Article | Comment by Per Alm | Response by Karrass, Conture and Walden

A new study has shown that young children who stammer may be more emotionally sensitive, and have greater difficulty regulating their feelings, than children who don't stammer. Findings suggest that these may contribute to the difficulties these children have establishing reasonably fluent speech and language.

The research team was led by Edward Conture, who said parents should not feel guilty about any role emotional control might play in stuttering. "What we're trying to find out is how emotions contribute to the onset and development of stuttering in children," Edward Conture told The Washington Post newspaper.

"There is no evidence that parents cause their children to stutter," Conture said. "The current findings tell us this is something we should look at."

The study involved 111 preschoolers between 3 and 5 years. Parents of 65 children who stutter completed a standardised and widely used behavioral questionnaire about their child's reactions to various situations and the ease with which they refocused their attention. Parents of 56 children who do not stutter completed the same inventory. Researchers assessed all children's language abilities.

Conture said that an analysis of the two groups yielded statistically significant differences in three areas: stutterers being more reactive than non-stutterers, and less able to regulate their reactions to everyday situations. They were also less able to refocus their attention when aroused - such as if another child grabbed a toy they were playing with - and were more likely to become fixated on the situation.

Conture said future studies will measure those differences more precisely using brain wave monitoring and other measurements that do not reply on reports from parents.

Barry Guitar, a professor of communication sciences at the University of Vermont, said he thought the study, at Vanderbilt University, might help answer one of the questions he hears most often as a speech pathologist: "Why does the problem ebb and flow unpredictably?"

Relation of emotional reactivity and regulation to childhood stuttering is published in the Journal of Communication Disorders Volume 39, Issue 6 doi:10.1016/j.jcomdis.2005.12.004

From the Autumn 2006 edition of Speaking Out, page 12

Comment by Per Alm, PhD, Uppsala University, Sweden, July 2008:
Error in figures, and group difference not clear.
In the original version this article from Speaking Out stated that the stuttering group was about 25 percent more emotionally reactive than non-stutterers (figures now deleted by the editor of Speaking Out). However, this was an error caused by a newspaper interview. The correct figure from this investigation was that the stuttering group, on a six-point scale, was about 0.25 steps more reactive than the non-stuttering group, which made 6.8 % group difference. An additional problem is that the two groups were not matched for gender, with a substantially higher percentage of boys in the stuttering group, which may have influenced the result. The difference in reactivity mainly was found in the 4-year age group, which also was the group with the largest gender difference: Among the stuttering 4-year-olds there were 7.3 boys for each girl, to be compared with 1.7 boys for each girl in the non-stuttering group. In comparison, the group of stuttering 5-year-olds actually was reported to be somewhat less emotionally reactive than the non-stuttering 5-year-olds. Therefore it seems likely that the real group difference, corrected for gender, was considerably lower than 6.8%, or non-existent.

Response by Jan Karrass, Edward Conture and Tedra Walden, Vanderbilt University, September 2008
Re: Karrass et al (2006) article, entitled, 'Relation of emotional reactivity to childhood stuttering.'
We thank you for the opportunity to reply to concerns recently expressed by Dr. Per Alm about our study (Karrass, J., Walden, T., Conture, E., Graham, C., Arnold, H., Hartfield, K., & Schwenk, K. (2006). Relation of emotional reactivity and regulation to childhood stuttering. Journal of Communication Disorders. 39, 402-423.) Some of Dr. Alm's concerns arise from differences between a popular press account and our published article, differences that have created some confusion about the actual findings. We assume full responsibility for contributing to that confusion and attempt in this letter to clarify the matter.

Basically, Dr. Alm expressed two concerns with Karrass et al's (2006) study of possible emotional contributions to childhood stuttering: (1) Data published in Karrass et al do not support the twenty-five percent difference in reactivity between children who do (CWS) and children who do not stutter (CWNS) reported in the popular press and (2) Karrass et al did not adequately, clearly or fully explore gender and age effects with our data, that is, perhaps gender and/or age significantly influenced our findings. Immediately below we will address each of these concerns; however, in doing so we will focus on the actual published article rather than the popular press characterizations of that article. Hopefully this coverage will help the BSA readership gain a clearer understanding of our procedures, findings and interpretation.

