Perceptual glimpses are locally accumulated and globally maintained at distinct processing levels

Through EEG experiments and modeling, this study reveals that while perceptual evidence from intermittent glimpses is transiently accumulated at the centroparietal level, it is simultaneously maintained as a sustained decision variable in motor beta lateralization to support continuous integration across gaps.

The Gist

The Big Picture: Making a Decision with "Glances"

Imagine you are trying to decide which way to walk through a busy, foggy park. You can't see the whole path at once. Instead, you get two quick glimpses of the path, separated by a moment of thick fog where you can't see anything.

• Glimpse 1: You see a path leading left.
• The Fog: You wait for a few seconds.
• Glimpse 2: You see the path again, still leading left.

Your brain has to combine these two quick looks to make a final decision: "Go Left!"

This study asked a simple question: How does our brain handle information when it comes in bursts, rather than a steady stream? Does it hold onto the first look perfectly while waiting for the second? Or does it forget a bit? And what is happening inside our brains while we wait in the fog?

The Two "Brain Managers"

The researchers used EEG (a cap with sensors on the head) to watch two specific "managers" in the brain doing the work. Think of them as a Construction Site:

1. The Foreman (The CPP): This manager is in charge of gathering the bricks (the evidence).

   • What they found: The Foreman works hard during the glimpses, stacking bricks quickly. But as soon as the fog rolls in (the gap), the Foreman packs up and goes home. They don't hold the stack of bricks; they just finish the job for that specific glimpse and reset to zero. When the second glimpse arrives, the Foreman starts stacking again from scratch.
   • The Analogy: It's like a cashier scanning items. They scan the first item, put it in a bag, and then the scanner turns off. When the second item comes, they scan it again. They don't keep the scanner "on" while you are looking for your wallet in the gap.

2. The Warehouse Manager (Motor Beta): This manager is in charge of holding the final decision and getting the body ready to move.

   • What they found: This manager is very different. Once the Foreman stacks the first batch of bricks, the Warehouse Manager holds onto that stack tightly. Even when the fog rolls in and the Foreman goes home, the Warehouse Manager keeps the stack safe, never dropping a single brick. They keep holding it until the second glimpse arrives, add the new bricks, and then finally send the order to the muscles to move.
   • The Analogy: This is like a person holding a heavy tray of drinks. Even if they stop walking for a moment (the gap), they don't put the tray down. They keep it balanced and ready until they reach the table.

The "Primacy" Effect: Why the First Glimpse Matters More

The study found something interesting about how people made their choices. Even though both glimpses were used, the first glimpse had more influence on the final decision than the second one.

• The "Early Exit" Theory: The researchers used a computer model to explain this. They found that our brains often act like a person with a patience limit.
  • If the first glimpse is very clear (high confidence), the brain says, "Okay, I'm sure! Let's stop looking!" and makes a decision immediately. The second glimpse is ignored because the decision was already "locked in."
  • If the first glimpse is blurry, the brain waits for the second one.
  • This explains why the first glimpse often feels more powerful: sometimes, we decide before we even get the second look.

The "Gap" Mystery: Did We Forget?

A common worry is that if you wait too long between glimpses, you might forget the first one (like trying to remember a phone number while waiting for a second part of it).

• The Result: The study found no forgetting. The accuracy didn't drop even if the "fog" lasted a long time.
• Why? Because of the Warehouse Manager. Even though the Foreman (CPP) went home, the Warehouse Manager (Motor Beta) kept the first glimpse safe in their memory the whole time. The information didn't leak out; it was just waiting in the warehouse.

The Attention Switch (The Alpha Waves)

The researchers also looked at how people paid attention. They found that when the "fog" (the gap) was on, the brain's attention signals actually turned up the volume to block out the noise.

• The Analogy: Imagine wearing noise-canceling headphones. When the "evidence" (the moving dots) stops, the brain puts on the headphones to ignore the random noise so it doesn't get confused. When the evidence starts again, the headphones come off. This shows our brains are very smart at knowing exactly when to listen and when to ignore.

Summary: The Takeaway

This study reveals that our brain has a two-stage system for making decisions when information is intermittent:

1. Local Processing (The Foreman): We process each piece of evidence separately. We don't try to hold a continuous "thought" about the evidence while waiting; we just process the current glimpse and reset.
2. Global Maintenance (The Warehouse): The result of that processing is held safely in a different part of the brain (motor preparation) so we don't lose it.

In everyday terms: When you get a text message, then wait a few minutes, then get another one, your brain doesn't try to "read" the first one continuously while you wait. It reads it, saves the meaning in your short-term memory, and then reads the second one. But if the first message was super clear, you might have already decided what to do before the second one even arrived!

Technical Summary

1. Problem Statement

Perceptual decision-making is widely understood as a process of evidence integration over time. While the neural architecture for continuous evidence accumulation is well-characterized (involving signals like the Centroparietal Positivity [CPP] and Motor Beta Lateralisation [MBL]), less is known about how these mechanisms operate when evidence is presented intermittently.

In real-world scenarios, decision-makers often receive "glimpses" of evidence separated by gaps (e.g., an object occluded by traffic). Previous behavioral studies using "gaps tasks" (two pulses of noisy motion separated by a variable gap) showed that participants use both pulses but often underutilize the second one, without a systematic loss of accuracy over time (ruling out simple information leakage). However, the neural underpinnings of how the brain handles, maintains, and integrates these discrete pulses, and the specific functional roles of intermediate (motor-independent) vs. downstream (motor) processing levels, remained unclear.

2. Methodology

The authors conducted two EEG experiments with human participants performing a motion direction discrimination task.

