Impact of an 18‐month non‐native language intervention on episodic memory and the underlying neural mechanisms in cognitively unimpaired older adults Authors: Anne‐Laure Turpin, Caitlin Ware, Brigitte Landeau, Florence Mézenge, Asrar Lehodey, Célia Soussi, Marion Delarue, Oriane Hébert, Vincent de la Sayette, Natalie L Marchant, Gaël Chételat, Julie Gonneaud, for the Medit‐Ageing Research Group Published: 2026 ABSTRACT AbstractINTRODUCTIONEpisodic memory decline impairs daily functioning in aging. This secondary analysis of data from the Age‐Well randomized controlled trial (RCT), we examined whether non‐native language training (NNLT) benefits memory and its neural correlates in older adults.METHODSParticipants from the Age‐Well RCT (aged ≥ 65) were assigned to an 18‐month NNLT (N = 45), or no‐intervention (N = 44) group. Episodic memory (California Verbal Learning Test Second Edition [CVLT‐II] and Logical Memory), and hippocampal and medial prefrontal cortex volumes and perfusion were assessed pre‐ and post‐intervention. Linear mixed models tested visit x group interactions. Regressions examined associations between memory and neuroimaging changes, adjusted for age, sex, education, and baseline memory performance.RESULTSA visit x group interaction was observed (F = 6.77, p = 0.01): the NNLT group maintained CVLT‐II delayed free recall performance, while the no‐intervention group declined. Within the NNLT group, changes in this score correlated positively with changes in hippocampal perfusion.DISCUSSIONThese findings suggest that NNLT may help preserve hippocampus‐dependent long‐term episodic memory in aging.Keywords: aging, cognitive reserve, dementia prevention, episodic memory, foreign language training, healthy older adults, hippocampus, medial prefrontal cortex, non‐pharmacological intervention BACKGROUND Population aging is associated with both cognitive and brain changes, leading to an increase in age‐related decline and neurodegenerative diseases such as Alzheimer's disease (AD). 1 These changes can significantly impact quality of life and the ability to maintain autonomy in daily life. 2 To promote healthy aging, non‐pharmacological interventions aiming to slow cognitive decline or enhance brain health have gained attention. 3 Resilience mechanisms including brain reserve (i.e., neurobiological capital that allows better coping with age‐related changes and pathology), cognitive reserve (i.e., brain's ability to cope with cerebral damage), and brain maintenance (i.e., the absence of biological brain changes), are thought to underlie the inter‐individual differences observed in cognitive and brain aging. 4 These mechanisms may buffer aging effects, supporting optimal cognitive functioning despite brain damage. Importantly, cognitively stimulating activities, such as learning new skills, can enhance these mechanisms and protect against age‐related decline, even in advanced ages. 5 , 6 Bilingualism is a cognitive activity that has been associated with better executive functioning and episodic memory 7 , 8 and that may promote resilience mechanisms over time. 9 Cross‐sectional studies suggest that bilingualism benefits cognitive performance and brain integrity in both young and older adults. 10 , 11 However, most studies focus on lifelong bilinguals, 12 , 13 limiting our understanding of the effects of language acquisition later in life. Thus, late‐life foreign language learning may offer a novel way to promote cognitive health and resilience during aging. 14 , 15 While late‐life language acquisition may not lead to bilingualism, it can still provide cognitive and brain benefits similar to those seen in lifelong bilinguals. Some studies indicate that second language training might enhance attention, executive control, and episodic memory, which are supported by changes in connectivity and activation patterns in frontoparietal, prefrontal, and hippocampal networks. 16 , 17 , 18 Despite some inconsistencies in the literature depending on the methodologies used, few interventional studies have examined cognitive and brain effects of new language learning in older adults over 65. 7 , 18 , 19 Recent systematic reviews 10 , 20 highlight that research on foreign language learning in older adults is still in its infancy. The limited interventional studies in older adults mostly focused on cognitive outcomes, addressing mainly global cognition, executive function, working memory, and attention, and only one study on episodic memory 21 and two investigating functional neuroimaging. 