Concern #1:
Data published in Karrass et al do not support the twenty-five percent difference in reactivity between children who stutter (CWS) and children who do not stutter (CWNS) reported in the popular press.

Our response:
The third author (Conture) mistakenly expressed to the popular press that CWS are 25% more emotionally reactive than CWNS. This was in error, and Conture assumes full responsibility for its occurrence. Nonetheless, no such claim, statement or suggestion was made in our original published paper. Indeed, we measured differences between CWS and CWNS using a six-point scale that has no absolute zero point. Children's scores could only range from a low of 1 to a high of 7. With no absolute zero point, it is inappropriate to talk about percent increases or decreases.

Importantly, as noted immediately above, our six-point scale lacked a zero (0) score because one cannot have absolutely no level of activity, for example; one can only be higher or lower in comparison to others. For example, 100°F is not twice as hot as 50°F (i.e., when converted into Celsius 38°C is not twice is high as 10°C), and 0°F indicates cold weather, not an absence of temperature. As such, the calculated percentage of 6.8% mentioned in the comment published in July 2008 also has no meaning. Additionally, talking about "steps more reactive" would not be meaningful, as there is no practical equivalent or interpretation of what or how large a "step" is (i.e., there is no uniform underlying metric).

One can, however, evaluate the strength of the finding that CWS were more emotionally reactive than CWNS by calculating the effect size, that is, the size of the difference between the two groups. One advantage of effect sizes is that they are comparable across studies, if one so desires, with one widely-used measure of effect size being Cohen's d. When this measure was applied to our data, results indicated a Cohen's d = .374, which is between a small (.20) and a medium (.50) effect according to Cohen's (1992) system of evaluating effect sizes.

Concern #2:
Karrass et al (2006) did not adequately, clearly or fully explore gender and age effects with our data, that is, perhaps gender and/or age significantly influenced the findings of Karrass et al.

Our response:
The original Karrass et al. (2006) article reported tests for gender and age effects on the group differences in emotional reactivity, emotion regulation and attention regulation, and none of those effects were statistically significant. For example, results of our regression analysis indicated that the effects of age (in months) on the relation between talker group and reactivity was close to zero, ß = .02, p = .8, that is, non-significant. However, we did not report in our 2006 study whether the difference found between CWS and CWNS in emotional reactivity, emotion regulation, and attention regulation was jointly impacted by age and gender.

To do that, we have subsequently tested three-way interactions between age, gender, and talker group on each of the three dependent variables (i.e., emotion reactivity, emotion regulation and attention regulation) using regression. None of these three-way (ageXgenderXtalker group) interactions were statistically significant, ß = -.06, p = .89 (reactivity), ß = .21, p = .61 (emotion regulation), and ß = .29, p = .47 (attention regulation). In essence, these subsequent three-way assessments indicate that the relations found between talker group and all three dependent variables were equivalent at every age for both genders.

Certainly, it is a possibility with this as well as any other empirical study of stuttering that gender and/or age may impact findings. However, it does not appear, based on our analyses, that the Karrass et al (2006) findings were so influenced. It is unknown, of course, whether the failure to find a significant influence of age and/or gender on Karrass et al's findings will hold up using different testing instruments and/or samples of children who do and do not stutter, a possibility that awaits further empirical study.

Again, we thank the BSA for this opportunity to respond to concerns expressed about Karrass et al (2006). While concerns with this or any other piece of empirical science are not uncommon, we believe our line of investigation - the emotional contributions to developmental stuttering - has important implications for the onset and development of stuttering. Further, we are unaware of any oral or written presentations in which we explicated or implied that emotions are the sole contributor to stuttering.

Instead, our findings can be taken to suggest that emotional processes - together with linguistic, motoric and/or neurological processes - are part of a matrix of variables that contribute to the cause of stuttering. As further scientific research brings this matrix into sharper relief, it is quite possible, as often happens, that the values of the variables in this matrix will form clusters. If that were to occur, it is hoped that such clustering would help us characterize a finite number of subgroups among people who stutter, rendering our assessment as well as treatment more effective by being more precisely targeted.