Accurately Measuring Nonconscious Processing Using a Generative Bayesian Framework

Despite considerable interest in subliminal effects, fundamental questions about the proper way of examining them remain unanswered, sowing doubts regarding the veracity of published results. A central question is whether observed effects result from nonconscious processing rather than from some stimuli being consciously perceived by participants which are missed due to error in the awareness measurement. Here, we suggest a solution that implements a Bayesian modeling approach to measure the behavioral effects due to nonconscious stimuli accurately. Our solution relies on a Bayesian estimate of the correlation between variables that accounts for measurement error by modeling their uncertainty. We use simulations to show that this method accurately estimates nonconscious effects. The method we suggest is easy to use, and we describe its implementation on a relevant data set. We recommend that researchers in the field of nonconscious processing employ this method to estimate the effects of nonconscious processing more accurately.