16 , 22 These studies suggest a positive impact on cognitive domains evaluated, but highlight a need for well‐controlled, randomized interventions with longer duration, combining cognition and neuroimaging to better understand the mechanisms behind non‐native language training (NNLT)‐related changes. A secondary analysis of the Age‐Well randomized controlled trial (RCT), which investigated notably the effects of an 18‐month NNLT intervention in older adults (≥ 65) 23 , suggested subtle (trend‐level) benefits on episodic memory. Other cognitive domains evaluated (i.e., global functioning, executive function, and attention) did not show such differences between the NNLT and no‐intervention groups, and will therefore not be further investigated. Specifically, the NNLT group showed less decline in episodic memory, as measured with a composite score, compared to the no‐intervention group, but this difference did not remain significant after correction for multiple comparisons. However, using a composite score prohibited determination of whether any specific aspects of episodic memory were significantly affected. Moreover, the brain substrates associated with changes in episodic memory after NNLT remain to be investigated. Given the major role of the hippocampus and the medial prefrontal cortex (mPFC) in episodic memory processes, 24 , 25 these regions appear as promising candidates to investigate whether any potential memory benefits are supported by functional and structural changes. Indeed, episodic memory relies on different neural substrates, including the hippocampus and the mPFC, which together support the encoding, consolidation, and retrieval of experiences. The hippocampus supports relational binding and reinstatement of item‐context detail (pattern separation and completion), which underlie recollection and the spatiotemporal organization of events. 26 , 27 In other words, it integrates and stabilizes contextual information to form coherent memory traces. In contrast, mPFC contributes to schema‐based integration and top‐down retrieval control and is increasingly engaged in consolidating memories. It contributes to the organization, strategic control, and flexible retrieval of these memories. 28 , 29 In this context, the objective of this study was twofold. First, using the longitudinal data from the Age‐Well trial, our goal was to refine previous findings by investigating which components of the episodic memory composite score may be influenced after an 18‐month NNLT intervention. Second, we aimed to explore the neural mechanisms supporting these memory changes, focusing specifically on their association with structural and/or functional changes in the hippocampus and/or the mPFC. MATERIALS AND METHODS General study design Age‐Well 30 is a monocentric, observer‐masked, randomized clinical trial with three parallel arms: an 18‐month meditation training intervention arm, an 18‐month NNLT (English) intervention arm, and an arm with no intervention. This trial includes 137 cognitively unimpaired older adults at baseline, recruited from the general population between November 2016 and March 2018 in Caen, France. The Age‐Well RCT's primary objective was to evaluate the effect of an 18‐month meditation intervention in cognitively unimpaired older adults compared to an 18‐month NNLT program and a no‐intervention group. 30 Participants completed a pre‐screening visit, with eligible participants invited to a baseline pre‐intervention visit, before being randomized. Participants were randomly assigned (1:1:1) according to a randomization list with permuted blocks of varying size (Methods S1 in supporting information). There were no major clinical differences in any demographic or clinical characteristics between arms. 31 A post‐intervention visit was performed after 18 months. The baseline and follow‐up visits comprised multimodal assessments, including cognitive and neuroimaging evaluations. Standard protocol approvals, registrations, and patient consents The Age‐Well trial, sponsored by INSERM, was approved by the ethics committee (CPP Nord‐Ouest III, Caen; ClinicalTrials.gov Identifier: NCT02977819; trial registration number: EudraCT: 2016‐002441‐36; IDRCB: 2016‐A01767‐44; registration date: 2016‐11‐25) and all participants gave written informed consent prior to participation. Design of the present study A total of 137 healthy older adults were randomized. Two participants were excluded for not meeting eligibility criteria, resulting in a sample of 135 participants who underwent cognitive and neuroimaging assessments at baseline and follow‐up (see flowchart in Figure S1 in supporting information). In the present study, we only focused on longitudinal episodic memory changes and their underlying neural mechanisms in the NNLT (N = 45) and the no‐intervention (N = 44) groups. All participants were aged ≥ 65 years old, native French speakers, retired for at least 1 year, and had at least 7 years of formal education. They performed within the normal range on standardized cognitive tests. Due to the characteristics of the trial, they must not have had present or past regular or intensive practice of meditation or comparable practices, nor speak fluent English, as measured using a placement test (i.e., individuals demonstrating a high level of proficiency were excluded). The main exclusion