Have you ever considered how your brain determines when you have clearly seen something and when you have not? It is possible to be shown something, such as an image, that our brain registers even if we are not consciously aware that we have seen it. Researchers have developed diverse theories that seek to explain these mechanisms of consciousness and unconsciousness, as well as the distinctions between them (Mudrik et al., 2025; Yaron et al., 2022).

Investigating this often involves recording people’s brain activity while they are shown stimuli at various perception levels and asked to report whether they have seen them or not. The recordings are then analysed to identify brain activity patterns associated with unconscious or conscious perception through a process known as “neural decoding”. However, some participants may report having seen a stimulus only when they are absolutely certain they have detected it, while others may report it even if they are not fully confident. This highly subjective threshold is known as the judgement criterion (Green and Swets, 1966). As it varies among individuals, it can introduce bias into experimental data, which often complicates studies of conscious experiences. Now, in eLife, Johannes Fahrenfort (Free University of Amsterdam) and colleagues in Leiden, Hamburg and Amsterdam report that differences in the judgment criterion have distinct effects on how neural signals associated with conscious and unconscious experiences are interpreted (Fahrenfort et al., 2024).

Fahrenfort et al. began by using a simple computational model to simulate how neural decoding might be influenced by differences in the judgement criterion. The team simulated two different scenarios. In the liberal condition, participants reported seeing a target stimulus even when the signal was weak or absent. On the other hand, in the conservative condition participants only reported seeing the stimulus when the signal was strong, causing them to report “seeing” it less often (Figure 1A). Analysis showed that the conservative condition led to higher neural decoding performance than the liberal condition, with brain activity patterns associated with a target stimulus being present or not more easily distinguished (Figure 1B). This apparent enhancement in neural decoding accuracy is a result of the conservative reporting itself rather than any change in the stimulus, which remained the same across the different conditions.

Figure 1

Download asset Open asset

To validate these findings, Fahrenfort et al. next compared the simulated results with two comparable experimental datasets. In the first dataset (which was reported by Kloosterman et al., 2019; Kloosterman et al., 2020), the experimental model matched the one used in the simulated experiments, with participants reporting on whether a visual stimulus was either "seen" or "unseen". In the second dataset, which was collected by Fahrenfort et al., participants used the Perceptual Awareness Scale (PAS; see Ramsøy and Overgaard, 2004) to report how clearly they saw the visual stimulus, with 0 indicating no conscious experience (similar to "unseen"), while 1, 2, and 3 represented the visual stimulus being increasingly clearer to see. Using this scale should make reporting more precise, as a value of 0 should only be given when the stimulus is truly “unseen”. In both experiments, researchers used feedback to shift participants' judgment criterion, successfully prompting them to respond more liberally or conservatively during the task. Fahrenfort et al. analyzed how neural decoding performance was influenced by these shifts in the judgement criterion. As in the simulations, adopting a conservative approach to reporting led to increased neural decoding performance. However, this effect was only observed in the “unseen” condition in the first dataset and the “seen” condition in the second dataset.

To better understand this inconsistency, Fahrenfort et al. used further computational modelling to show that neural decoding performance is not only affected by the judgement criterion but also by a participant’s signal sensitivity, which is how well they are able to detect a stimulus regardless of their judgement criterion. This can be measured by comparing how often a participant correctly detects a visual stimulus with how often they mistakenly report seeing one, revealing their actual perceptual ability. When participant sensitivity was low, a more conservative judgement criterion significantly inflated the neural decoding performance in the “seen” condition. However, when participant sensitivity was high, a conservative bias significantly inflated the neural decoding accuracy under the “unseen” condition (Figure 1B).

Taken together, these findings confirm the long-held prediction that subjective reporting biases can influence how well conscious and unconscious experiences can be distinguished from one another experimentally. This highlights the importance of carefully controlling the effects of the judgement criterion in future studies.