When and How to Interpret Null Results in NIBS: A Taxonomy Based on Prior Expectations and Experimental Design

Abstract

Experiments often challenge the null hypothesis that an intervention, for instance application of non-invasive brain stimulation (NIBS), has no effect on an outcome measure. In conventional statistics, a positive result rejects that hypothesis, but a null result is meaningless. Informally, however, researchers often do find null results meaningful to a greater or lesser extent. We present a model to guide interpretation of null results in NIBS research. Along a “gradient of surprise,” from Replication nulls through Exploration nulls to Hypothesized nulls, null results can be less or more surprising in the context of prior expectations, research, and theory. This influences to what extent we should credit a null result in this greater context. Orthogonal to this, experimental design choices create a “gradient of interpretability,” along which null results of an experiment, considered in isolation, become more informative. This is determined by target localization procedure, neural efficacy checks, and power and effect size evaluations. Along the latter gradient, we concretely propose three “levels of null evidence.” With caveats, these proposed levels C, B, and A, classify how informative an empirical null result is along concrete criteria. Lastly, to further inform, and help formalize, the inferences drawn from null results, Bayesian statistics can be employed. We discuss how this increasingly common alternative to traditional frequentist inference does allow quantification of the support for the null hypothesis, relative to support for the alternative hypothesis. It is our hope that these considerations can contribute to the ongoing effort to disseminate null findings alongside positive results to promote transparency and reduce publication bias.