criteria consisted of a history of major neurological or psychiatric disorders, the presence of a chronic illness or an acute unstable condition, and current/recent treatments that might affect cognitive functioning. Visual and auditory impairments were not systematically documented at baseline; however, any such deficits were confirmed to be corrected prior to cognitive assessment. Full inclusion/exclusion criteria are available in supporting information (Table S1). The NNLT The NNLT program is fully detailed in supporting information (Methods S1). Briefly, the NNLT was a cognitively stimulating intervention consisting of English exercises designed to reinforce English comprehension (oral and written), writing, and speaking skills. It followed a structured curriculum and was delivered by experienced and highly trained instructors. The sessions were tailored to baseline English proficiency levels, including activities targeting oral comprehension and expression to facilitate vocabulary acquisition as well as grammatical structures. The intervention followed a standardized protocol and consisted of weekly 2‐hour group sessions, daily home practice (minimum 20 minutes), and a 5‐hour intensive practice session (i.e., a trip to Jersey) held once during the 18‐month program. Participants were encouraged to fully engage in all activities specific to this intervention. After the intervention, participants were classified as responders or non‐responders based on a scale completed by teachers, assessing their perception of each participant's intervention benefit, and the improvement on an English language test. Details of the responder/non‐responder classification are provided in supporting information (Methods S1). The no‐intervention group Participants in the no‐intervention group were requested not to change their habit and continue living as they used to before engaging in the study and throughout the intervention period. They were specifically asked not to engage in meditation or NNLT. 30 Cognitive measures All participants underwent a detailed neuropsychological assessment in their native language (i.e., French). Full details on neuropsychological tests are provided in supporting information (Methods S2). This study focused on the different verbal episodic memory scores included in the episodic memory composite score, namely, California Verbal Learning Test Second Edition (CVLT)‐II sum of trials 1 to 5, immediate free recall and delayed free recall, as well as the immediate and delayed recall of the Wechsler Memory Scale (WMS)‐IV Logical Memory test (story B). Neuroimaging data Neuroimaging acquisition and processing are fully detailed in supporting information (Methods S3). Briefly, participants were all scanned on the same magnetic resonance imaging (MRI) and positron emission tomography (PET) cameras at the Cyceron Center (Caen, France) at baseline and after 18 months. High‐resolution T1‐weighted MRI were acquired in all participants and early florbetapir‐PET scan, performed to measure brain perfusion, was acquired in all participants except one. Gray matter volume and perfusion in the hippocampus and the mPFC were extracted at baseline and follow‐up. Changes in gray matter volume and cerebral perfusion were computed as the difference (delta) between follow‐up and baseline values. Statistical analysis First, linear mixed models (LMMs) with participant‐level random intercepts, estimated via restricted maximum likelihood, were fitted for each episodic memory score between the NNLT and the no‐intervention group to evaluate intervention effects. LMMs were conducted with RStudio using lme4, lmerTest, and emmeans packages. The first two packages fit mixed‐effects models and assess between‐group differences, while the latter also allows post hoc pairwise comparisons and estimation of within‐group changes based on the predictions from LMM. Baseline age, sex, and education were included as fixed effects, with continuous variables mean‐centered. For episodic memory scores showing a significant group x visit interaction, we further assessed whether episodic memory changes in the NNLT group was associated with changes in hippocampal/mPFC volume or perfusion. To do so, we conducted multiple linear regressions in RStudio, using changes in cognitive performance (delta) as the outcome and changes in volume/perfusion, in both the hippocampus and the mPFC, as predictors. For the sake of interpretation, the same analyses were conducted in the no‐intervention group. These analyses were corrected for baseline age, sex, education, and baseline episodic memory score. All analyses were considered significant at p < 0.05, without corrections for multiple comparisons. Sensitivity analyses were conducted to assess the influence of non‐responders on the associations (i.e., participant showing minimal engagement and benefit from the intervention). All the above‐described analyses were replicated after excluding