Unexpected Twists: Electrophysiological Correlates of Encoding and Retrieval of Events Eliciting Prediction Error

Abstract

According to the predictive processing framework, our brain constantly generates predictions based on past experiences and compares these predictions with incoming sensory information. When an event contradicts these predictions, it results in a prediction error (PE), which has been shown to enhance subsequent memory. However, the neural mechanisms underlying the influence of PEs on subsequent memory remain unclear. This study investigated the electrophysiological correlates during encoding and retrieval of events eliciting PEs. We employed a statistical learning task in which participants were presented with pairs of objects in sequence. Subsequently, while recording electroencephalography (EEG), we introduced PEs by replacing the second object of each pair with new objects and we then tested the participants' memory. Behaviorally, PEs did not enhance memory. During retrieval, we observed higher amplitudes in the recollection-related late positive component for violation items that were remembered compared to those that were forgotten. In contrast, no evidence for the presence of the FN400 component associated with familiarity was found. These results suggest that recollection, but not familiarity, plays a crucial role in the interplay between PE and memory. Contrary to our hypothesis, we did not observe a relationship between PEs and the P3 component during encoding. In conclusion, our study contributes to the growing body of knowledge concerning the intricate relationship between PEs and episodic memory. It sheds light on the underlying neural mechanisms involved and emphasizes the importance of recollection in this context.

Keywords: electroencephalography; episodic memory; event‐related potentials; prediction error; predictive processing; statistical learning.

FIGURE 1

Study design. (A) During prediction learning (Days 1 and 2), participants viewed pairs of sequentially presented objects and were asked to indicate whether each object was bigger or smaller than the previous object. (B) In the prediction violation phase (Day 3), new object pictures were inserted into the sequence of objects, either instead of the second object in the pair (violation) or after the second object in a pair (non‐violation). (C) Following the violation phase, participants completed an item recognition memory test (Day 3) where they were presented with violation and non‐violation targets, similar lures, or new items, and asked to indicate whether each item was old, similar, or new. Memory for the original predictive pair was also tested (associative memory) by presenting participants with the first object in a pair and asking them to identify which of three objects followed the top object.

FIGURE 2

Response rates and classification index. The raincloud plot shows the distribution of response rates and classification index for violation and non‐violation conditions. (A) Proportion of old responses to old items. (B) The proportion of correct responses (true positives and true negatives) out of all instances. The box plots display the median, interquartile range, and 95% confidence interval for each group, while the density plots show the distribution of the data points for each experimental condition. The individual data points are displayed as scatter plots.

FIGURE 3

P3 component during the violation phase. Stimulus‐locked ERPs during the prediction violation phase. (A) Color‐coded ERP grand average recorded at centroparietal electrodes with highlighted time window in gray. (B) Topographical map plot of violation minus non‐violation difference in the P3 time window.

FIGURE 4

Average ERP amplitude values. Average ERP amplitude values for each condition within the relevant component‐specific time windows. Error bars represent the within‐participant standard error of the mean. (A) Mean amplitude values of P3 component during the violation phase. (B) Mean amplitude values of LPC during the item recognition phase. **p < 0.01.

FIGURE 5

LPC during the item recognition phase. Stimulus‐locked ERPs during the item recognition phase. (A) Color‐coded ERP grand average recorded at parietooccipital electrodes with highlighted time window in gray. (B) Topographical map plot of remembered violation minus forgotten violation difference in the recollection component time window.

FIGURE 6

FN400 during the item recognition phase. Stimulus‐locked ERPs during the item recognition phase. (A) Color‐coded ERP grand average recorded at frontocentral electrodes with highlighted time window in gray. (B) Topographical map plot of remembered violation minus forgotten violation difference in the recollection component time window.