Children exhibit superior memory for attended but outdated information compared to adults

The study investigates how memory selection—specifically, the ability to prevent attended but outdated information from entering working memory—develops from early childhood to adulthood. Traditional accounts position attention as a gateway to working memory, implying that attended information is automatically stored. However, findings from attribute amnesia show that even fully attended information can fail to be consolidated into working memory, indicating a subsequent memory selection stage beyond attention. Developmentally, attentional selectivity improves from childhood to adulthood, with children often showing broader attention and increased processing of task-irrelevant information. The current research asks whether memory selection follows a similar developmental trajectory. Using attribute amnesia paradigms in which a key feature is attended and used for the task but is not required for later report, the authors compare young children and adults to test whether differences in surprise-memory performance reflect developmental differences in memory selection rather than attentional allocation.

Prior work shows that selective attention matures with age, with adults better at focusing on relevant information and filtering distractors, whereas children often process more task-irrelevant input (e.g., flanker effects, change detection, and visual search). Classical theories proposed that attended information is automatically encoded into working memory. Attribute amnesia studies challenge this, demonstrating that a fully attended key feature can be poorly reported on a surprise test due to lack of consolidation, implying a reselection process in working memory (memory selection). Research suggests active inhibition can block attended but outdated information from memory, a process linked to executive control, which itself matures over childhood. Developmental studies of inhibition (e.g., directed forgetting, Think/No-Think) show children’s reduced capacity to suppress irrelevant or outdated information. This background motivates testing whether children’s broader attention extends to weaker memory selection, leading to greater retention of attended but outdated features.

Design: Five experiments using an attribute amnesia framework compared adults and young children. In pre-surprise trials, participants performed a location task that required attending to a key feature to find a target. In a surprise trial, participants were unexpectedly asked to report the key feature (identity). Subsequent control trials, identical to the surprise trial, assessed performance when memory for the key feature was explicitly required. Participants: Convenience samples from Zhejiang University (adults) and a kindergarten in Hangzhou, China (children). Experiments 1–4: N=20 adults and N=20 children per experiment; Experiment 5: N=50 per group. Ages: adults ~19–24 years; children ~5.5–6.1 years. Inclusion: normal or corrected-to-normal vision; no reported developmental issues for children. Ethics approval obtained; parental/adult consent provided. Some participants in Exp.4–5 were replaced for eye-tracking failure or low pre-surprise accuracy. Apparatus and software: MATLAB with Psychtoolbox on a 14-inch 60 Hz monitor (1024×768). Viewing distance ~50 cm. Background medium gray. Adults responded via keyboard; children gave verbal responses recorded by experimenters (except in Experiment 3, where both were verbally queried in the surprise test). Investigators were not blinded. Tasks and stimuli:

• Experiment 1: Four placeholders at corners of an invisible square (5.2°×5.2°). Trial: fixation (0.6°) among placeholders, then array (250 ms) with one target English letter (A, B, D, or E; 0.6°×0.8°) and three distractor digits (2–9; 0.6°×0.8°), followed by 500-ms fixation. Location task in 11 pre-surprise trials; 12th trial was the surprise identity query; then four identity control trials. Adults read prompts and responded by keys; children answered verbally. Eight practice trials at slower pace.
• Experiment 2 (pre-registered on OSF; March 18, 2023): Target stimuli were four highly familiar animals (chicken, dog, rabbit, cat) selected via a prior familiarity survey with 20 children; distractors were clothing items. Stimuli displayed at corners of a larger square (7.8°×7.8°), presented for 1000 ms followed by 400-ms blank. Surprise and control identity tests probed the animal identity.
• Experiment 3: Target was a number larger than five (6,7,8,9) among distractors smaller than five (1–4), requiring numerical quantity access. Stimulus duration 1000 ms. In the surprise test, the experimenter directly asked both adults and children to report the target number identity.
• Experiment 4: Same as Experiment 3, with eye-tracking (EyeLink Portable Duo, 1000 Hz, left eye). 5-point calibration, fixation-required trial start, recalibration as needed. Stimuli placed on 7.8°×7.8° square. Total 32 trials: 27 pre-surprise, 1 surprise, 4 control.
• Experiment 5: Replication of Experiment 2 with larger samples (N=50 per group), identical procedures; not pre-registered. Analyses: Accuracy on location tasks (pre-surprise and surprise trial) and identity report in surprise and control trials. Chi-square tests for proportion comparisons; Bayes factors reported for nonsignificant outcomes. For eye-tracking (Experiment 4), analyzed proportion of fixation duration on target vs distractor areas during pre-surprise arrays; two-sided t-tests and ANOVAs used; Shapiro–Wilk normality tests used; Greenhouse–Geisser corrections applied as necessary. OSF links provided for stimuli familiarity scoring and data. Power analyses with G*Power informed sample sizes (ω=0.49 for attribute amnesia; ω=0.30 conservative estimate for Experiment 5).

Across five experiments, children consistently outperformed adults in reporting the attended but outdated key feature on the surprise test, while both groups performed well on the primary location task in pre-surprise trials.

