2003 IFA Congress: Montreal, Canada

Fluency Development and Temperament in Fluent Children and Children Who Stutter

Peter Howell1, Stephen Davis1, Hina Patel1, Paul Cuniffe1, Deborah Downing-Wilson1, James Au-Yeung1 and Roberta Williams2
1University College London, 26 Bedford Way, London WC1H 0AP
2City University, Northampton Square, London EC1V OHB


Some interim results of a longitudinal study into indicators of fluency development are reported. A variety of language tests were performed. OCDI (Oxford Communication Development Inventory) scores were obtained at about 18 months and errors in naming pictures were obtained on a group of the children around this age. MLU (mean length of utterance) and a receptive measure of syntax (ROST - Reception of Syntax Test) were obtained when the children were around three years. CD1 correlated negatively with picture naming errors and this might reflect lexical development. MLU correlated with ROST score and this might reflect syntactic development. Measures of temperament obtained at 3 years did not correlate markedly with any language measure. It is concluded that there are two language factors that affect fluency development (lexical and syntactic), neither of which is related to temperament. In older fluent children and children who stutter temperament was found to differentiate the groups (CD1, MLU and picture naming are not appropriate tests for these older children). It is concluded that temperament may differentiate children who stutter from fluent speakers independent of their language performance.

  1. Introduction
Children who stutter are typically not seen in clinics until they are about three years old. This is some time after language onset. It is important to know whether these children had problems in early language development before they first attended clinic. Though it would, in principle, be possible to assess the language of large groups of children some of whom would subsequently be found to suffer from stuttering, practically this is impossible as incidence rates of stuttering are too low. Fluent children show all the features of stuttering that children designated as stuttering do, and this has been used to argue that stuttering is placed at one end of a fluency continuum with fluent speech at the other end (the continuity hypothesis). According to this hypothesis, the factors that affect fluency development in children developing normally, would also operate in children who stutter. Language development can be examined in fluent children at ages prior to, and around, the age of stuttering onset to see what factors relate to fluency development. If measures assess the same aspect of fluency in language, then they should correlate with each other (children scoring high on fluency according to one measure should score high on other measures that assess the same dimension of fluency and so forth). In like manner, behavioral measures other than language that also influence fluency development, should correlate with the language measures. Two-year old children with expressive language delay (less than 50 words) have been found to be more shy, fearful and to have more behavioral problems than children with normal language development (Caulfield, 1989). Paul and Kellog (1997) found that a group of 6 year old children who were identified as having delayed expressive language at 2 years were more shy and less outgoing than peers with a normal language history. Cohen (1975) indicated that children with expressive language delay were more active, inattentive and difficult to manage. According to the continuity hypothesis, language factors that correlate together, and behavioral factors that correlate with them, would indicate potential sources of difficulty in fluency that, in extreme form, surfaces as stuttering. This report has two parts: The first gives details about tests made on a group of fluent children when they were between 18 months and three years on several language tests that are related to fluency development and on a questionnaire that indicates the child’s temperament. These data are examined to see which measures correlate to get some idea what factors influence fluency development around the age of stuttering onset. The second study examines whether a slightly older group of children who stutter have different temperament scores to age- and sex-matched fluent control children.

  1. Experiment 1
The question to be addressed in the analysis of this battery of language tests is whether performance on any language test correlates with performance on the other language tests. Scores from the temperament questionnaire are then used to see whether these behavioral indices relate to language performance. Language tests were selected that a) it was possible for children at the designated ages to do, and b) to cover a range of abilities that could relate to fluency development. The language tests employed were: a) A measure of vocabulary (the Oxford version of the Communication Development Inventory, OCDI, Hamilton et al., 2000). b) A test that requires the child to name pictures they know that uses number of errors in naming the pictures as the performance measure (based on Gershkoff-Stowe & Smith, 1997). c) Mean length of utterance (MLU) that is a standard measure for the development of productive syntax at these ages (Brown, 1973). d) A measure of the child’s ability to understand syntactic relations that does not require them to produce sounds (the reception of syntax test, ROST, Howell et al., 2003).

