Is it possible to make unconscious inferences? The limits of the influence of unconsciously perceived stimuli on conscious decisions.

We often assume that our interactions with the world are primarily driven by conscious experiences. However, according to numerous experimental and medical reports, stimuli that are not consciously perceived can exert an influence on behaviour. For example, during subliminal semantic priming, an unconsciously presented prime can influence reaction times of subsequently presented target words. One of the open questions in the field of consciousness research is to what extent unconscious perception can influence our behaviour. A seemingly good strategy to approach this problem is to compare the complexity of tasks in existing experiments and see at what level of complexity unconscious perception effects disappear. However, various methodologies are used to investigate unconscious perception (e.g., different methods of rendering stimuli subliminal), which makes it difficult to draw conclusions from such comparisons.The purpose of our study was to address this issue and investigate the limits of unconscious perception. To achieve this, we designed an experiment in which participants were making various decisions based on information that was presented unconsciously with the use of the sandwich-masking technique. The experiment contained three conditions in which the same stimulus presentation parameters were used, but which differed in the complexity of the decisions (simple decisions without a context, medium decisions with a context, and complex decisions with a context and inference). Essentially, the study compared the impact of unconscious perception across varying levels of difficulty while keeping the methodology and stimuli presentation settings constant.We tested 40 participants. With the increase of complexity of decisions, we observed a decrease in decisions accuracy. Simple decisions had the highest accuracy rate (58%), which was significantly above the random level of answers (50%). According to Bayesian analysis, the evidence for the difference between the random level and the accuracy obtained was extreme. A lower accuracy was observed in the medium condition (54%; the Bayes factor suggested substantial evidence for the difference between the obtained accuracy and the random level of the answers). We found no evidence for a difference between random answer level and the accuracy in the complex condition, and anecdotal evidence for the opposite hypothesis. In a separate procedure, all the participants performed the objective tests of awareness. The average visibility score in the objective awareness test was 50.08%. The Bayesian factor suggested substantial evidence that the average score in the awareness test did not differ from the random level. The observed changes in effects across decision complexity, combined with random results in the awareness test, indicate that unconsciously perceived stimuli do not influence decisions based on logical inference.