The study found evidence consistent with unconscious priming in a direct, person-by-person test. In three of six participants, the indirect task had an advantage at short prime-mask intervals, but the interpretation holds only if the awareness task was at least as sensitive as the priming task. That condition matters because the result does not establish that consciously available information was absent from the indirect task.
Participants completed a direct prime-discrimination task and an indirect priming task. In the direct task, they reported the direction of the prime. In the indirect task, researchers measured how quickly they responded to the mask, comparing trials in which the prime and mask pointed in the same direction with trials in which they conflicted. The question was framed at the single-subject level, using a Bayesian framework across multiple experimental sessions.
A test built around repeated sessions
Six volunteers were recruited from the Free University Amsterdam lab pool. The group included one man and five women, with a mean age of 19.8 years and a reported standard deviation of 1.33 years. No participants were excluded, and the experiment was approved by the university's Ethical Review Board. The conclusions therefore rest on data from these six volunteers.
Each trial used a left- or right-pointing arrow as the prime, followed by a mask. The prime appeared for 12.5 milliseconds and the mask for 125 milliseconds. Researchers varied the delay between their onsets, known as the stimulus-onset asynchrony, across 12.5, 25, 37.5, 50, 62.5 and 75 milliseconds. A 300-millisecond condition was added from session two.
After practice or a warm-up, later sessions alternated blocks in which participants judged the prime and blocks in which they responded to the mask. Data collection used Bayesian sequential stopping, meaning sessions continued until the analysis produced strong evidence for an effect or for no effect at at least four of the six main delays in both tasks. The stopping rule was applied at the individual-participant level.
The first session was deliberately different. It used little to no training and was collected to recreate the original data pattern. Because it contained fewer trials, it was treated as a control-like replication rather than the source of the main statistical conclusions. That session broadly reproduced the original qualitative pattern.
The group result hid a split between tasks
When the later, modified experiment was considered at group level, participants identified prime direction above chance overall, and their accuracy rose as the prime-mask gap lengthened. The awareness effects were smaller than the priming effects. The same data showed longer mask reaction times when the prime and mask conflicted than when they matched, with the priming difference becoming more pronounced at longer delays.
That group picture was not a good description of everyone. Some participants performed above chance on prime discrimination across most or all delays, while others were at chance or near chance across many of them. Priming was more uniform: all six participants showed the same broad profile, with no priming effect at the shortest delay and gradually stronger priming as the delay increased.
The crucial comparison
To compare awareness and priming on one scale, the researchers converted reaction times from the mask task into a binary accuracy measure using a median split. In plain terms, responses were divided into faster and slower-than-median outcomes before being compared with correct and incorrect prime reports. The approach made the direct and indirect tasks comparable within one statistical test.
In the pooled comparison, the two tasks did not differ at the shortest delay. The indirect measure was higher at the remaining SOAs, while the direct awareness measure was higher at 300 milliseconds. Subject-level and group-level summaries also reported an indirect-task advantage across all subjects and at multiple SOAs. The authors interpret this pattern as strengthening the case for unconscious priming only under the assumption that the direct awareness task was at least as sensitive as the indirect task.
The pattern also sets a boundary on what the experiment can say. It does not show that awareness was absent at longer delays, since direct prime discrimination increased with SOA. Nor does it establish unconscious priming for every participant or every delay. The initial session's lower trial count further means its replication pattern was supportive context, not a full test of the later analysis.
At the same time, the evidence is limited to six volunteers performing one masked-arrow laboratory task. The large gap in awareness between participants remains an open question, and the central interpretation remains conditional on the relative sensitivity of the two tasks. The result is therefore a focused behavioral finding from one specific design, rather than a general demonstration that unconscious priming occurs in every setting.
Paper data and sources
Original title: Evidence for unconscious priming using a Bayesian single-subject approach.
Authors: Nicolás Sánchez-Fuenzalida, Simon van Gaal, Zazie van den Hurk et al.
Journal/Repository: Neuroscience of consciousness
Status: Peer-reviewed
First online: 2026-08-20
DOI: 10.1093/nc/niag052
Original paper