Subjective scoring of divergent thinking: Examining the reliability

The present research examined the reliability of three types of divergent thinking tasks (unusual uses, instances, consequences/implications) and two types of subjective scoring (an average across all responses vs. the responses people chose as their top-two responses) within a latent variable framework, using the maximal-reliability H statistic. Overall, the unusual uses tasks performed the best for both scoring types, the instances tasks had less reliable scores, and the consequences tasks had poor reliability and convergence problems. The discussion considers implications for test users, differences between average scoring and top-two scoring, and the problem of whether divergent thinking tasks are interchangeable.