The current paper criticizes the concept, research methodology, data analyses, and validity of the conclusions made in Hardy, Woodman, and Carrington’s (2004) article published in this journal. In their repeated-measures analysis of data from the performances of 7 golfers, they did not examine changes in performance scores on successive holes. Instead, Hardy et al. used several ANOVA models to examine how performance varied with respect to somatic and cognitive anxiety level and self-confidence interaction. By doing so, their findings produced effects which we argue to be conceptually and empirically limited. We also address problems associated with dichotomization of continuous variables, measurement errors when splitting data, eradication of random significant effects, cell sizes in segmental quadrant analysis, and correlation between somatic and cognitive anxiety. We believe these difficulties prevent any reliable conclusions and/or generalizations from being made.
The authors are with the College of Education, 307 Stone Bldg., Florida State University, Tallahassee, FL 32306.