• Task Design (The "Gaps Task"):

  • Participants viewed random dot kinematograms (RDK).
  • Trials consisted of one or two pulses of coherent motion (yellow dots) indicating a direction (Left/Right), separated by a variable temporal gap of incoherent motion (blue dots).
  • Experiment 1: $N=22$; Gap durations: 0, 0.5, 1.0 s.
  • Experiment 2: $N=21$; Gap durations: 0, 0.12, 0.36, 1.08 s (skewed toward shorter gaps).
  • Coherence levels were titrated per participant to target ~70% accuracy on single-pulse trials.
  • Responses were delayed until after the trial ended (delayed-response design).

• Neural Measures:

  • Motor Beta Lateralisation (MBL): Beta band (13–30 Hz) power over motor cortex. Used as a marker for motor preparation and the sustained state of the Decision Variable (DV).
  • Centroparietal Positivity (CPP): A centro-parietal ERP component. Used as a marker for evidence accumulation (motor-independent).
  • Occipital Alpha (8–12 Hz): Used to index attentional gating (desynchronization indicates attention to the stimulus).

• Analysis & Modeling:

  • Behavioral: Logistic regression to assess the influence of pulse coherence, order, and gap duration on accuracy.
  • Neural: Time-resolved regressions and cluster-based permutation tests to analyze CPP and MBL dynamics relative to pulse coherence and trial outcome.
  • Computational Modeling: A bounded diffusion model was fitted to behavioral data. The model simulated a single Decision Variable (DV) that accumulates evidence only during pulses, pauses during gaps, and terminates upon hitting a bound.
  • Neural Simulation: The model was extended to simulate MBL (tracking the signed, sustained DV) and CPP (tracking the absolute, local accumulation $|\Delta DV|$ within each pulse).

3. Key Contributions

1. Functional Dissociation of Neural Signals: The study demonstrates a clear functional separation between the CPP and MBL in intermittent evidence tasks. The CPP reflects local, transient accumulation (resetting between pulses), while MBL reflects global, sustained maintenance of the decision state.
2. Mechanism of Underutilization: The authors identify that the behavioral "underutilization" of the second pulse is not due to memory decay or leakage, but rather a bounded accumulation process where decisions are sometimes terminated early (primacy effect) once a bound is reached during the first pulse.
3. Attentional Gating: Evidence that occipital alpha power dynamically gates attention, suppressing processing during the irrelevant gap periods and enhancing it during evidence pulses.

4. Key Results

Behavioral Findings

• Primacy Effect: Participants were more accurate when the first pulse had high coherence compared to when the second pulse had high coherence (HL > LH). This indicates a stronger weighting of the first piece of evidence.
• No Leakage: Accuracy did not decrease as a function of gap duration, confirming that information from the first pulse was perfectly maintained during the gap.
• Suboptimal Integration: Participants did not achieve the accuracy of a "perfect accumulator" (who would integrate all evidence perfectly), suggesting they engaged in a bounded process that sometimes stopped early.

Neural Findings

• MBL (Motor Preparation):
  • Showed a sustained buildup following the first pulse.
  • Maintained a coherence-dependent level throughout the gap and until the response.
  • This indicates the motor system holds a continuous, global representation of the accumulated evidence (the DV) without loss.
• CPP (Evidence Accumulation):
  • Exhibited transient dynamics. It built up during each pulse, peaked, and then fell back to baseline during the gap.
  • Pulse 1 (CPP-P1): Amplitude scaled with Pulse 1 coherence.
  • Pulse 2 (CPP-P2): Amplitude scaled with Pulse 2 coherence but was generally smaller than CPP-P1. Crucially, CPP-P2 was larger when Pulse 1 was low coherence (because the bound was less likely to have been hit, allowing more accumulation space).
  • This pattern suggests the CPP tracks the absolute belief update ($|\Delta DV|$) of the current pulse, not the running sum of the total DV.
• Attention (Occipital Alpha):
  • Alpha power decreased (desynchronized) during yellow evidence pulses and increased (synchronized) during blue gap periods, confirming active gating of irrelevant information.

Computational Modeling

• A simple bounded diffusion model successfully recapitulated the behavioral primacy effect and the lack of gap-dependent leakage.
• Simulation of Neural Signals:
  • Simulating MBL as the signed, sustained DV matched the empirical MBL data (sustained through gaps).
  • Simulating CPP as the absolute, local accumulation that resets to zero when evidence stops (or a bound is hit) matched the empirical CPP data (transient peaks, baseline reset).
  • Models assuming the CPP sustained the DV during the gap failed to reproduce the observed drop to baseline and the specific amplitude relationships between P1 and P2.

5. Significance

• Refining the Neural Architecture of Decision Making: The study challenges the view that the CPP always tracks a continuous, running sum of evidence. Instead, it shows the CPP is a flexible, local integrator that can reset, while the maintenance of the decision state is offloaded to downstream motor areas (MBL).
• Context-Dependent Dynamics: It highlights that neural decision signals are not static; their functional roles (sustained vs. transient) depend on the temporal structure of the evidence (continuous vs. intermittent).
• Explanation of Suboptimal Behavior: The findings provide a neural mechanism for "bounded rationality" in decision making: the brain may terminate accumulation early to save cognitive effort, leading to primacy effects, rather than integrating all available information perfectly.
• Attentional Mechanisms: It links alpha oscillations to the active filtering of irrelevant temporal windows, ensuring that accumulation only occurs when evidence is present.

In summary, the paper establishes that for intermittent evidence, the brain employs a two-level architecture: a local, transient accumulator (CPP) that processes discrete glimpses, feeding into a global, sustained motor buffer (MBL) that maintains the decision state across time gaps.