Consciously detecting and recognizing a past visual word after its sensory trace is gone

Background:
Conscious perception refers to the subjective experience of sensory information, the ability to know and report what one is perceiving at a given moment. It reflects not just sensory processing but awareness of that information. A broad consensus holds that the initial feedforward sweep of sensory processing – a fast, bottom-up flow of neural activity from sensory receptors (like the retina) through the cortex within ~100–150 ms occurs without awareness. During this stage, the brain extracts basic features such as edges, motion, or sound frequency, but this processing is automatic and not yet accessible to conscious perception.

In practice, it has been difficult to clearly separate consciousness from unconscious perception at these early stages. The study discussed here tests whether a briefly seen (and masked) stimulus can still leave behind a hidden meaning (semantic information, i.e., what the stimulus represents, like the concept of a word or object), even if its visual details fade away. It shows that a later cue, if it is related in meaning, can help bring this information into awareness. This suggests that people can access the meaning of a stimulus even when they are not aware of its sensory details. Building on this idea, the preprint tests a stronger hypothesis: that even when the visual details of a masked stimulus are lost, its meaning (semantic information) can still remain and be brought into awareness by a related cue presented later. In other words, it asks whether people can become aware of what something is without seeing how it looked. This directly challenges the idea that consciousness depends only on ongoing sensory signals, and instead suggests that perception is flexible over time, where earlier unconscious information can later become consciously accessible through broader brain processes.

Figure 1. Schematic depicting the preprint’s hypothesis – masking erases a word’s low-level features, but a congruent retro-cue can reactivate its semantic representation, allowing conscious access to meaning without the original sensory details. (refer to Preprint Figure 1, made available under a CC-BY-NC-ND 4.0 International license.)

Key findings:

• Congruent retro-cues improved participant’s identification and sensitivity to detect the presence of the words.

One of the first experiments in this study tested a key question: can awareness of a stimulus be triggered after it has already disappeared? To examine this, participants viewed rapid visual masks in which a word briefly appeared for just 12–48 ms, close to the limit of visibility. Some trials had no word, allowing the researchers to measure detection accurately. Importantly, about 200 ms after the word disappeared, participants heard a retro-cue (a cue presented after the stimulus, often related in meaning) that was either related or unrelated to the word. They then reported the word’s identity, its visual form (upper/lower case), and how clearly they saw it.

For comparison, in other conditions researchers can use pre-cues (cues presented before the stimulus) to guide attention in advance. By using many different word pairs and controlling for guessing, the study ensured that any improvement in performance was due to the cue itself, not simple response strategies.

The results from this experiment revealed a striking dissociation. When the retro-cue was congruent, participants were better at identifying the word and detecting its presence but not at reporting its visual features. This pattern could not be explained by partial awareness of letters or orthographic fragments, as multiple control analyses ruled out these alternatives. Instead, the findings suggest that the cue can reactivate a latent semantic trace, allowing access to meaning

even when the underlying sensory representation has faded. In this way, the experiment provides strong evidence that conscious perception is not fixed at the moment of stimulus presentation, but can emerge later through top-down processes that selectively recover high-level information while leaving low-level details behind.

Figure 1: In the trials, participants were presented with one stream of centered stimuli. Response screens asked for word identification, case discrimination and overall visibility rating, in this order. Case discrimination performance was quantified when word identification is correct, with a specific estimate for retro-perceived trials. Error bars represent 95%  bootstrapped confidence intervals. (refer to Preprint Figure 2, made available under a CC-BY-NC-ND 4.0 International license.)

• Matching cues shown after a word can reveal its meaning

Building on the previous results, the authors clarify the link between visibility and confidence. Participants rated confidence separately for word identification and case discrimination. Congruent retro-cues improved word identification and confidence for that task, but case discrimination remained unchanged. This reveals a dissociation between high-level and low-level features, that is, participants could access the meaning of the word (semantic information) without recovering its visual details (like uppercase vs lowercase).

• Both pre- and retro-cues improved detection, but only pre-cues improved recognition of visual details.

Figure 2: Detection sensitivity (a) and case discrimination performance (b) are shown for pre-cueing (left panels) and retro-cueing (right panels). (refer to Preprint Figure 5, made available under a CC-BY-NC-ND 4.0 International license.)

So far, it is shown that the participants  knew what the word was, but not how it looked. This suggests that retro-cues enhance access to meaning without improving sensory representations, and that better detection is not due to misplaced confidence about visual details. This dissociation was further explored by testing whether it is unique to retro-cues or also occurs with pre-cues. Four preregistered experiments compared pre-cues and retro-cues with neutral cues, focusing participants on letter-case and visibility rather than word meaning. Both cue types were found to improve detection sensitivity, but only pre-cues enhanced case discrimination. Retro-cues left low-level visual performance unchanged, confirming that the dissociation between semantic access and sensory details is specific to retro-cueing. Detection improvements with retro-cues were smaller than in earlier experiments, likely due to task differences or reduced interference from incongruent cues.

What I like about the preprint?

The authors have done remarkable work that breaks the perception of seeing is equal to detailed perception. This preprint shows you can consciously extract semantic content without the need for sensory format. I especially liked the section on how the brain can reconstruct awareness thereby challenging the feed forward models. These findings could possibly have implications for hallucinations and predictive coding frameworks. 

Questions to the authors:

1. How was the length of time that the visual stimulus is presented determined, including the optimal delay between the auditory and visual stimuli?
2. What are the effects of pairing a non-associated auditory cue with the visual stimulus?

Future directions:

This preprint addresses and challenges theories of conscious perception that emphasize either local sensory processing or higher-level, global access to information. Future work could investigate whether there is a critical time window after stimulus offset during which retro-cues selectively enhance high-level awareness without restoring low-level details. Although the authors note that the timing of retro-cue effects is flexible, it would be valuable to determine the precise temporal range. Additionally, it remains an open question whether retro-cues preferentially enhance semantic meaning over other cognitive features, such as emotional content or syntactic structure.

References:

1. Dehaene, S., Changeux, J. P., & Naccache, L. (2011). The global neuronal workspace model of conscious access: from neuronal architectures to clinical applications. Characterizing consciousness: From cognition to the clinic?, 55-84.
2. Koch, C., Massimini, M., Boly, M., & Tononi, G. (2016). Neural correlates of consciousness: progress and problems. Nature reviews neuroscience, 17(5), 307-321.
3. Lamme, V. A. Towards a true neural stance on consciousness. Trends in cognitive sciences 10, 494–501 (2006).
4. Rosenthal, D. M. 2005 Consciousness and mind. Oxford, UK: Clarendon Press.
5. Sergent, C., & Naccache, L. (2012). Imaging neural signatures of consciousness: ‘What’,‘When’,‘Where’ and ‘How’ does it work?. Archives italiennes de biologie, 150(2/3), 91-106.

Tags: consciousness, perception, semantic cues, visual cues

Read preprint (No Ratings Yet)