Research Design Issues in Relationships Between Anxiety and Stuttering Comments on Craig Letter to the Editor
Letter to the Editor  |   October 01, 1991
Research Design Issues in Relationships Between Anxiety and Stuttering
 
Author Affiliations & Notes
  • Joseph S. Attanasio
    Montclair State College Upper Montclair, NJ
Article Information
Speech, Voice & Prosodic Disorders / Fluency Disorders / Research Issues, Methods & Evidence-Based Practice / Speech / Letters to the Editor
Letter to the Editor   |   October 01, 1991
Research Design Issues in Relationships Between Anxiety and Stuttering
Journal of Speech, Language, and Hearing Research, October 1991, Vol. 34, 1079-1080. doi:10.1044/jshr.3405.1079
History: Received June 4, 1990 , Accepted September 26, 1990
 
Journal of Speech, Language, and Hearing Research, October 1991, Vol. 34, 1079-1080. doi:10.1044/jshr.3405.1079
History: Received June 4, 1990; Accepted September 26, 1990
In his article, “An Investigation Into the Relationships Between Anxiety and Stuttering” (1990), Craig suggested certain cautions concerning the interpretation of the results of his study. I would like to add several other cautions.
Craig suggested that previous studies on self-report measures of anxiety may have failed to uncover significant differences between stutterers and nonstuttering controls because those studies used small sample sizes. The use of small sample sizes, he stated, may have resulted in insufficient statistical power to reject the null hypothesis. In his study, Craig used relatively large samples (102 stutterers and 102 nonstuttering matched controls). Differences in trait and state anxiety between the two groups reached statistical significance. Now, as Bakan (1968)  has pointed out, it is easier to reject the null hypothesis with large rather than with small sample sizes because large samples provide the opportunity for the appearance of variables with even low levels of impact on deviations from null. The corollary is that in small samples a variable that surfaces as having an impact on deviation from null is most likely stronger in its impact than one that requires large samples before its effects become apparent. The question, then, is how important or significant in the nonstatistical sense is a variable that requires the use of large sample sizes before its effects surface? An estimate of practical significance can be achieved by computing γ (gamma), which is a measure of the magnitude of the population effect size (Welkowitz, Ewen, & Cohen, 1976).
First Page Preview
First page PDF preview
First page PDF preview ×
View Large
Order a Subscription
Pay Per View
Entire Journal of Speech, Language, and Hearing Research content & archive
24-hour access
This Article
24-hour access