Children aged between 18 months and 3 years are developing other skills apart from language and the former skills might also impact on fluency development. The behavioral dimension of child’s temperament was assessed. This dimension is raising a lot of interest as a potential indicator of early stuttering (Anderson et al., 2003; Embrechts et al., 2000). In this study, the question is whether temperamental variables relate to factors in early language development. The Behavioral Style Questionnaire (BSQ) is used and it is a parent-report questionnaire developed by McDevitt & Carey (1978) for assessing temperament in 3-7 year-old children. The BSQ is based on large normative samples and has its theoretical foundation in the New York Longitudinal Study of Thomas and Chess. The BSQ measures temperament along nine dimensions: activity level, rhythmicity, approach-withdrawal, adaptability, intensity of reactions, quality of mood, attention- span/persistence, distractibility, and threshold of responsiveness. The temperament measures for each of these dimensions was examined in two ways a) whether each dimension correlates with any of the language measures (if the scores relate to the same underlying ability, they should correlate), b) median splits were made between each language score and each temperament dimension. For each combination of language and temperament dimension, a two by two contingency table was produced with one dimension being classification of children who score high or low on the language test and the other dimension being classification of children who score high or low on the temperament dimension. Each contingency table was then examined to see whether there was an association (by Fisher-Exact test) between the language measures and temperament dimensions (if the scores relate to the same underlying ability, there should be an association).

2.1 Method


Twenty eight children were assessed on up to four language tests and on the BSQ temperament scales. The first ten children (group N1) did not do the version of the picture naming experiment reported here. Sixteen of the remaining 18 children did picture naming. There are 112 (28 X 4) remaining assessments and all but 23 were completed (20.5% incomplete). Twenty one of these (18.8%) were from seven children whose parents provided a CD1 score. Four of these children did the reported version of the picture naming experiment. The age that tests were made on each child and gender are given in Table 1. The children were tested in their own home (usually in the morning when the child was most alert). These children were brought up in a monolingual English environment. They were recruited from birth announcements in local papers etc.


Table 1.

The Oxford Communication Development Inventory (OCDI)

Productive vocabulary data were obtained from parental report using the OCDI (Hamilton et a1., 2000). OCDI is a checklist composed of 416 words, which are among the first to appear in the vocabulary of young English-speaking children. The words are divided into the following categories - Sounds, Animals, Vehicles, Toys, Eating/Drinking, Body Parts, Furniture/Rooms, Out, House Objects, People, Games/Routines, Action Words, Descriptive Words, Question Words, Time, Pronouns, Prepositions, and Quantifiers. The OCDI was used to obtain indications of the child’s productive vocabulary. Parents recorded productive Vocabulary by filling in the form.

Picture naming

The children in group N2 were tested on picture-narning using a variant of the procedures described in Gershkoff-Stowe and Smith (1997). Gershkoff-Stowe and Smith (1997) used a fixed set of test pictures throughout their tests. The variation used here customised the pictures for each child. This was done by selecting pictures for the child to name that represented objects from the OCDI (based on words the child produced) that the child had acquired between visits made 6 and 3 weeks previously. This controlled, to some extent, for familiarity with the objects that had to be named. The pictures were down-loaded onto a laptop for presentation (Gershkoff-Stowe & Smith, 1997 used a picture book for material presentation).

Mean Length of Utterance (MLU)

Mean Length of Utterance (MLU) data were obtained from analysis of audio recordings made in the child’s homes at the time of the ROST assessment. The researcher and child interacted in an unstructured way for approximately 30 minutes. The session was recorded on a Digital Audio Tape recorder and the recording was orthographically transcribed. The first one hundred utterances were used for analysis and the MLU was calculated using the criteria presented in Brown (1973).

Reception of syntax test, ROS T

In the reception of syntax test (ROST), the child looked at two pictures displayed on the screen of a laptop computer. The child heard a sentence played from the computerthat corresponded with one of the pictures. If the child has acquired the target syntactic dimension, he or she should be able to link the picture with its sentence. This was indicated by touching the picture appropriate to the sentence (the touch response was detected automatically by the computer). Before testing, all subjects were given four training items consisting of two pictures of different objects and a recording of a single word utterance to ensure they knew how to make the responses.

