Value Shapes How Young and Older Brains Prioritize Meaningful Sentences

Every day the brain confronts far more information than it can possibly retain, forcing it to make split-second decisions about what deserves precious cognitive resources. A new study published in Frontiers in Cognition by Yu Min W. Chung and Kara D. Federmeier of the University of Illinois Urbana-Champaign reveals that this triage system operates with surprising sophistication, and that it changes in subtle but measurable ways as people age. Using a clever adaptation of a classic memory paradigm combined with real-time brain recordings, the researchers uncovered a fundamental dissociation between the neural machinery that predicts upcoming words and the machinery that builds meaning from sentences, showing that only the latter responds flexibly to cues about importance.

The research builds on the value-directed remembering paradigm, a well-established tool in memory science in which participants study items each tagged with a point value representing the reward for recalling it. Decades of work with this approach have demonstrated that people can selectively encode and retrieve higher-value items, proving that memory is shaped not only by capacity limits but also by automatic and strategic processes that prioritize important content. What remained unclear was whether this prioritization extends to complex, temporally unfolding material like sentences, and whether the neural processes engaged during comprehension itself are modulated by value or only downstream memory outcomes.

To address these questions, the team ran three experiments. In the first, 48 younger adults aged 18 to 28 read 96 sentences word by word using rapid serial visual presentation, with each word appearing for 200 milliseconds separated by 300-millisecond intervals. Before each sentence, a colored numerical cue appeared for three seconds, signaling whether remembering that sentence would earn 10 points or a mere 3 points. The sentences varied systematically in how strongly they constrained a particular ending: half were strongly constraining, meaning the best completion had a cloze probability of at least 0.68, while half were weakly constraining with cloze probabilities at or below 0.42. Critically, each sentence could end with either its most expected word or an unexpected but plausible alternative with near-zero cloze probability, allowing the researchers to isolate the neural signature of prediction violation from the signature of semantic facilitation.

Electroencephalography captured the brain’s electrical responses to each sentence-final word, and three event-related potential components served as the study’s neural workhorses. The N400, a negative-going potential peaking around 400 milliseconds after word onset, indexes semantic access and shrinks in amplitude when contextual information has already activated features of an incoming word. The post-N400 anterior positivity, observed between roughly 600 and 900 milliseconds over prefrontal sites, appears when readers encounter unexpected but plausible completions of strongly constraining sentences and is thought to reflect the effortful revision of a disconfirmed prediction. Finally, the frontal negativity, also in the 600 to 900 millisecond window over frontal electrodes, is elicited by expected completions and has been linked to deeper message-level integration, such as resolving competition among alternative interpretations or refining the evolving sentence representation.

Behaviorally, both younger and older adults successfully used value information to prioritize memory, recalling significantly more high-value than low-value sentences across all experiments. But the neural data told a more nuanced story. In young adults, the anterior positivity elicited by unexpected endings in strongly constraining sentences was present and robust, yet completely insensitive to value. Whether a sentence was worth 10 points or 3 made no difference to the strength of this prediction-violation signal. This finding aligns with recent work suggesting that prediction may operate more like a state that is either engaged or not, rather than a process that can be graded by attention or effort. Prediction, in other words, appears to run in the background regardless of how important the material is.

In sharp contrast, value reliably modulated the other two components. High-value sentences elicited larger N400 expectancy effects in young adults, indicating that contextual information provided greater semantic facilitation when the material mattered. The frontal negativity also grew larger for high-value sentences, consistent with increased engagement of message-level processing. Together, these patterns suggest that value cues do not simply boost overall processing effort but instead selectively enhance specific components of comprehension, deepening the construction of meaning without altering the degree to which predictive mechanisms are engaged.

Experiment 2 shifted the focus to 96 older adults aged 55 to 81, tested online, and asked whether the timing of value cues shapes memory differently in aging. Half the participants saw the value cue before each sentence, allowing it to guide processing as the sentence unfolded, while the other half saw it after the sentence, when it could only influence later rehearsal or evaluation strategies. Unlike younger adults in prior work, older adults showed relatively similar memory patterns across the two timing conditions, suggesting reduced flexibility in adapting encoding strategies to task structure. Nevertheless, they recalled more high-value sentences and showed elevated false recognition of lure words, the expected completions of sentences that had actually ended unexpectedly, providing behavioral signatures consistent with prediction-based memory distortions.

Experiment 3 then recorded ERPs from 43 older adults in the laboratory while they completed the same value-cued sentence task. The most striking result was the emergence of a clear anterior positivity to prediction violations in this group, a pattern that prior studies using identical materials and passive reading instructions had found to be reduced or absent in older adults. Because participants in the current study were never explicitly instructed to predict, only to maximize their score by prioritizing high-value sentences, the authors argue that the value-directed memory task provided sufficient top-down support to encourage predictive processing in a group that does not routinely engage it. Notably, for low-value sentences, trials that elicited a larger anterior positivity were more likely to be later recalled, suggesting that the effortful revision of a violated prediction can actually benefit memory when participants are not already investing heavily in the material.

Yet the value modulation in older adults was more selective than in their younger counterparts. The frontal negativity was again larger for high-value sentences, and direct statistical comparison showed no reliable age difference in this value-based modulation of message-level processing. However, the N400 expectancy effect, which young adults enhanced for high-value items, showed no value-related modulation in older adults. The authors interpret this asymmetry in terms of the cognitive demands of real-time sentence processing: older adults can strategically engage processes that come online relatively late, such as revising predictions or building message-level representations after encountering a word, but they show deficits in rapidly adjusting the activation states of semantic information, a process that may be more automatic and less amenable to strategic control.

Backsorting the ERPs by subsequent recall revealed little evidence that sentence-final word processing robustly predicted later memory in either age group, indicating that the processes indexed at this single position are not, on their own, sufficient to account for memory outcomes of the full sentence. The authors suggest that memory benefits from different encoding strategies likely accrue across the entire sentence and may also be augmented by post-encoding rehearsal. Taken together, the findings paint a picture of value-based prioritization as a layered phenomenon: prediction operates in a state-like manner largely indifferent to importance, while semantic facilitation and message-level integration respond flexibly to value cues, with older adults preserving the latter capacity even as the former becomes less modifiable. The work underscores that knowing what matters can reshape how the brain builds meaning, but that the reach of this strategic control narrows in specific, principled ways with age.

Subject of Research: Value-directed remembering, predictive sentence processing, and age-related differences in event-related brain potentials during comprehension

Article Title: Prioritizing meaning: value, prediction, and age differences in sentence processing

Article References: Prioritizing meaning: value, prediction, and age differences in sentence processing. (n.d.). https://doi.org/10.3389/fcogn.2026.1862899

Image Credits: AI Generated

DOI: 10.3389/fcogn.2026.1862899

Keywords: value-directed remembering, sentence processing, event-related potentials, N400, prediction, aging, memory, frontal negativity, anterior positivity, semantic facilitation, cognitive control, language comprehension