Association for Behavior Analysis International

To date, mainstream behavior analysis has not been tainted by the so-called “replication crisis” that has embroiled much of the behavioral and social sciences. In this symposium, we summarize key elements of the replication crisis, identify the methods of behavior analysts that have tended to insulate our field from the crisis, describe trends in published research suggesting that behavior analysts are drifting away from these methods, consider the professional contingencies (for example, from granting agencies) that support these trends, and explore the implications of these trends for the long-term vitality of our field. We suggest that our field’s success may be attributed to its longstanding embrace of rigorous single-case experimental methods, and we caution against the adoption of group experiments and correlational studies that rely on statistical inference, especially those that survey subjects about their potential behavior in hypothetical situations instead of observing what subjects actually do under controlled environmental conditions.

Translational research on neuro-behavioral disorders (e.g., schizophrenia; Autism) is in a replication crisis. Rather than failing to replicate results across experiments, positive results from laboratory models fail to reproduce similar results in clinical trials with humans. This problem is so severe that no novel treatment for any behavior disorder has come from animal models in the last 20 years. In this replication crisis, the fault lies less in the design and statistical inference of the laboratory research than it does in the assumptions made about human and animal behavior, and the failure to engage in proper cross species replication at intermediate stages of treatment development and testing. This paper will argue that the fault lies largely in the "behavioral assays" used to test animal models. These rely largely on formal similarity of behaviors across species (face validity) rather than functional similarity of the behavioral process operating in the laboratory and clinical settings. Using single-subject designs, behavior analysis readily replicates results across animal and human subjects because it stresses functional models (similarity in subject's reactions to common environmental variables) rather than formal models. This emphasis of cross species replicability can be of tremendous use to translational research to produce novel treatments.