The test pictures and sentences probed nine syntactic dimensions. Examples of test sentences that the child had to understand to pick the appropriate picture were: 1) SVO word order were “The boy kisses the girl”/“The girl kisses the boy”. 2)Article, “a fish”/ “some fish”. 3) Plural morpheme, “cat”/“cats”. 4) Pronoun (he/she), “He runs”/“She runs”. 5) Pronoun (his/her), “his cat”/“her cat”. 6) Pronoun (he/they), “he danced”/“they danced”. 7) Preposition (in/on), “in the box”/“on the box”. 8) Other prepositions (such as up/down), “going up”/“going down”. 9) Compound noun, “car cake”/ “cake car”, where “eat cake” was meant to signify a cake made in the shape of a car and “cake car” was meant to signify a car for delivering cakes. Each of the nine categories was tested eight times each. A child had to pass 7/8 to pass the category as this criterion gives statistical significance by a Sign test (p <.O5). (For further details see. Howell et al., 2003.) The ROST scores used here represent number of blocks passed.


Though trained observers can be used to assess temperament, the information acquired is limited to the range of behaviors that can be readily observed during a single interview. Parent reports can yield a more comprehensive View of the child’s temperamental characteristics, reflecting an intimate knowledge of the child that is unavailable to others. In addition, the use of parental input has a long history in the study of speech-language pathology and in psychology and is known to be reliable (MacArthur Communicative Development Inventories, Fenson, Dale, Reznick, Thal, Bates & Hartung 1993). The BSQ was used as a parent-report questionnaire for assessing temperament.

2.2 Results

Inter correlations between all the language tests (OCDI, picture naming, MLU and ROST) were examined. CDI correlated with picture naming. Thus the Pearson product moment correlation coefficient was 0.42 for the 16 subjects who did CD1 and picture naming (p<0.05 one tail). MLU correlated with ROST scores. Thus the Pearson product moment correlation coefficient was 0.45 for the 21 subjects who did MLU and ROST (p<0.05, one tail).

Correlations between each of the four language tests and nine temperament scales resulted in significance between ROST and rhythmicity (L: -0.58, p <0.05 one tail) and between ROST and persistence (r = 0.58, p <0.05 one tail).

When the 2 x 2 contingency tables of language by temperament variables (described above) were analyzed by Fisher-Exact tests, significant associations were obtained in the case of ROST and adaptability (chi square = 5.65, p < 0.05, one tail) and ROST and persistence (chi square = 5.65, p < 0.05, one tail).

2.3 Discussion

Two of the language variables that correlate (CD1 and picture naming) may relate to lexical or vocabulary development and the other two (MLU and ROST score) may indicate syntactic development. The fact that there are no correlations across lexical and syntactic dimensions may suggest that fluency in development of these language skills is independent in children of this age. Other explanations are also possible given that the variables that correlate were obtained around the same age (but, in the case of CD1, not at the same visit).

Temperament can be measured before a child starts to use language, and if it affects language skills, measures of temperament might offer an early indication of potential fluency problems. We have also found with a subset of our children that there is an impressive relationship between temperament variables at 15 months and 3 years (across different forms of the temperament questionnaire). Somewhat disappointingly given the promising prospects for temperament, there were only occasional significant correlations and associations with language measures.

The only significant correlations obtained were between ROST and rhythmicity and ROST and persistence. At this age, the ROST test may be a) a pure measure of syntactic ability, or b) tap the two abilities of concentrating on the task in hand as well as syntactic ability per se. If the latter is the case, a child may have developed syntax but not be capable of concentrating long enough on the task so that this ability can be assessed. The two temperament variables that correlate with ROST (rhythmicity and persistence) might be inherently linked to language ability (consistent with alternative a)), or reflect how well a child applies him or her self to a task (consistent with alternative b)). In the latter case, a child who is erratic in his or her behavior and who finds it hard to sustain attention on an activity, would perform poorly on ROST because this task makes just these demands. It is not possible at present to favor possibility b) over a), though the longitudinal nature of this study will allow these alternatives to be distinguished in the future. If temperament is directly linked to language as in a), then the correlations would be expected to hold up as the children get older. If the temperament variables are merely an indication of how well a child can concentrate on the ROST task, the correlation would be expected to disappear as age increases.

Thus, the effects of temperament on language are not dramatic (either in terms of the relationships themselves or the number of temperament variables associated with language measures). Nevertheless, children who stutter may have temperamental problems that are independent of language performance that affects whether they stutter but does not affect specific language measures. In the next study, groups of (slightly older) children who stutter and fluent children were compared on measures of temperament to see whether temperament dissociates the groups.