the two non‐responders in the NNLT arm. Data availability statement Data are available upon request after signing a formal data sharing agreement and approval by the Medit‐Ageing consortium and executive committee. The data sharing request form can be downloaded online (https://silversantestudy.eu/2020/09/25/datasharing). RESULTS Demographic characteristics of participants included in the longitudinal analyses are summarized in Table 1. Out of the 135 eligible participants, our analyses focused exclusively on those randomized to the NNLT group (N = 46) and the no‐intervention group (N = 45). Two participants were excluded for not meeting eligibility criteria in the NNLT group and one non–study‐related death was reported at follow‐up in the no‐intervention group. One participant was randomized to the no‐intervention group but attended NNLT and was therefore analyzed within the NNLT arm. Thus, analyses were conducted in 45 participants in the NNLT and 44 in the no‐intervention group (see flowchart in Figure S1). TABLE 1Participant characteristics.CharacteristicsNNLT groupNo‐intervention groupSample size (N)4544Age (years)70.38 ± 4.5267.97 ± 2.55Education (years)12.16 ± 3.0214.36 ± 2.87Sex, N (% M/F)20/25 (44.44/55.56)18/26 (40.91/59.09)MMSE (/30)29.98 ± 1.0129.2 ± 0.90 Notes: Demographic variables for the participants of the NNLT group (N = 45) and the no‐intervention group (N = 44). Values represent mean ± standard deviation or number of participants.Abbreviations: MMSE, Mini‐Mental State Examination; NNLT, non‐native language training. Participant characteristics. Notes: Demographic variables for the participants of the NNLT group (N = 45) and the no‐intervention group (N = 44). Values represent mean ± standard deviation or number of participants. Abbreviations: MMSE, Mini‐Mental State Examination; NNLT, non‐native language training. Longitudinal cognitive changes after an 18‐month NNLT intervention LMM revealed a visit x group interaction for CVLT long‐term memory (i.e., delayed free recall) score (F interaction = 6.77, p interaction = 0.01; Table 2, Figure 1). No visit x group interaction was found for the other episodic memory scores (p interaction > 0.05; Table 2, Figure 1). TABLE 2Statistics of the between‐group changes from the linear mixed models.Interaction group x visitDifference in change: NNLT vs. no interventionEpisodic memory outcomesVisit F value p b (95% CI) p CVLT II sum of trials (1–5) scoreV3‐V12.310.132.32 (−0.746 ; 5.39)0.14CVLT II delayed free recall scoreV3‐V16.77 0.01 1.24 (0.282 ; 2.20) 0.01 CVLT II immediate free recall scoreV3‐V10.040.840.09 (−0.84; 1.02)0.85Logical Memory (story B) immediate recall scoreV3‐V10.340.560.40 (−0.96 ; 1.75)0.56Logical Memory (story B) delayed recall scoreV3‐V13.910.0511.27 (−0.02; 2.56)0.054 Notes: Values are expressed as estimated change (b, 95% confidence interval). All analyses were adjusted for age, sex, and education. For between‐group differences, negative values reflect the decrease in the no‐intervention group from baseline (V1, pre‐intervention) to follow‐up (V3) compared to the NNLT group. The significant values (p < 0.05) were originally highlighted in bold.Abbreviations: CI, confidence interval; CVLT, California Verbal Learning Test; NNLT, non‐native language training. Statistics of the between‐group changes from the linear mixed models. Notes: Values are expressed as estimated change (b, 95% confidence interval). All analyses were adjusted for age, sex, and education. For between‐group differences, negative values reflect the decrease in the no‐intervention group from baseline (V1, pre‐intervention) to follow‐up (V3) compared to the NNLT group. The significant values (p < 0.05) were originally highlighted in bold. Abbreviations: CI, confidence interval; CVLT, California Verbal Learning Test; NNLT, non‐native language training. Episodic memory scores at baseline and follow‐up in the NNLT and the no‐intervention groups. Linear mixed models show no significant interaction between groups and visits for the CVLT‐II sum of trials, CVLT‐II immediate free recall, Logical Memory immediate recall, and delayed recall scores, controlling for age, sex, and education. For each episodic memory score, boxplots show medians and interquartile ranges at pre‐ and post‐intervention for both the NNLT and no‐intervention groups, while lines connect the means for each group at each time point. Only a significant interaction between group and visits was observed for the CVLT‐II delayed free recall score (CVLT long‐term score). The CVLT‐II delayed free recall score significantly decreased from baseline (pre‐intervention) to follow‐up (post‐intervention) in the no‐intervention group only. * p < 0.05. CVLT, California Verbal Learning Test; NNLT, non‐native language training. Considering the visit x group interaction for the CVLT‐II delayed free recall score, post hoc analyses revealed that the no‐intervention group showed a decline in this