Testing Theories of Consciousness

CASE STUDY: Adversarial Testing of Global Neuronal Workspace and Integrated Information Theories of Consciousness

How does consciousness arise from physical neural activity in the human brain? To answer this question, an international team of researchers set up a massive open-science adversarial collaboration involving data collected from 256 participants across seven international laboratories. Published in Nature, this landmark study tested preregistered, contrasting predictions from both theories using functional MRI (fMRI), magnetoencephalography (MEG), and intracranial EEG (iEEG), while EyeLink eye trackers monitored participants’ gaze during the experiments. Here is a look at who proposed these theories, what they actually claim in practical terms, how the study tested them, and what the results mean for the future of consciousness research.

Global Neuronal Workspace Theory (GNWT)

• Who Proposed GNWT: Developed primarily by cognitive neuroscientist Stanislas Dehaene and neurobiologist Jean-Pierre Changeux (building on Bernard Baars‘ original psychological workspace model).
• Philosophical Framework—Materialist / Functionalist: GNWT assumes that consciousness is produced entirely by physical brain activity and defines consciousness by what it does—its functional role in broadcasting information across the brain.
• How It Defines Consciousness: Imagine the brain as a large organization. Most brain regions work like specialized departments, quietly processing their own information. According to GNWT, information becomes conscious when it is selected from localized sensory areas and “broadcast” across a distributed frontoparietal workspace—a network linking the prefrontal cortex (PFC) with the parietal cortex. This network is thought to overlap substantially with the Frontoparietal Network (sometimes called the Central Executive Network), although the proposed global workspace is a theoretical functional network rather than a precisely defined anatomical one. Once broadcast, the information becomes globally available for memory, planning, reasoning, and decision-making.
• When Consciousness Is Activated: GNWT predicts two brief bursts of widespread brain activity during a conscious visual experience:
  • Stimulus Onset (0.0 s): An image first appears and begins being processed unconsciously in localized visual areas.
  • Onset Ignition (300–500 ms later): A sudden burst of activity (“ignition”) spreads through the frontoparietal workspace, particularly involving the PFC, making the information consciously available.
  • Activity-Silent Maintenance: While the image remains visible (for example, 1.0 or 1.5 seconds), the conscious information is thought to be maintained without continuous widespread firing, through temporary synaptic changes.
  • Stimulus Offset: When the image disappears, GNWT predicts a second ignition approximately 300–500 ms later as the frontoparietal workspace updates to the new conscious scene.
• Predictions
  • PFC Ignition: The PFC is predicted to show brief bursts of activity during the 300–500 ms windows following both stimulus appearance and disappearance.
  • Global Broadcast Representation: Information about the stimulus—including its category and, ideally, finer details such as identity and orientation—should become represented throughout the frontoparietal workspace, making it globally available independently of task demands.

Integrated Information Theory (IIT)

• Who Proposed It: Formulated by neuroscientist and psychiatrist Giulio Tononi, and prominently championed alongside neuroscientist Christof Koch.
• Philosophical Framework—Phenomenological / Intrinsic: IIT starts with the properties of subjective experience rather than brain function. It proposes that consciousness depends on the amount of integrated information generated by a system. Because the theory allows integrated information to exist in systems beyond brains, many philosophers argue that IIT has panpsychist or panpsychist-like implications.
• How It Defines Consciousness: Imagine thousands of musicians performing independently. Individually they produce sounds, but only when they play together in perfect coordination does a symphony emerge. IIT argues that consciousness works similarly: experience arises because information becomes integrated into a single unified whole. According to IIT, the brain region with the highest level of integrated information is proposed to be anchored in the brain’s posterior “hot zone” at the back of the head, combining the occipital cortex (colors/lines), parietal cortex (space/depth), and temporal cortex (identity/faces/sound).
• When Consciousness Is Activated: Unlike GNWT’s brief ignition events, IIT predicts that conscious experience remains actively supported throughout the entire time an object is perceived.
• Prediction:
  • Sustained Connectivity: IIT predicts that lower-level visual areas (V1/V2) and higher-level visual areas (such as the fusiform face area) should remain continuously synchronized throughout the entire time a stimulus is consciously perceived. This ongoing synchronization within the brain’s posterior cortex is proposed to support conscious experience.
  • Posterior Feature Representation: The posterior hot zone should continuously represent conscious stimulus features—including category, identity, and orientation—throughout the experience, independent of task demands.

The Methodology: Separating Perception from Task Performance

One of the biggest challenges in consciousness research is distinguishing the neural activity associated with consciously seeing something from the brain activity involved in paying attention, making decisions, or pressing a button.

Images appeared for 0.5, 1.0, or 1.5 seconds. To reduce these task-related influences, participants were instructed before each block to watch for two specific target images—for example, one particular face and one particular object. This naturally created three kinds of trials:

• Target Stimuli (Perception + Attention + Action): When participants saw one of the designated targets, they pressed a button.
• Task-Relevant Non-Targets (Perception + Attention):Other faces or objects required attention because they belonged to the attended categories, but participants made no response.
• Task-Irrelevant Stimuli (Minimizing Task Demands): Letters and abstract symbols also appeared while participants searched for faces and objects. Because these stimuli were irrelevant to the task, participants did not have to respond to or actively use this information, even though the stmuli remained clearly visible. These trials provided the researchers with their closest approximation of conscious visual perception while substantially reducing—but not completely eliminating—attention and task-related processing.

The researchers then compared activity across these different trial types to distinguish neural activity associated with conscious visual perception from activity related to attention, decisions, and motor responses. This allowed them to evaluate each theory’s predictions while minimizing common experimental confounds.

To ensure high-quality neural recordings, EyeLink 1000 Plus eye trackers continuously monitored gaze position throughout the experiment. This allowed researchers to verify stable fixation and identify trials in which blinks or eye movements could have affected the neural recordings. Eye tracking was synchronized in real-time alongside three complementary neuroimaging technologies: iEEG (high-gamma neural spiking), MEG (fast magnetic brain signals), and fMRI (whole-brain blood oxygenation).

Key Takeaways: Neither Theory Passed Completely

When the data was analyzed, neither theory emerged unscathed.

• Where Dehaene’s GNWT Struggled: The prefrontal cortex showed a strong ignition burst when images first appeared (roughly 300–500 ms after stimulus onset), consistent with one of GNWT’s central predictions. However, the predicted offset-related prefrontal ignition was not consistently observed after images disappeared. Because GNWT proposes that the global workspace updates conscious experience through both onset and offset ignition events, the result represents one of the theory’s strongest challenges. Furthermore, granular details like stimulus orientation were largely absent in the prefrontal cortex.
• Where Tononi’s IIT Struggled: Although posterior visual regions did maintain activity tracking stimulus duration, the results did not show sustained gamma-band synchronization at the level predicted by IIT between early visual areas (V1/V2) and high-level category areas over time. Additionally, representations of stimulus orientation decayed rapidly in posterior regions rather than lasting the full duration of the percept.

Both theories captured important aspects of conscious processing, but neither accounted for all of the preregistered observations. For GNWT, the absence of the predicted prefrontal offset ignition challenges its proposed mechanism for updating conscious experience over time. For IIT, the absence of sustained posterior synchronization challenges the claim that ongoing connectivity within the posterior cortex specifies conscious experience. Broadly, this landmark study demonstrates how preregistered adversarial collaborations can move consciousness research beyond competing narratives toward more rigorous, empirically verified theory testing.

For information regarding how eye tracking can help your research, check out our solutions and product pages or contact us. We are happy to help!