Image Memorability Predicted N400 Encoding Signals More Than Recognition

TL;DR: A 2026 study in Cognitive, Affective, & Behavioral Neuroscience found that image memorability, the stable tendency of an image to be remembered across people, predicted N400 brain responses at encoding even after later recognition success was modeled.

Source: Cognitive, Affective, & Behavioral Neuroscience (2026) | Deng and Federmeier

Image memorability is not just whether a single viewer happened to remember a single picture. It is an item-level property: some images are remembered more reliably across people, even when viewers differ in attention, strategy, and context.

N400 Subsequent-Memory Effects May Reflect Image Memorability

The subsequent memory effect is a common memory-research pattern: brain activity during encoding predicts whether an item will be remembered later. In many studies, later-remembered items produce less negative N400 responses than later-forgotten items.

N400 is an event-related potential linked to semantic processing, or how the brain maps an object, word, or image into meaning. A less negative N400 often suggests easier semantic processing.

The interpretive problem is that memorable images also produce less negative N400 responses. A later memory-linked response might therefore reflect the image itself, not only the participant’s temporary encoding state.

Participants Studied Images and Returned for Recognition Testing

Researchers recruited 32 participants and analyzed 24 after exclusions for missed sessions, data quality, or unusable recordings. Participants first viewed images while ERPs were recorded.

After 2 days, participants completed a recognition test. The design let researchers classify each image by both its memorability score and the participant’s later outcome: recognized hit or missed item.

• Memorability score: A normed item-level estimate of how likely the image is to be remembered across people.
• Recognition outcome: Whether a specific participant recognized that image at the later test.
• ERP windows: Researchers focused on N300, N400, and LPC time windows during the initial image presentation.

Participants recognized previously seen images at an average rate of 0.55, with a false-alarm rate of 0.22. Higher memorability reliably predicted recognition, with a z value of 9.17 and p < 0.001.

High-Memorability Images Produced Less Negative N400 Signals

When images were split into high- and low-memorability groups, high-memorability images produced less negative N400 amplitudes at encoding. The averages were -5.69 microvolts for high-memorability images and -7.87 microvolts for low-memorability images.

The same direction appeared in the N300 and LPC windows. High-memorability images had less negative N300 responses and more positive LPC responses, and those effects remained when participant confidence was included in the models.

Recognition outcome also showed the expected pattern when tested alone. Later-recognized images produced less negative N400 amplitudes than later-missed images, and later-recognized images also produced more positive LPC amplitudes.

Joint Models Put Memorability Ahead of Recognition Status

The key analysis modeled memorability and recognition together. For N400 amplitude, memorability remained significant with t = 5.66 and p < 0.001.

Recognition status was not significant in that same N400 model, with t = -0.26 and p = 0.794. The interaction between memorability and recognition was also not significant.

The LPC model showed a similar pattern. Memorability remained significant with t = 4.95 and p < 0.001, while recognition status was not significant.

The likelihood-ratio tests added nuance. Adding recognition still improved model fit for both N400 and LPC, so recognition was not irrelevant.

The fixed-effect pattern still showed that much of the encoding response was carried by image memorability. That distinction keeps the interpretation tied to both stimulus properties and participant-level memory outcome.

Memory Brain Signals Are Not Always Pure Encoding-State Markers

The study changes how N400 subsequent-memory effects should be interpreted. A less negative N400 before later recognition may not mean only that a participant happened to encode the item better in that moment.

Some of the encoding response can come from the image’s stable memorability. Images that are easier to map semantically may produce more fluent processing and also be more likely to be remembered later.

• Item-level contribution: The image itself can carry features that make recognition more likely.
• Person-level contribution: Attention, confidence, and encoding state still influence whether a person remembers the image.
• Modeling implication: Memory studies should account for stimulus memorability before treating encoding responses as purely participant-driven.

The Result Fits Visual Memory Better Than General Memory Claims

This was an image-recognition experiment with 24 analyzed participants, not a broad test of all memory types. The findings apply most directly to visual episodic memory and image-based N400/LPC effects.

Even with that boundary, the study changes the interpretation of remembered-versus-forgotten contrasts. Memory experiments often sort trials into remembered and forgotten bins, then interpret the brain difference as encoding success.

This study shows that the item being encoded can explain a large part of the same brain-response difference.