score between baseline and follow‐up timepoints (p = 0.02; b = −0.84, 95% confidence interval [CI: −1.52; −0.16], Table 3) whereas the NNLT group showed no significant changes over time (p = 0.24; b = 0.40, 95% CI [−0.27; 1.07], Table 3). TABLE 3Statistics of the within‐group changes of the linear mixed models.Estimated changeNNLT groupNo‐intervention groupEpisodic memory outcomesVisit b (95% CI) p b (95% CI) p CVLT II sum of trials (1–5) scoreV3‐V12.78 (0.62; 4.94) 0.01 0.45 (−1.73; 2.6)0.68CVLT II delayed free recall scoreV3‐V10.40 (−0.27; 1.07)0.24−0.84 (−1.52; −0.16) 0.02 CVLT II immediate free recall scoreV3‐V10.00 (‐.65; .65)1.00−0.09 (−0.75; 0.57)0.75Logical Memory (story B) immediate recall scoreV3‐V1−0.38 (−1.33; −0.57)0.43−0.78 (−1.73; 0.19)0.11Logical Memory (story B) delayed recall scoreV3‐V10.11 (−0.77; 1.04)0.77−1.14 (−2.05; −0.22) 0.02 Notes: Values are expressed as estimated change (b, 95% confidence interval). All analyses were adjusted for age, sex, and education. For within‐group standardized estimated changes, positive values reflect improvement within a trial group from baseline (V1) to follow‐up (V3); negative coefficients indicate the opposite. The significant values (p < 0.05) were originally highlighted in bold.Abbreviations: CI, confidence interval; CVLT, California Verbal Learning Test; NNLT, non‐native language training. Statistics of the within‐group changes of the linear mixed models. Notes: Values are expressed as estimated change (b, 95% confidence interval). All analyses were adjusted for age, sex, and education. For within‐group standardized estimated changes, positive values reflect improvement within a trial group from baseline (V1) to follow‐up (V3); negative coefficients indicate the opposite. The significant values (p < 0.05) were originally highlighted in bold. Abbreviations: CI, confidence interval; CVLT, California Verbal Learning Test; NNLT, non‐native language training. Association between changes in the CVLT‐II delayed free recall score and changes in the hippocampal and mPFC volume/perfusion Changes in CVLT‐II delayed free recall score were associated with changes in hippocampal perfusion (p = 0.03, b = 24.66, 95% CI [1.91; 47.40]; Table 4, Figure S2 in supporting information). By contrast, changes in CVLT‐II delayed free recall were not associated with changes in hippocampal gray matter volume (p = 0.79; b = 1.62, 95% CI [−10.38; 13.62], Table 4, Figure S2), mPFC volume (p = 0.93; b = 0.09, 95% CI [−1.29; 1.44]; Table 4, Figure S2 ), or mPFC perfusion (p = 0.32; b = 11.59, 95% CI [−11.71; 34.88]; Table 4, Figure S2). To better interpret the results, the same analyses were conducted in the no‐intervention group and revealed no association between changes in the CVLT‐II delayed free recall score and changes in the hippocampal and mPFC volume/perfusion (ps > 0.30; Table S2 in supporting information, Figure S2). TABLE 4Association between changes in CVLT long‐term score and changes in gray matter volume and cerebral perfusion in the hippocampus and the mPFC in the NNLT group. NNLT group Delta CVLT delayed free recall score b (95% CI) t p Changes in gray matter volumeHippocampal volume1.62 (−10.38; 13.62)0.270.79mPFC volume0.06 (−1.29; 1.43)0.090.93Changes in cerebral perfusionHippocampal perfusion24.66 (1.91; 47.40)2.19 0.03 mPFC perfusion11.59 (−11.71; 34.88)1.000.32 Notes: Statistical values were obtained from multiple linear regressions including changes in volume or cerebral perfusion of the hippocampus and the mPFC as predictors of delta CVLT long‐term score in the NNLT. The significant values (p < 0.05) were originally highlighted in bold.Multiple linear regressions were adjusted for age, sex, education and baseline CVLT long‐term score.Abbreviations: CI, confidence interval; CVLT, California Verbal Learning Test; mPFC, medial prefrontal cortex; NNLT, non‐native language training. Association between changes in CVLT long‐term score and changes in gray matter volume and cerebral perfusion in the hippocampus and the mPFC in the NNLT group. Notes: Statistical values were obtained from multiple linear regressions including changes in volume or cerebral perfusion of the hippocampus and the mPFC as predictors of delta CVLT long‐term score in the NNLT. The significant values (p < 0.05) were originally highlighted in bold. Multiple linear regressions were adjusted for age, sex, education and baseline CVLT long‐term score. Abbreviations: CI, confidence interval; CVLT, California Verbal Learning Test; mPFC, medial prefrontal cortex; NNLT, non‐native language training. Sensitivity analyses Two participants were classified as non‐responders to the NNLT. Restricting analyses to responders yielded essentially the same findings concerning the CVLT‐II delayed recall score (Tables S3–S5 in supporting information). The only exception was that a visit x group interaction was also observed for the Logical Memory delayed recall (F interaction = 4.32, p