Does Working Memory Selectively Modulate Subjective Perception?

⚕️

This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

The Big Question: Does Your Brain's "To-Do List" Change What You See?

Imagine your brain has two main jobs happening at the same time:

The Watcher: Looking at the world right now (Perception).
The Keeper: Holding a few things in your mind, like a phone number or a grocery list (Working Memory).

Scientists have long wondered: Does the "Keeper" job change how the "Watcher" sees things?

Specifically, they wanted to know if holding a specific image in your mind makes you actually see a matching image better, or if it just changes your feeling that you saw it. This is like asking: If I'm thinking about a red apple, does a blurry red dot look clearer to me, or do I just think I saw it more clearly?

The Experiment: The "Face" Game

The researchers set up a game to test this. They asked volunteers to memorize four faces (the "Keeper" task). While they were holding those faces in their minds, they had to play a quick game of "Spot the Face" (the "Watcher" task).

They created three scenarios:

The Twin Match (Stimulus Congruent): The face they were memorizing was the exact same face they had to spot.
The Family Match (Category Congruent): They were memorizing faces, and they had to spot a different face. (Same category, different person).
The Mismatch (Category Incongruent): They were memorizing faces, but they had to spot a picture of a scene (like a park or a building). This is like trying to find a needle while your brain is busy holding a brick.

The Two Ways of Measuring "Seeing"

To figure out if the memory was changing actual vision or just confidence, they used two different tests:

1. The "Forced Choice" Test (The Objective Score)

The Analogy: Imagine a magician shows you two cards. One has a face, one is a scrambled mess. You must pick which one has the face.
Why it matters: This is a pure test of your eyes and brain's processing power. It's hard to fake. If you get this right, you actually saw the difference.

2. The "Yes/No" Test (The Subjective Score)

The Analogy: The magician shows you one card. You have to shout "Yes, I see a face!" or "No, it's just noise."
Why it matters: This relies on your internal feeling. It's easier to be biased here. If you are thinking about faces, you might shout "Yes!" even if you aren't 100% sure.

The "Blindsight" Connection

The researchers were inspired by a rare condition called Blindsight. In this condition, people can point to an object they "don't see" (they pass the Forced Choice test) but insist they are blind to it (they fail the Yes/No test). This proves that "seeing" (objective) and "feeling like you see" (subjective) can be separated.

The team wanted to see if working memory could create a similar split: Could holding a face in your mind make you feel like you see a face better, even if your actual eyes aren't any sharper?

What Happened? (The Results)

The Pilot Study (The Small Guess):
In a small test with 8 people, they saw a hint that when people held a face in their mind but had to look for a scene (the Mismatch), their "Yes/No" confidence dropped. It looked like the memory was messing with their subjective feeling.

The Big Study (The Real Test):
They repeated the experiment with 16 people and made the test very precise. They adjusted the brightness of the images so that everyone got the "Forced Choice" test right about 79% of the time. This ensured that everyone's "eyes" were working at the exact same level.

The Verdict:
The effect disappeared.
When the "eyes" were working equally well, holding a face in memory did not selectively change how people felt about what they saw.

If the memory made it harder to see, it made it harder for both tests (Forced Choice and Yes/No).
It didn't just mess with the "feeling" (subjective); it messed with the whole system.

The Conclusion: A Simple Analogy

Imagine your brain is a radio.

The Pilot Study suggested that if you were thinking about Jazz, and someone played Country music, the radio would still play the Country music clearly, but you would feel like it sounded fuzzy.
The New Study found that if you are thinking about Jazz and someone plays Country music, the radio just gets a little bit of static for everyone. The volume goes down for both the "objective signal" and your "subjective feeling."

In short: Working memory doesn't act like a magical filter that tricks your brain into feeling like you saw something when you didn't. Instead, if your memory is busy with the "wrong" thing, it just makes your whole brain a little slower and less sharp, affecting everything equally.

What's Next?

Because the results didn't match their first small guess, the authors are planning to run the experiment one more time with a slightly different setup (removing some of the "noise" in the test) to see if they can find that original effect. They are being very careful to make sure they aren't just seeing patterns that aren't there.

1. Problem Statement & Research Question

The study investigates the interaction between Working Memory (WM) and Perception, specifically addressing whether WM content selectively modulates subjective perception (conscious experience) distinct from general perceptual processing capacity (objective discrimination ability).

Theoretical Context: Previous research suggests WM and perception share neural mechanisms (e.g., in V1/V2) and that WM can bias perception (e.g., repulsion effects). However, it remains unclear if this influence is a general reduction in sensory capacity or a specific alteration of conscious awareness.
The "Blindsight" Analogy: The authors draw on the phenomenon of blindsight, where patients show intact objective discrimination (2-IFC) but impaired subjective detection (Yes/No). They hypothesize that holding incongruent WM content might similarly dissociate these two measures, selectively impairing subjective sensitivity ( $d_a$ ) while leaving objective capacity intact.
Pilot Discrepancy: A previous pilot study ( $n=8$ ) suggested a trend where categorically incongruent WM content impaired sensitivity in a Yes/No (Y/N) task but not in a Two-Interval Forced Choice (2-IFC) task. This study aimed to rigorously test and replicate that finding with a larger sample.

2. Methodology

Experimental Design

Design: A $2 \times 3$ $2 \times 3$ within-subject factorial design.
- Perceptual Tasks: 2-IFC (Objective capacity) vs. Yes/No (Subjective detection).
- Congruency Conditions:
  1. Stimulus Congruent (SC): WM sample is physically identical to the perceptual stimulus.
  2. Category Congruent (CC): WM sample belongs to the same category (faces) but is a different image.
  3. Category Incongruent (CI): WM sample belongs to a different category (scenes) while the perceptual stimulus is a face.
Participants: 16 university students (mean age 23.6).

