Electrophysiological differences in older and younger adults’ anaphoric but not cataphoric pronoun processing in the absence of age-related behavioural slowdown

The study addresses whether syntactic processing, specifically pronoun resolution, declines with age and if any decline reflects online interpretive (per-millisecond) processes or post-interpretive decision processes. Prior work offers mixed evidence: some studies report slower and less accurate performance in older adults on sentence judgment tasks, particularly under high working memory (WM) load, whereas other neuroimaging studies suggest preserved syntactic processing accompanied by compensatory recruitment of additional neural regions. Most prior designs emphasize offline judgments and English/Germanic languages, leaving the time-course of online processing and cross-linguistic generalization (e.g., French clitic pronouns) less explored. Using ERPs, the authors test French speakers’ online processing of anaphoric (referent-first) and cataphoric (pronoun-first) pronouns, predicting similar behavioral performance across ages if syntactic processing is resilient, potential ERP differences reflecting compensatory mechanisms in older adults, and possible asymmetries between anaphoric and cataphoric processing due to differences in memory storage/retrieval vs. antecedent search.

Background work shows that pronoun resolution often elicits a P600 in gender-mismatch paradigms, interpreted as syntactic reanalysis/integration and referential processing. Cataphoric pronoun studies are fewer; some report anterior negativities under constrained dependencies, interpreted as failures to find appropriate antecedents. Ageing literature indicates older adults may show slower/more error-prone behavior, especially with higher WM demands, but neuroimaging often reveals preserved syntax with compensatory recruitment (e.g., bilateral or right-hemisphere homologues of left IFG). ERP studies sometimes find reduced/delayed components in older adults (e.g., N400), though pronoun-related P600 can be comparable across ages; WM load may modulate amplitudes, particularly in older adults.

Design: ERP/EEG study examining anaphoric and cataphoric pronoun processing with gender match/mismatch manipulations. Participants: 33 native French speakers after preprocessing exclusions (young n=18, 11F, age 19–35, mean 22.05; older n=15, 11F, age 57–88, mean 65.73). All right-handed (Edinburgh inventory >70%), normal/corrected vision, MMSE >23. Cognitive measures: verbal STM (vSTM; WAIS forward digit span), non-verbal STM (Corsi block-tapping), print exposure (Author Recognition Task). Older group showed lower vSTM than younger; no group differences in nvSTM or print exposure. Materials: 52 base sentences, each realized in four conditions: Anaphor match/mismatch and Cataphor match/mismatch (total 208 items). Manipulated gender agreement between clitic subject pronouns (il/elle) and proper-name antecedents. Male/female names counterbalanced (half female, n=26), selected to be stereotypically gendered. Pretests assessed name gender stereotypes, pronoun bias, and cloze probability. Procedure: Word-by-word moving-window visual presentation (Courier New 32pt): each word 500 ms with 350 ms blank. End-of-sentence acceptability judgment (j=yes, f=no). Latin-square distribution over two lists; each participant viewed 104 experimental sentences (26 per condition) across eight balanced blocks. Anaphor and cataphor variants of a sentence appeared in different blocks; participants saw either match or mismatch per item. Session duration 50–70 min (~2 h including EEG setup). EEG acquisition: 64-channel 10–20 Quick-Cap (Neuroscan amplifier), ground near Fz; horizontal/vertical EOG and mastoid references recorded. Sampling 1000 Hz; impedances <10 kΩ. Preprocessing in EEGLAB: down-sampled to 256 Hz, re-referenced to mastoids, band-pass 0.1–46 Hz. Epochs -400 to 1600 ms time-locked to the later pronoun/referent presentation; baseline -400 ms. Artifact rejection >100 µV (mean rejection: young 8.4%, older 9.2%; <20% per condition/subject). ICA (Picard) and ICLabel used to remove ocular/muscle/line-noise components. ERP analysis: Electrodes grouped into 15 ROIs (frontal, fronto-central, central, central-parietal, parietal, parietal-occipital; left/midline/right). Mean amplitudes analyzed in 300–500 ms, 500–700 ms, 700–900 ms windows. Omnibus rmANOVA: Condition (Anaphor vs Cataphor) × Mismatch (Match vs Mismatch) × Group (Older vs Younger) × Region (15). Follow-up rmANOVAs per condition (Mismatch × Group × Region). Linear regressions probed modulation of P600 amplitudes by vSTM. Behavioral analysis: Mixed-effects models in R: log RTs (linear) and accuracy (logistic), with fixed effects Group, Condition, Mismatch (sum-coded) and vSTM; random intercepts (and slopes where applicable) for subjects and items.