interaction = 0.04; Table S3). Post hoc analyses revealed no change in the NNLT group (p = 0.71, b = 0.22, 95% CI [−0.71; 1.14]; Table S4) but a decrease of performance in the no‐intervention group (p = 0.02, b = −1.14, 95% CI [−2.07; 0.22]; Table S4). However, these effects were not associated with any brain changes (ps > 0.10; Table S6 in supporting information). DISCUSSION This study aimed to refine previous findings about the effects of an 18‐month NNLT intervention in cognitively unimpaired seniors. Specifically, we investigated the effects of NNLT on different aspects of verbal episodic memory, and related brain changes, particularly in the hippocampus and the mPFC. Our results revealed a group x visit interaction on a measure of longer term episodic memory: both groups performed similarly at baseline, but after an 18‐month intervention, the NNLT group maintained its score while the no‐intervention group declined. This association remained when analyses were restricted to responders. These findings align with some studies showing a maintenance of cognitive functioning among older adults learning a new language compared to a control group. 16 Moreover, our observations go further by suggesting that NNLT may protect episodic memory, particularly long‐term memory, consistent with previous research. 32 , 33 , 34 NNLT engages multiple cognitive processes (i.e., working memory, encoding of new information, recall strategies, and information integration) that may strengthen brain networks underlying episodic memory, particularly those supporting contextual and temporal retrieval. 35 , 36 Furthermore, the changes in the CVLT long‐term score were positively associated with hippocampal perfusion, indicating that individuals who improved the most in memory performance were also those who showed the most increased perfusion in this region, conversely. This association was NNTL specific, as no associations were evidenced in the no‐intervention group. First, our findings align with research suggesting that engagement in language learning may be associated with benefits in episodic memory performance. 21 , 37 Learning a foreign language in late‐life may thus provide cognitive benefits similar to those associated with bilingualism and represents a form of cognitively stimulating activity that supports memory preservation and the maintenance of cognitive reserve in aging, potentially delaying the onset of dementia. 38 , 39 Second, the specific involvement of the hippocampus suggests that language learning enhances episodic memory via hippocampus‐dependent detail/recollection rather than mPFC‐mediated schema‐based integration or top‐down retrieval control. In particular, language learning may enhance relational binding and the reinstatement of item‐context information, processes by which the hippocampus links elements of an experience—such as words, meanings, and context—into coherent memory traces, while the mPFC supports strategic retrieval. These hippocampus‐dependent mechanisms may support efficient encoding and flexible retrieval of linguistic knowledge underlying recollection and the spatiotemporal organization of events. 27 , 40 Such processes are central to language learning programs, and our NNLT intervention, with its emphasis on acquisition of vocabulary, sentence structures, contextual meanings, and active recall, may have further strengthened them, 10 , 41 contributing to the observed maintenance of hippocampal‐dependent long‐term episodic memory. Moreover, some studies suggest that bilingualism appears to enhance cognitive reserve via functional brain changes before structural modifications. 33 Our results, in line with previous studies linking brain perfusion to episodic memory, 33 , 34 indicate that language learning could similarly promote cognitive reserve through functional mechanisms supporting memory performance in regions involved in memory retrieval. 36 , 42 When analyses were restricted to responder participants, the association between changes in CVLT delayed recall and hippocampal perfusion remained significant in the NNLT group. We also observed a visit x group interaction for Logical Memory delayed recall with maintenance of performance in the NNLT group, while the no‐intervention group declined. However, changes in Logical Memory performance were not associated with hippocampal or mPFC volume/perfusion. This suggests that this measure may be less sensitive for detecting intervention effects (i.e., significant only when considering responder participants). In addition, it may rely on neural mechanisms not captured by the regional measures assessed here, such as network‐level or coupling‐based processes. Future studies using connectivity‐based approaches may help clarify these mechanisms. Overall, both CVLT and Logical Memory indices (long‐term recall) showed maintenance over the 18‐month intervention, which may reflect mitigation of age‐related decline. 