A dynamic affective surprise signal influences episodic memory

Overview

This research examines how affective surprise—quantified as deviations in expected affective states across valence and arousal dimensions—functions as a learning signal modulating episodic memory formation. The study operationalizes affective surprise through continuous affect ratings and investigates its influence on temporal context binding and item recognition in memory. The investigation bridges literatures on prediction error-driven learning and affective neuroscience by proposing that dynamic changes in subjective emotional states, independent of valence direction, enhance memory consolidation.

Methods and approach

The research employed a two-phase design: reanalysis of an existing dataset followed by an independent replication study. In both phases, participants encoded item sequences while listening to emotional music. During a subsequent listening session, participants provided continuous ratings of felt affect across valence and arousal dimensions. A novel computational metric for affective surprise was derived from these individual-level ratings, operationalized as deviations from recent affective state history. Episodic memory was assessed 24 hours post-encoding, measuring both temporal source memory (when items occurred in sequences) and item recognition memory. The temporal specificity of affective surprise effects was examined in relation to memory outcomes.

Results

Affective surprise significantly enhanced memory for temporal context, with greater deviations from recent affective state history associated with improved memory for item sequence position. This effect held across both studies, indicating robust temporal binding enhancement. Conversely, affective surprise effects on item recognition memory proved inconsistent between the reanalyzed dataset and the replication study. Valence-related surprise demonstrated more consistent effects than arousal-related surprise. The directionality of affective surprise—whether increases or decreases in affect—did not differentially impact temporal memory enhancement, suggesting valence surprise operates as a bidirectional learning signal.

Implications

The findings establish affective surprise as a potential learning signal with specific consequences for episodic memory architecture, particularly for binding items to temporal contexts rather than supporting item-specific memory. The proposed computational metric offers a methodological framework for quantifying dynamic affective deviations from individual baselines, advancing precision in affective neuroscience measurement. The inconsistent effects on recognition memory versus robust temporal context effects suggest domain-specific roles for affective surprise in memory systems.