Conference Agenda

Our subjective reality corresponds to a continual stream of perceptual updates. Yet, how the brain constructs this stream remains a scientific mystery. Although experience appears to occur in real-time, psychophysical evidence supports a two-stage model of perception: conscious updates, corresponding to simplified post-hoc interpretations, are preceded by long-lasting periods of unconscious processing, which retain high spatiotemporal fidelity to the incoming information. Here, we used magnetoencephalography and multivariate decoding across time and frequency bands to delineate the neural underpinnings of phenomenology (stage 2) versus sensory processing (stage 1). Subjects performed a motion-induced blindness (MIB) task, in which the illusory disappearances and reappearances of a salient target reflect the closure of unconscious windows leading to perceptual updates. Notably, we kept decoders blind to MIB-related decisional and motor processes by training them on brief localizer trials embedding no task.

Our analyses uncovered a dissociation between high-frequency activity patterns (30-60Hz) and the slow cortical potentials (SCP, 0.05-5Hz). Starting ~500ms before behavioral reports, SCP classifiers successfully discerned target reappearance/disappearance whereas those trained on gamma performed at chance. Crucially, gamma classifiers succeeded on phenomenally-matched control trials, confirming sufficient SNR. Further inspection of their estimates revealed that in MIB trials gamma classifiers always assigned high probability to target’s presence regardless of perceptual state, indicating that gamma activity faithfully tracked the physical input. Overall, our results suggest that while SCP supports conscious contents, gamma provides a substrate for unconscious analysis. These findings reveal a frequency-based division of labor, with distinct bands orchestrating the temporal structure of perceptual awareness.

One of the central aims of visual consciousness research is to understand how perception transitions from unconscious to conscious states. Although numerous studies have investigated the thresholds and emergent processes of visual consciousness in adults, relatively little is known about how these processes develop in childhood. Here, we examined the thresholds and emergent processes of visual consciousness in color and category in preschoolers and adults using a backward masking paradigm.

Sixteen 5–6-year-old children and fifteen adults completed two discrimination tasks (color and category discrimination tasks) with varying stimulus onset asynchronies (SOAs; 16.7–216.7 ms for children, 16.7–83.4 ms for adults). Objective discrimination accuracy was recorded, and subjective awareness was assessed using a four-scale Perceptual Awareness Scale (PAS). We fit the data with a four-parameter nonlinear psychometric function to estimate each participant's threshold (center of the slope) and gradualness (range of the slope).

The results revealed that the thresholds of objective discrimination and subjective awareness were significantly larger in 5–6-year-olds than in adults, without the main effect of task and interaction between age and task. Furthermore, the emergent processes of objective discrimination and subjective awareness were significantly more gradual in 5–6-year-olds than in adults, without the main effect of task and interaction.

These findings indicate that visual consciousness of color and category emerges at larger thresholds and more gradually in 5–6-year-olds, with no differences in the task performance of color and category. Such developmental differences may be linked to the maturation of the fronto-parietal network.

Introduction: Aphantasia, the absence of visual mental images, has recently been recognized as part of a continuum of normal function and can offer insights into mental imagery in general. While many brain regions have been correlated with visual mental imagery in functional imaging studies, lesions that lead to a loss of mental imagery can identify regions that are causal.

Methods: We performed a literature review to identify lesions associated with aphantasia. Lesions were mapped and compared to a large resting-state connectome. Locations sensitive (T>7, >75% overlap) and specific (family-wise error p<0.05) for aphantasia were identified. The lesion locations were then compared to neuropsychiatric syndromes (n=31) previously studied in our laboratory for similarities.

Results: We identified 11 cases of lesion-induced aphantasia with focal injury to multiple brain locations including the frontal, temporal, parietal and occipital lobes and subcortical structures, without significant lesion overlap. In contrast, greater than 90% of lesions causing aphantasia were functionally connected to the fusiform gyrus (L>R). This location was both sensitive and specific (family-wise error p<0.05) for aphantasia. The aphantasia network was most similar to the lesion-induced blindsight network when compared to 31 neuropsychiatric syndromes.

Conclusion: Aphantasia causing lesions connect to a key region along the ventral visual pathway recently termed the fusiform image node adding causal evidence to the hypothesis that this region is necessary for visual mental imagery. Our results demonstrate similarities between the networks involved in aphantasia and blindsight suggesting shared mechanisms possibly related to non-conscious imagistic representations in these two previously disparate syndromes.

The COGITATE Consortium recently led an adversarial collaboration [1] to test the predictions of two prominent consciousness theories: integrated information theory (IIT) and global neuronal workspace theory (GNWT). Initial evidence failed to distinguish between these theories’ predictions from inter-areal functional connectivity (FC). Here, we leverage a data-driven approach to identify FC metric(s) that support pre-registered theoretic predictions.

We obtained the same magnetoencephalography (MEG) data [1] collected from N=94 healthy adults (22.7 +/- 3.5 years old) while viewing different visual stimuli. We compared stimulus decoding performance of 245 different FC metrics, computed between brain areas integral to IIT and GNWT. We then developed neuro-dynamical models based on top-performing features, tuned to predictions from either GNWT (central workspace competition) or IIT (sensory region recurrence).

Multiple FC metrics distinguished between stimulus categories in one or more region–region pairs, with strong overall performance from the barycenter [2], which captures the time-resolved ‘center of mass’ of two processes. The barycenter maximum detects periods of tightly-yoked signal increases in both regions, while the mean captures sustained inter-areal synchrony throughout the epoch. We computed these metrics from simulated time series from the two theory-based models, finding that IIT-predicted dynamics better matched those in empirical MEG data throughout stimulus presentation.

We present a systematic, highly comparative approach to uncover previously unexplored FC measures relevant to processing visual stimuli, with the barycenter as a promising candidate metric for quantitatively evaluating theoretic predictions using empirical and simulated data.

[1] COGITATE Consortium, arXiv (2023)

[2] Petitjean, Pattern Recognition (2010)

The more peripheral in the visual field a stimulus is presented, the harder it gets to detect it and to consciously perceive a displayed item. It is unclear whether human concept cells respond to different levels of detectability in a graded or an all-or-nothing manner. In up to now 37 experimental sessions, we recorded 3405 units in the medial temporal lobe from 8 neurosurgical patients implanted with hybrid Behnke-Fried depth electrodes, while displaying stimuli at eccentricities of 5, 10, and 15 degrees of visual angle, and performing simultaneous eye-tracking. Subjects had to fixate centrally and report the concept they perceived in the periphery. The analysis of our preliminary dataset reveals that the firing rates of concept cells are not only higher for detected versus undetected concepts, but that the firing rates in response to detected concepts also decrease with increasing eccentricity. In addition, the response latencies of these neurons are not only shorter for detected versus undetected stimuli, but also show an increase with increasing eccentricity for detected concepts. Furthermore, the concept cells showed higher firing rates in response to undetected preferred versus undetected non-preferred stimuli. These findings suggest that both firing rates and neuronal response latencies exhibit a graded relationship with the detectability of preferred concepts in the peripheral field of view, providing insight into how conscious perception emerges in the brain during peripheral vision.