Stimuli & Apparatus

Visual Stimuli: Grayscale faces (Chicago Face Database), scenes, and scrambled faces (generated via Fourier phase scrambling to preserve low-level properties like spatial frequency while removing configural identity).
Display: CRT monitor (60 Hz) to ensure high temporal precision.
Task Structure:
1. WM Encoding: Participants memorize 4 images (faces or scenes) for 2000ms.
2. Delay/Perception: During the WM delay, participants perform either the 2-IFC or Y/N task.
  - 2-IFC: Two intervals; one contains a face, the other a scrambled face. Participants identify the interval containing the face.
  - Y/N: A single interval containing either a face or a scrambled face. Participants judge "Yes" (face present) or "No" (face absent).
3. WM Probe: Participants identify which of the 4 original images was replaced.

Critical Control: Contrast Titration

To ensure that any difference in Y/N performance was due to subjective modulation rather than general capacity differences:

The contrast of the face stimuli in the 2-IFC task was adaptively titrated using a 1-up-3-down staircase to target ~79% accuracy for each congruency condition.
The resulting contrast threshold for each condition was then fixed and applied to the subsequent Y/N task. This ensures that the "processing capacity" (signal strength) is matched across conditions before measuring subjective sensitivity.

Data Analysis

Signal Detection Theory (SDT): Sensitivity ( $d_a$ ) and criterion were calculated using an unequal-variance SDT model.
Type-1 zROC: Reaction time (RT) data was used to estimate the slope ( $s$ ) of the zROC to account for unequal variances between signal and noise distributions.
Metacognition: Metacognitive sensitivity ( $meta-d_a$ ) and efficiency ( $M_{ratio}$ ) were inferred from RTs (dividing trials into confidence bins).

3. Key Results

Working Memory Performance

WM accuracy was well above chance (~63–68%) but did not reach ceiling.
There was no significant effect of congruency or perceptual task type on WM accuracy ( $p = .56$ ), indicating participants successfully maintained the memory load across conditions.

Perceptual Capacity (2-IFC)

Contrast Matching: The adaptive staircase successfully matched performance across conditions (Mean accuracy $\approx$ 76–78%).
Contrast Thresholds: The physical contrast values required to reach threshold were statistically identical across SC, CC, and CI conditions ( $p = .516$ ).
Sensitivity ( $d_a$ ): No significant differences in 2-IFC sensitivity were found across congruency conditions ( $p = .200$ ). This confirms that general perceptual processing capacity was successfully matched.

Subjective Perception (Y/N)

Sensitivity ( $d_a$ ): Contrary to the pilot study and the authors' hypothesis, there was no significant effect of congruency on Y/N sensitivity ( $p = .829$ ).
Task Comparison: As theoretically expected, 2-IFC sensitivity was higher than Y/N sensitivity. When 2-IFC $d_a$ was scaled by $\sqrt{2}$ , it closely matched Y/N $d_a$ across all conditions, confirming the theoretical relationship between the two tasks was preserved.
Numerical Trends: While a slight numerical reduction in $d_a$ was observed in the CI condition for both tasks, the reduction was parallel (affecting both 2-IFC and Y/N) rather than selective. The ratio between Y/N and 2-IFC sensitivity remained stable.

Metacognition

The study analyzed metacognitive efficiency ( $M_{ratio}$ ), but the primary focus remained on the dissociation between objective and subjective detection, which was not found.

4. Key Contributions

Rigorous Control of Capacity: The study employed a sophisticated titration method (staircasing 2-IFC to match contrast thresholds) to isolate subjective perception from general sensory capacity, a methodological improvement over previous studies that often relied on fixed stimulus intensities.
Replication Failure: The study failed to replicate the pilot finding that incongruent WM selectively impairs subjective perception (Y/N) while sparing objective capacity (2-IFC).
Theoretical Clarification: The results suggest that incongruent WM content may act as a general sensory interference (reducing signal strength globally) rather than a specific mechanism that dissociates conscious awareness from sensory processing (like blindsight).
Pre-registration and Transparency: The authors transparently reported the discrepancy between their pilot and main study, pre-registered a follow-up study to test the original pilot design, and decided against a planned follow-up experiment that was no longer motivated by the current null results.

5. Significance & Discussion

Re-evaluating the "Selective Modulation" Hypothesis: The findings challenge the idea that WM selectively targets the "subjective" component of perception. Instead, the data suggests that WM interference likely operates by reducing the overall signal-to-noise ratio, affecting both objective discrimination and subjective awareness equally.
Task Design Implications: The authors speculate that the failure to replicate the pilot result may be due to the specific Y/N design used here (where scrambled faces were used as "noise" in target-absent trials). This might have turned the task into a discrimination process rather than pure detection, reducing sensitivity to congruency effects.
Future Directions: The authors pre-register a follow-up study using the original pilot design (pure detection Y/N with no stimulus in target-absent trials) to determine if the selective modulation effect exists under different task constraints. They also note that future work should distinguish between WM and mental imagery, as the latter might interact with perception differently.

Conclusion: Under the current experimental conditions where perceptual capacity was strictly matched, working memory content did not selectively modulate subjective perception. The interaction between WM and perception appears to be a general reduction in perceptual sensitivity rather than a dissociation between objective and subjective processing.