Further evidence for the deficit account of the test anxiety–test performance relationship from a high-stakes admission testing setting

• Published in: Intelligence (Norwood) 2015-11, Vol.53, p.72-80
• Main Authors: Sommer, Markus, Arendasy, Martin E.
• Format: Article
• Language: English
• Subjects: Admissions policies; Anxiety; Applicants; Bias; Cognitive ability; Cognitive ability tests; Correlation analysis; Deficit model; Emotionality; High-stakes setting; Interference; Interference model; Item response theory; Measurement; Measurement bias; Test anxiety; Trait anxiety; Trait test anxiety
• ISSN: 0160-2896; 1873-7935
• DOI: 10.1016/j.intell.2015.08.007

Based on meta-analytic evidence of a moderate negative correlation between test anxiety and test performance some researchers hypothesized that trait test anxiety may induce measurement bias. Two competing models have been advanced to account for the observed test anxiety–test performance relationship: the deficit hypothesis and the interference hypothesis. The interference hypothesis predicts that trait test anxiety induces measurement bias in items of intermediate difficulty, while the deficit hypothesis claims that test anxiety has no causal effect on test performance. Despite the practical relevance of this topic only few studies directly tested these predictions and none of them was conducted in real life high-stakes settings, which may limit the ecological validity of their findings. Therefore we tested these competing predictions in a high-stakes admission testing situation by means of structural equation modeling and item response theory analyses. A total of N=1768 applicants to a medical university participated in the present study. After completing four cognitive ability tests as part of the admission test respondents filled a trait test anxiety questionnaire, which measured trait worry, trait task-irrelevant thinking and trait emotionality. In line with previous findings the results indicated that none of the trait test anxiety components induced measurement bias across different levels of trait test anxiety. Thus, the present findings were most consistent with a deficit account of the test anxiety–test performance relationship. •Competing predictions deduced from the deficit and interference hypothesis were tested in a real-life high-stakes setting.•Item response theory and structural equation models were used to disentangle deficit- and interference effect.•Similar to previous studies, the results indicated measurement invariance across trait test anxiety levels.•Trait test anxiety components were more strongly related to more g-saturated tests than to less g-saturated tests.•The results were consistent with the deficit models of the test anxiety-test performance relationship.