Short and sweet: Comparing strategies for the reduction of questionnaires on self-criticism and social safeness while...

Measuring psychological traits with standardised questionnaires is an essential component of clinical practice and research; however, patients and participants risk fatigue from overly long and repetitive measures. When developing the short form of a questionnaire, the most widely used method for selecting an item subset uses factor analysis loadings to identify the items most closely related to the psychological construct being measured. However, this approach will tend to select highly correlated, homogeneous items and might therefore restrict the breadth of the construct examined. In this study, we will present Yarkoni's genetic algorithm for scale reduction and compare it with the classical scale reduction method. The algorithm will be applied to the shortening of three instruments for measuring self-compassion and social safeness (two unidimensional measures and a three-factor measure). We evaluated the shortened scales using correlation with long-form scores, internal reliability and the change in the correlations observed with other related constructs. Findings suggested that the classical method preserves internal reliability, but Yarkoni's genetic algorithm better maintained correlations with other constructs. An additional qualitative assessment of item content showed that the latter method led to a more heterogeneous selection of items, better preserving the full complexity of the constructs being measured.