43 Alternatively, ceiling effects related to high baseline performance in our highly educated sample might lead to a limited potential for improvement. 44 In contrast, no association was observed between changes in CVLT‐II scores and mPFC perfusion or with hippocampal/mPFC structural measures. This finding contrasts with some studies suggesting that learning a foreign language may improve brain integrity and increase brain volume in cognitively unimpaired older adults. 18 , 45 It is possible that significant long‐term structural benefits may be noticeable in old age if such an intervention is carried out at an earlier period of life or for a longer period of time. 46 The absence of structural correlates may then indicate that functional brain changes precede structural adaptations in response to cognitively stimulating activities such as language learning. Regarding the mPFC, the lack of perfusion‐related effects may reflect subtler or more diffuse contributions of this region to episodic memory, which might be better captured by other cognitive or functional measures. Taken together, these findings suggest that an 18‐month NNLT intervention may help preserve long‐term memory performance in healthy older adults, likely supported by hippocampus‐dependent mechanisms—specifically increased hippocampal perfusion—rather than mPFC structural or functional changes. STRENGTHS AND LIMITATIONS One of the main strengths of this study is the 18‐month duration of the NNLT intervention. As highlighted previously, 10 the Age‐Well dataset is valuable for testing cognitive interventions in healthy older adults, partially because its interventions are longer than in most comparable studies in the field. 19 , 47 Thus, this study allows analysis of long‐term effects of sustained cognitive activity, including stabilization or amplification of initial benefits over time. Moreover, the inclusion of neuropsychological and neuroimaging (hippocampal/mPFC volume and perfusion) assessments provides an integrative approach linking episodic memory performance to brain substrates. This study also has limitations. While adequate for an intervention study, the sample size (45 participants in the NNLT and 44 in the no‐intervention group) may limit statistical power and the detection of subtle effects or robust cognitive–neurobiological correlations. Future longer longitudinal studies with larger sample sizes and perhaps a more intensive language program would be necessary to strengthen the findings and provide more reliable conclusions about the relationship between language learning and cognitive changes. It should be noted that bilingualism was not formally documented at baseline. Consequently, some bilingual participants may have been included in the trial, although likely in very small numbers. While bilinguals might differ from monolinguals, beneficial influence of learning a new language should still be expected. Future studies should systematically account for mono‐ versus bilingualism, and evaluate how it could modulate cognitive and neural outcomes associated with language learning. Although this study includes structural and cerebral perfusion measurements, it was not designed to examine network‐level functional changes such as functional connectivity or resting/task‐based functional MRI. Future studies specifically designed to assess functional connectivity including hippocampus–mPFC coupling, may provide complementary insight into the network‐level processes underlying cognitive changes associated with foreign language learning. Additionally, the cognitively unimpaired, highly educated, and generally active profile of our participants—with relatively healthy lifestyles—may have introduced healthy‐volunteer selection bias and blunted group differences as the no‐intervention group likely remained engaged in cognitively stimulating daily activities. Although changes in other cognitively stimulating activities cannot be entirely excluded, available trial data did not indicate differential pre–post lifestyle changes between groups. Therefore, these factors are unlikely to fully account for the present findings. Future studies are needed to assess the generalizability of these results to more diverse populations. Despite these limitations, these findings highlight NNLT as a promising, non‐invasive, accessible, and scalable strategy to mitigate age‐related episodic memory decline and enhance brain health in older adults. CONCLUSION Overall, these findings support the view that learning a language can benefit cognitive and brain health in cognitively unimpaired older adults. Engagement in complex cognitive activities in late‐life, such as language learning, may help maintain long‐term episodic memory and promote cognitive reserve through functional mechanisms supporting hippocampus‐dependent detail/recollection during memory retrieval. Thus, this study offers insights into the prevention of cognitive decline in older