One approach to plagiarism prevention focuses on improving students' authorial identity, but work in this area depends on robust measures. This paper presents the development of a psychometrically robust measure of authorial identity -the Student Attitudes and Beliefs about Authorship Scale. In the item generation phase, a pool of items was developed and assessed for content validity by subject matter experts. In the exploratory phase, data from 439 higher education students were used to identify a latent variable model with three factors: 'authorial confidence', 'valuing writing' and 'identification with author'. In the confirmatory phase, data from 306 higher education students were used to test the three-factor model's reliability and validity. The three-factor structure was confirmed, and the results showed the SABAS has a stronger psychometric basis than previously available measures. This measure of authorial identity can be used with confidence in research and pedagogy to help students improve their authorial identity.Keywords: plagiarism; authorial identity; academic writing; psychometrics; pedagogy
IntroductionOne approach to reducing unintentional plagiarism is to improve students' authorial identity, so that they understand the role of the author better, and take a more authorial role in the production of their university assignments. This approach has been adopted in psychology (Elander et al. 2010; Kinder and Elander 2012), accounting (Ballantine and Larres, 2012;Ballantine, Guo, and Larres 2013) and health (Maguire, Reynolds, and Delahunt, 2013). All of those studies used the only available questionnaire measure of authorial identity, the Student Authorship Questionnaire (SAQ) (Pittam et al. 2009).This 18-item questionnaire has three scales measuring attitudes to authorship ('confidence in writing', 'understanding authorship', and 'knowledge to avoid 4 plagiarism') and three measuring approaches to writing ('top-down', 'bottom-up' and 'pragmatic').However, the SAQ has substantial psychometric limitations. First, its content validity is questionable, because the items were not systematically generated.Furthermore, the approaches to writing scales are not part of the core authorial identity construct, whereas other important aspects of authorial identity may have been omitted. Second, the six-factor structure may not be valid, for the Eigenvalue-over-one rule that was used to extract factors has been heavily criticised by measurement theorists (e.g., Zwick and Velicer 1986;Hayton, Allen, and Scarpello 2004;Velicer, Eaton, and Fava 2000). There were also multiple cross-loadings (i.e., items that loaded strongly onto more than one factor in the model), and items were only interpreted according to their highest loading, when in one case the difference between loadings was only .03 (Pittam et al., 2009). Third, internal consistency was poor, with Cronbach's (1951) alphas for three scales ranging from poor to moderate, and three other scales with only two items each where Cronbach's alpha was not report...