Psychometric evaluation of the Student Authorship Questionnaire: a confirmatory factor analysis approach

Abstract

This research provides new insights into the measurement of students' authorial identity and its potential for minimising the incidence of unintentional plagiarism by providing evidence about the psychometric properties of the Student Authorship Questionnaire (SAQ). Exploratory and confirmatory factor analyses (EFA and CFA) are employed to investigate the measurement properties of the scales which comprise the SAQ using data collected from accounting students. The results provide limited psychometric support in favour of the factorial structure of the SAQ and raise a number of questions regarding the instrument's robustness and generalisability across disciplines. An alternative model derived from the EFA outperforms the SAQ model with regard to its psychometric properties. Explanations for these findings are proffered and avenues for future research suggested.

Keywords:

• authorial identity
• unintentional plagiarism
• SAQ
• discipline
• voice

Notes

1. Swales (1990) describes a discourse community as a group which uses communication to achieve its common goals.

2. Drawing on the judgements of academic faculty members and subsequent analysis, Biglan (1973b) used multidimensional scaling to develop a taxonomy which classifies academic disciplines along three dimensions, namely [1] ‘the existence of a single paradigm (hard-soft) … [2] concern with practical application (pure-applied) … and [3] concern with life systems [life-non-life]’ (207).

3. A copy of the questionnaire is available on request from the authors.

4. An explanation for the two unsatisfactory alpha coefficients is that the associated factors comprise too few items (Hinkin 1995).

5. In Pittam et al.'s (2009) study Cronbach's alpha coefficients were only reported for the factors with more than two items: .69 for confidence in writing, .62 for knowledge to avoid plagiarism and .46 for pragmatic approach to writing.

6. A cut-off value of .90 is used for CFI, and the combinational rule of RMSEA < .06 and SRMR < .08 is used as it is ‘extremely sensitive in detecting models with mis-specified factor co-variances’ (Hu and Bentler 1999, 26).