  1. Experiment 2
Embrechts et al. (2000) found 3 - 7;7 year-old children who stutter were less successful in adapting to their environment than their non-stuttering counterparts and that they were more easily distracted than fluent children. A recent study by Anderson et al., (2003) compared temperament of 31 children who stutter with 31 matched control children who did not stutter (aged 3-5;4) and found three differences (summarized in Table 2). Though our earlier study suggested, little language- related effect of temperament, these studies suggest that temperament might still differentiate fluency groups when temperament scores alone are compared. This study compares temperament of a group of children who stutter with an age- and seX-matched group of fluent children.

3.1 Method

Ten CWS and age-and sex-matched controls. For the children who stutter, seven were boys, three were girls and ages ranged from 3 years 7 months to 7 years 2 months with a mean of 5 years 10 months (sd 1.36 years). For the children who do not stutter, there were also seven boys and three girls with ages ranging from 3 years 8 months to 7 years 2 months and with a mean of 5 years 4 months (sd 1.16 years). The parents of these children filled out the BSQ as in the previous study.

3.2 Results

Fluency groups differed on four of the nine temperament dimensions when assessed by independent t-tests. The t values, as well as a summary of findings from the Anderson et al. (2003) and Embrechts et al. (2000) studies, are given in Table 2.


Table 2.

3.3 Discussion

In this study, there are a healthy number of temperament dimensions that differentiate children who stutter from fluent controls (4/9). However, three of these (activity, mood and persistence) were not found in either of the two previously reported studies on temperament and stuttering. Comparison of the Anderson et al.and Embrechts studies alone, also reveals discrepancies in whether variables differentiated speaker groups (rhythmicity) and in the direction of effects (distractability). In fact, the only consistent finding across all three studies is that children who stutter are not as adaptable as fluent children. This may merit more study, as well as ways of modeling how this can have an independent effect on language. However, in the next study to be made on the longitudinal group (used in study 1), we plan to take a different approach and examine whether motor factors correlate with the language measures.

This work was supported by the Wellcome Trust. Deborah Downing-Wilson was supported by a vacation research student grant from the British Stammering Association.

Anderson, J .D., Pellowski, M.W., Conture, E.G. & Kelly, E.M. (2003). Temperamental characteristics of children who stutter. Journal of Speech, Language and Hearing Research, 46, 1221-1233

Brown, R. (1973). A first language: the early stages. Cambridge, MA: Harvard University Press.

Caulfield, M. (1989). Communication difficulty: A model of the relation of language delay and behavior problems. SRCD Abstracts, 7, 212.

Cohen, D. (1975). Childhood Personality Scale. Washington, DC: NIMH

Embrechts, M., Ebben, H., Franke, P., & van de Poel, C. (2000). Temperament: A comparison between children who stutter and children who do not stutter. 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 on Fluency Disorders in Nyborg, Denmark (pp. 557- 562). Nijmegen: Nijmegen University Press.

Fenson, L., Dale, P.S., Reznick, J.S., Thal, D., Bates, E., & Hartung, J .P. (1993). The MacArthur Communicative Development Inventories: User’s Guide and Technical Manual. San Diego: Singular Publishing Group.

Gershkoff-Stowe, L. & Smith, L.B. (1997). A curvilinear trend in naming errors as a function of early vocabulary growth. Cognitive Psychology, 34, 37-71.

Hamilton, A., Plunkett, K. & Schafer, G. (2000). Infant vocabulary development assessed with a British Communication Development Inventory: Lower scores on the UK then in the USA. Journal of Child Language, 27, 309-343.

Howell, P., Davis, S. & Au-Yeung, J . (2003). Syntactic development in fluent children, children who stutter, and children who have English as an additional language. Child Language Teaching and Therapy, 19, 331-337

McDevitt S.C., & Carey, W.B. (1978). The measurement of temperament in 3-7 year old children. Journal of Child Psychology and Psychiatry and Allied Disciplines, 19, 245-253.

Paul, R., & Kellogg, L. (1997) Temperament in late talkers. Journal of Child Psychology and Psychiatry, 38, 803-811
join button


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


Journal of Fluency DisordersBrowse the current issue

The official journal of the International Fluency Association
IFA Members receive online access to JFD as a member benefit.

Read more: JFD

Twitter Feed @official_IFA

IFA on Facebook