benjamins (German original version published in 1988).

Clahsen, H. & Hansen, D. (1991). COPROF - Computerunterstiitzte Profilanalyse. Koln: Focus.

Crystal, D., Fletcher, P., & Garrnan, M. (1990). Grammatical analysis of language disability. (2 ed.) London: Whurr.

Grimm, H., Scholer, H. (1991). H-S-E-7: Heidelberger Sprachentwicklungstest (2. Aufl.). Gottingen: Hogrefe.

Johannsen, H. S., Schulze, H., Rommel, D., & Hage, A. (1994). Stuttering in childhood: A five- year longitudinal study in progress. Special Issue: Fluency disorders. Folia Phoniatrica et Logopaedica, 46, 241-249.

Kaufman, A.S. & Kaufman, N.L. (1983). K-ABC: Kaufman assessment battery for children. Circle Pines, MN: American Guidance Service. (German Version by Melchers, P. & Preuli, U., Amsterdam: Swets & Zeitlinger, 1991).

Kiese, C. & Kozielski, P.-M. (1979). AWST 3-6, Aktiver Wortschatztest fiir drei- bis sechsjéihrige Kinder. Weinheim: Beltz Testgesellschaft. '

Kloth, S. A. M., Kraaimaat, F. W., Janssen, P., & Brutten, G. J . (1999). Persistence and remission of incipient stuttering among high-risk children. Journal of Fluency Disorders, 24, 253-265.

Onslow, M., Bernstein Ratner, N., & Packman, A. (2001). Changes in linguistic variables during operant, laboratory control of stuttering in children. Clinical Linguistics and Phonetics, 15, 651-662.

Rommel, D. (1995). Psycholinguistic Aspects of Stuttering in Childhood. Cross-sectional results with 50 pre-school stuttering children. In C.W.Starkweather & H. F. M. Peters (Eds.), Stuttering: Proceedings of the first world congress on fluency disorders. Munich, Germany, August 8-11, 1994. (pp. 168-174). Nijmegen, University Press: The International Fluency Association.

Rommel, D. (1998). Psycholinguistic variables and the development of stuttering in childhood. In E.C.Healey & H. F. M. Peters (Eds.), 2nd world congress on fluency disorders. Proceedings. San Francisco, August I 8-22, 1997. (pp. 124-130). Nijmegen, University Press: The International Fluency Association.

Rommel, D. (2000). The influence of psycholinguistic variables on stuttering in childhood. In H.- G.Bosshardt, J. S. Yaruss, & H. F. M. Peters (Eds.), Fluency disorders: Theory, research, treatment and self-help. Proceedings of the third world congress of fluency disorders in Nyborg, Denmark (pp. 195-202). Nijmegen: Nijmegen University Press.

Rommel, D. (2001). Die Bedeutung der Sprache fur den Verlauf des Stotterns im Kindesalter. Sprache - Stimme - Geho'r, 25, 25-33.

Rommel,D. & Burgmaier, B. (1993). Theorie und Therapie psycholinguistischer Faktoren. In H.S.Johannsen & H. Schulze (Eds.), Praxis der Beratung und Therapie bei kindlichem Stotter n. Werkstattbericht. (pp. 52-104). Ulm: Phoniatrische Ambulanz.

Rommel, D., Hage, A., Kalehne, P., & Johannsen, H. S. (2000a). Development, maintenance, and recovery of childhood stuttering. Prospective longitudinal data 3 years after first contact. 414 Theory, research and therapy in fluency disorders

In K.L.Baker, L. Rustin, & F. Cook (Eds.), Proceedings of the Fifth Oxford Dysfluency Conference (pp. 168-182). Leicester, UK: Kevin L. Baker.

Rommel, D., H'age, A., Kalehne, P., & Johannsen, H. S. (2000b). Die Entwicklung des Stotterns im Vorschulalter. Teilergebnisse der Ulmer prospektiven Langsschnittstudie nach Ende der 4 1/2-j.";ihrigen Beobachtungsphase. Forschungsbericht der Phoniatrischen Ambulanz der Universita't Ulm, Ulm, 68.

Ryan, B. P. (2001). A longitudinal study of articulation, language, rate, and fluency of 22 preschool children who stutter. Journal of Fluency Disorders, 26, 107-127.

Watkins, R. V., Yairi, E., & Ambrose, N. G. (1999). Early childhood stuttering 111: Initial status of expressive language abilities. Journal of Speech, Language, and Hearing Research, 42, 1125- 1135.

Yairi, E., Ambrose, N., Paden, E. P., & Throneburg, R. N. (1996). Predictive factors of persistence and recovery: Pathways of childhood stuttering. Journal of Communication Disorders, 29, 51-77.


In preparation for the 2018 World Congress the IFA is implementing Japanese translations of some pages on the site. 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

Follow the Joint World Congress