Behavioral: Older adults were not less accurate or slower than younger adults on mismatch trials. Accuracy: Older slightly higher overall (85% vs 81%); significant Group × Condition interaction driven by better performance by older in cataphor conditions (86% vs 78%; B=1.01, SE=0.35, z=3.082, p=0.002), no difference in anaphor (85% vs 83%; B=0.43, SE=0.34, z=1.25, p=0.21). vSTM significantly modulated mismatch detection accuracy with a Group × Mismatch × vSTM interaction: older with lower vSTM detected mismatches more poorly (B=0.89, SE=0.17, z=5.17, p<0.001); effect smaller in younger (B=0.25, SE=0.08, z=2.96, p=0.003). RTs: Main effects of Mismatch (mismatch faster: 1786 ms vs match 2151 ms) and Condition (cataphor slower: 2091 ms vs anaphor 1853 ms); no main Group effect. Group × Mismatch interaction: older slower than younger on match (2351 vs 1975 ms; B=0.34, SE=0.15, p=0.02), no significant group difference on mismatch (1999 vs 1599 ms; p=0.15). ERPs: Both anaphoric and cataphoric mismatches elicited a central/parietal P600 with similar timing and amplitude across age groups. Omnibus ANOVAs: 500–700 ms and 700–900 ms windows showed main effects of Mismatch (mismatch > match) and Condition (anaphor > cataphor), with some interactions involving Region but not robust Group × Mismatch interactions. Anaphoric conditions: Clear P600; onset ~300–370 ms posteriorly; more sustained positivity in younger beyond 900–1000 ms. Older adults exhibited an additional anterior negativity over frontal/fronto-central ROIs (FL, FR, FCL, FCR) in 500–700 ms for mismatches, absent in younger (post-hoc t-tests over frontal ROIs p≤0.02). P600 amplitudes (mismatch–match) in older correlated positively with vSTM over PL (β=1.51, SE=0.55, t=2.73, p=0.017), CPL (β=1.84, SE=0.52, t=3.50, p=0.003), CPR (β=1.60, SE=0.57, t=2.79, p=0.015); no such modulation in younger. Cataphoric conditions: Robust P600 (onset ~450–480 ms) without frontal negativity; amplitudes and topography similar across groups; no relationship with vSTM (all ps>0.46). Early N170 (100–170 ms) observed across conditions, reflecting orthographic processing.

Findings indicate that online syntactic/pronoun processing, as indexed by P600 effects, is preserved across age, with similar timing and amplitude for both anaphoric and cataphoric mismatches. Behaviorally, older adults matched or exceeded younger adults’ accuracy (notably for cataphors) and did not show generalized slowing on mismatch trials, suggesting no age-related decline in the core syntactic/referential process. However, older adults uniquely showed an anterior frontal negativity for anaphoric mismatches, consistent with additional recruitment of frontal resources during antecedent retrieval and increased memory demands. The P600 amplitude in older adults scaled with verbal short-term memory for anaphoric mismatches, indicating individual differences in compensatory capacity; younger adults showed no such modulation. Cataphoric processing evoked a P600 comparable across groups without frontal negativity, and elicited longer decision RTs and smaller positivity than anaphoric conditions, supporting asymmetries between storage/retrieval (anaphor) and antecedent search (cataphor) mechanisms. Overall, results support resilience of syntactic processing in healthy aging, with compensatory neural recruitment when demands are higher (anaphora), and highlight the role of vSTM in older adults’ neural responses and accuracy.

Pronoun processing during sentence comprehension is not measurably compromised in healthy older adults: both age groups show robust P600 effects to gender mismatches in anaphoric and cataphoric dependencies and comparable behavioral performance. Older adults, however, recruit additional frontal mechanisms during anaphoric processing (anterior negativity) and exhibit vSTM-dependent modulation of P600 amplitudes, suggesting compensation for increased memory demands associated with antecedent retrieval. Cataphoric mismatches elicit smaller P600s and longer judgment RTs than anaphoric ones, indicating processing asymmetries across dependency types. Future work should localize compensatory regions via functional imaging, examine broader domain-general cognitive contributions, extend to other languages and structures, and further characterize individual differences in memory capacities.

The study did not include functional imaging, so precise anatomical localization of the additional recruited regions is unknown. Domain-general cognitive measures beyond memory were not comprehensively assessed, limiting interpretation of post-interpretive processes. Sample size was modest. Results are based on French clitic pronouns, which may limit generalizability to other languages and pronoun systems. Individual variability, particularly in verbal short-term memory, indicates that resilience is not uniform across older adults.