• Experiment 1 (N=20 per group): Pre-surprise location accuracy: adults 99%, children 94%. Surprise identity accuracy: adults 50% (10/20), significantly lower than Control1 90% (χ²(1,N=40)=7.619, p=0.006, φ=0.436); children 80% (16/20), not different from Control1 65% or later controls (BF10≤0.631). Children > adults on surprise identity: 80% vs 50% (χ²(1,N=40)=3.956, p=0.047, φ=0.314).
• Experiment 2 (familiar animal targets; N=20 per group): Surprise identity: adults 55% (11/20) < Control1 95% (χ²=8.533, p=0.003, φ=0.462); children 85% (17/20), not different from controls (BF10 up to 0.806). Children > adults: 85% vs 55% (χ²=4.286, p=0.038, φ=0.327).
• Experiment 3 (number comparison; both groups verbally queried; N=20 per group): Surprise identity: adults 60% (12/20) < Control1 90% (χ²=4.800, p=0.028, φ=0.346); children 90% (18/20), not different from controls (BF10≤0.971). Children > adults: 90% vs 60% (χ²=4.800, p=0.028, φ=0.346).
• Experiment 4 (eye-tracking; N=20 per group): Surprise identity: adults 50% (10/20) < Control1 100% (χ²=13.333, p<0.001, φ=0.577); children 100% (20/20), not different from controls (BF10≤0.806). Children > adults: 100% vs 50% (χ²=13.333, p<0.001, φ=0.577). Eye-tracking: Adults fixated proportionally more on target vs each distractor (57% vs 9%; t(19)=13.652, p<0.001, d=3.053), while children showed broader allocation (40% vs 19%; t(19)=11.123, p<0.001, d=2.487); group difference significant, F(1,38)=42.293, p<0.001, ηρ²=0.527.
• Experiment 5 (larger sample; N=50 per group): Surprise identity: adults 46% (23/50) < Control1 98% (χ²=33.532, p<0.001, φ=0.579); children 86% (43/50), not different from controls (BF10≤0.343). Children > adults: 86% vs 46% (χ²=17.825, p<0.001, φ=0.422). Breslow–Day test showed greater improvement from surprise to Control1 in adults vs children (χ²(1)=7.532, p=0.006), indicating different attribute amnesia patterns between groups. Additional observation: In Experiment 5, children’s location performance in the surprise trial was worse than adults (58% vs 94%), a pattern also seen in earlier experiments, suggesting greater disruption of children’s location memory by the unexpected identity query.

The findings directly address the research question by demonstrating that memory selection—the process preventing attended but outdated information from entering working memory—is underdeveloped in young children. Despite attending to and using the key feature for target localization, adults frequently failed to report it on a surprise test (attribute amnesia), whereas children retained it, yielding a developmental reversal pattern. Eye-tracking confirmed that adults, not children, focused more on the target, ruling out the explanation that children performed better due to greater attentional allocation to the target or its features. Controls in Experiment 3 addressed strategy differences (e.g., categorization without identity access), and procedure adjustments (verbal questioning for both groups) ruled out modality-related memory disruption in adults. Together, these results imply that children’s reduced inhibitory control prevents efficient filtration of attended but irrelevant information at the point of memory consolidation. This has implications for interpreting developmental differences in tasks that infer attention from memory performance: superior memory for irrelevant information in children may reflect weaker memory selection rather than weaker attentional selection alone. The results align with theoretical accounts positing an inhibitory gate that allows goal-relevant information into working memory while blocking off-goal information, a mechanism that develops with executive control. The broader relevance includes understanding how children’s broader, less selective encoding may burden working memory yet facilitate exploration and pattern learning in some contexts. The demonstrated disruption of children’s location memory by the surprise probe suggests their working memory is more susceptible to interference from unexpected events.

This work introduces developmental evidence for memory selection as a distinct facet of cognitive selectivity and shows it is not fully developed in young children. Across five experiments, children consistently outperformed adults in reporting attended but outdated key features on surprise tests, even when stimuli were highly familiar, strategies were controlled, and attention allocation was measured. The study cautions against equating memory performance with attentional allocation in developmental research and underscores the role of inhibitory mechanisms in regulating memory contents. Future research should: (1) directly examine the inhibitory processes that block attended outdated information from entering memory and their developmental trajectory; (2) compare prevention (pre-encoding inhibition) versus removal (post-encoding inhibition) mechanisms to determine shared or distinct developmental pathways; (3) broaden samples and contexts to test generalizability across ages, cultures, and task domains; and (4) integrate multimodal measures (e.g., neurophysiology, finer-grained eye-tracking/feature-level attention indices) to dissociate attentional selection from memory selection.

• Initial small sample sizes (N=20 per group in Experiments 1–4); addressed by a larger replication (Experiment 5), but small-sample concerns may still apply to specific manipulations.
• Convenience sampling from a single university and a single kindergarten in China limits generalizability across cultures and socioeconomic contexts.
• Investigators were not blinded to allocation or outcomes.
• Response modality differed by age group in most experiments (adults: keyboard; children: verbal), although Experiment 3 matched verbal questioning for both groups; residual modality effects cannot be entirely excluded.
• Eye-tracking cannot directly assess attention at the feature level; thus, while target vs distractor allocation was quantified, feature-specific attention to identity was inferred indirectly.
• Only Experiment 2 was pre-registered; other experiments were not.
• Tasks focused on specific visual search/location paradigms within the attribute amnesia framework; results may not generalize to all forms of attended but outdated information or to other sensory modalities.