A multi-year study of the reliability of published social & behavioral science research has found that only about half of published results are replicated by new studies.