Concluding causation from correlation: comment on Burns and Spangler (2000).

This commentary critiques the study conducted by D. D. Burns and D. Spangler (2000) in which the relationship between homework compliance and therapy outcome was estimated using structural equation modeling (SEM). Although the authors of the commentary advocate the use of SEM, they suggest greater caution in the indiscriminate endorsement of a causal relationship based on (a) correlational data and (b) retrospective accounts of the main predictor variable within a cross-sectional design. This discussion also highlights a need to address the issue of therapist competence in homework administration. The results of Burns and Spangler's report are consistent with existing empirical evidence suggesting that compliance with homework facilitates therapeutic outcome. However, the gold standard for determining causal inferences rests on prospective, experimental research rather than on retrospective, correlational models.