# A compensatory relationship between working and long-term memory for emotional faces

## TL;DR

The study shows that better memory for emotional faces in the short term is linked to worse memory for the same emotions in the long term.

## Contribution

It reveals a compensatory relationship between working and long-term memory for emotional faces at the individual level.

## Key findings

- Group-level analysis showed no significant difference in working memory for happy and angry faces.
- Long-term memory showed a strong advantage for happy faces over angry ones.
- Individuals with better working memory for one emotion had reduced long-term memory advantage for the same emotion.

## Abstract

Previous research has presented contradictory findings regarding emotional memory enhancement, often demonstrating an “angry face advantage” in working memory (WM) and a “happy face advantage” in long-term memory (LTM). However, the relationship between emotional memory enhancements in WM and LTM, particularly at the individual level, remains unclear. This study investigated the change in emotional memory for happy versus angry faces from immediate to delayed recognition and explored the individual-level correlation between WM and LTM performance.

A memory experiment was conducted over two consecutive days. On the first day, participants completed a change-detection task to measure WM for angry and happy faces. Approximately 24 h later, a surprise LTM recognition task was administered to assess the memory of the facial identities encountered during the WM task. This study measured WM capacity and LTM sensitivity for both emotional faces, and a correlation analysis was performed between individual bias scores (difference in memory for happy versus angry faces) in WM and LTM tasks.

There was no significant difference in memory capacity in the group-level analysis of WM performance for happy and angry faces. In contrast, the delayed LTM test showed a robust happy face advantage, with significantly better recognition of faces initially encountered with happy expressions than with angry expressions. Most critically, the correlation analysis revealed a significant negative relationship between WM and LTM emotional biases, indicating that individuals with relatively better WM performance for one emotional valence subsequently exhibited a reduced LTM advantage for the same valence.

The negative correlation between WM and LTM emotional biases aligns with the arousal-biased competition model, suggesting that the cognitive resources allocated to maintaining threatening information in WM may come at the cost of elaborate encoding processes necessary for robust LTM consolidation. These results provide novel evidence for a compensatory processing mechanism wherein the cognitive system manages immediate and delayed memories of emotional faces in a complementary rather than cumulative fashion.