Without an experiment, one can never be sure of the origin of the observed relationships. Correlation is different from causation.