Abstract: Despite considerable academic interest in the construct of statistics anxiety, we know little about how it operates, or even what it is. Statistics anxiety is said to be distinct from mathematics anxiety, but existing evidence suffers methodological limitations and relies heavily upon correlations that cannot take subscale specificity into account. Yet, if statistics and mathematics anxieties are the same constructs (i.e., have fallen prey to the jangle fallacy), the statistics anxiety field could draw upon this advanced knowledge and extrapolate considerable empirical and theoretical insights. The present study addresses this, examining the extent to which statistics anxiety (as measured by the Statistics Anxiety Rating Scale; STARS) and mathematics anxiety (as measured by the Revised Mathematics Anxiety Rating Scale; R-MARS) overlap. Across three core analyses, each conducted in two samples (n = 489 and n = 245), we consistently evidence strong construct overlap. The factor analyses showed items from the scales did not load onto statistics or mathematics anxiety specific factors, the latent profile analyses showed it was very rare (< 2%) for an individual to have statistics anxiety but not mathematics anxiety, and our experimental studies revealed that individuals with statistics anxiety did not experience a greater increase in state anxiety when taking a statistics test than a mathematics test, or vice-versa. Counter to the prevailing narrative, our results suggest a jangle fallacy, but we advocate for further research to fully understand what is driving responses to STARS items. 

Abstract: The Statistics Anxiety Rating Scale (STARS) is a popular self-report measure of statistics anxiety. It is commonly used to explore the role of statistics anxiety in the learning and teaching of statistics, but lacks construct validity evidence, meaning we cannot be sure which construct(s) the STARS is capturing. It is particularly unclear whether the STARS measures anything distinct from mathematics anxiety, a construct that overlaps considerably with statistics anxiety. The present study addresses this gap, examining the extent to which the STARS overlaps with the Revised Mathematics Anxiety Rating Scale (R-MARS). Using a large (N = 5,885) sample of undergraduate students from 83 universities across 33 countries, we demonstrate strong overlap according to four criteria. First, the scales were strongly correlated at both the aggregate level and within each university. Second, factor analyses revealed items from the scales did not cluster into statistics or mathematics anxiety specific factors (with the possible exception of those related to arithmetic). Third, the scales were related to other variables in statistically equivalent ways. Fourth, the scales could not account for meaningfully different amounts of variance in a range of outcome variables. We conclude that the STARS and R-MARS do not measure what they intend to and call for the development of more precise construct definitions and measurement scales that map onto those definitions.