adults, proposing a targeted approach for maintaining long‐term brain health through non‐pharmacological interventions, warranting further exploration in larger, long‐term studies. AUTHOR CONTRIBUTIONS Concept and design: Julie Gonneaud and Gaël Chételat. Acquisition, analysis, or interpretation of data: Anne‐Laure Turpin, Caitlin Ware, Natalie L. Marchant, Brigitte Landeau, Florence Mézenge, Asrar Lehodey, Célia Soussi, Marion Delarue, Oriane Hébert, Vincent de la Sayette, Julie Gonneaud, and Gaël Chételat. Drafting of the manuscript: Anne‐Laure Turpin. Critical revision of the manuscript for important intellectual content: Caitlin Ware, Natalie L. Marchant, Brigitte Landeau, Florence Mézenge, Asrar Lehodey, Célia Soussi, Marion Delarue, Oriane Hébert, Vincent de la Sayette, Julie Gonneaud, and Gaël Chételat. Statistical analysis: Anne‐Laure Turpin. Obtained funding: Julie Gonneaud, Gaël Chételat. Supervision: Julie Gonneaud, Gaël Chételat. Julie Gonneaud has full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Anne‐Laure Turpin served as first author. Julie Gonneaud served as last author. Gaël Chételat, Brigitte Landeau, Florence Mézenge, Caitlin Ware, Natalie L. Marchant, Asrar Lehodey, Célia Soussi, Marion Delarue, Oriane Hébert, and Vincent de la Sayette served as co‐authors. CONFLICT OF INTEREST STATEMENT Gaël Chételat reports grants, personal fees, and non‐financial support from Institut National de la Santé et de la Recherche Médicale (Inserm), Fondation FondaMental in collaboration with Lilly laboratory (i.e., Lilly funding) and has received research support from Fondation Alzheimer, Fondation Recherche Alzheimer, Région Normandie, Association France Alzheimer et maladies apparentées, Fondation Vaincre Alzheimer, and Fondation Bettencourt Schueller outside the submitted work. Julie Gonneaud was supported by a Young Researcher Grant from the Fondation Alzheimer and Fondation de France. Institut National de la Santé et de la Recherche Médicale (INSERM) is the sponsor. The other authors report no relevant disclosures. Author disclosures are available in the Supporting Information. CONSENTS Concerning the Age‐Well RCT, the study was conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent prior to the examinations, and the Age‐Well randomized clinical trial was approved by the ethics committee (Comité de Protection des Personnes Nord‐Ouest III, Caen, France; trial registration number: EudraCT: 2016‐002441‐36; IDRCB: 2016‐A01767‐44; ClinicalTrials.gov Identifier: NCT02977819; registration date: 2016‐11‐25). Supporting information Supporting Information: dad270381‐sup‐0001‐SupMat.docx Supporting Information: dad270381‐sup‐0002‐ICMJE.pdf ACKNOWLEDGMENTS The authors thank all the Medit‐Ageing Research Group members and the Cyceron MRI‐PET staff members for their help with recruitment, data acquisition, or administrative support; the EUropean CLInical Trials Platform & Development (EUCLID) team, the sponsor (INSERM), and the participants of the study for their contribution. The full Medit‐Ageing Research Group is listed in the Appendix. The Age‐Well randomized clinical trial is part of the Medit‐Ageing project and is supported by the European Union's Horizon 2020 Research and Innovation Program (grant 667696), Région Normandie (Label d'Excellence), and Fondation d'Entreprise MMA des Entrepreneurs du Futur. APPENDIX: COLLABORATORS—THE MEDIT‐AGEING RESEARCH GROUP NameLocationRoleContributionAllais Florence, BAEUCLID/F‐CRIN Clinical Trials Platform, Bordeaux, FranceData managerData managementAndré Claire, PhDInstitut National de la Santé et de la Recherche Médicale, Caen, FrancePhD studentAcquisition, analysis, or interpretation of dataArenaza Urquijo Eider, PhDInstitut National de la Santé et de la Recherche Médicale, Caen, FrancePostdoctoral researcherStudy design; acquisition, analysis, or interpretation of dataBaez Lugo Sebastian, MScUniversity of Geneva, Geneva, SwitzerlandPhD studentAcquisition, analysis, or interpretation of dataBejanin Alexandre, PhDInstitut National de la Santé et de la Recherche Médicale, Caen, FrancePostdoctoral researcherAcquisition, analysis, or interpretation of dataBotton Maelle, MScInstitut National de la Santé et de la Recherche Médicale, Caen, FranceNeuropsychologistAcquisition, analysis, or interpretation of dataChampetier Pierre, PhDInstitut National de la Santé et de la Recherche Médicale, Caen, FrancePhD studentAcquisition, analysis, or interpretation of dataChauveau Léa, MScInstitut National de la Santé et de la Recherche Médicale, Caen, FrancePhD studentAcquisition, analysis, or interpretation of dataChételat Gaël, PhDInstitut National de la Santé et de la Recherche Médicale, Caen, FranceCoordinator, Work Package LeaderObtained funding, study designChocat Anne, MDInstitut National de la Santé et de la Recherche Médicale, Caen, FranceBiological doctorInvestigating doctorCollette Fabienne, PhDUniversity of Liege, Liege, BelgiumWork Package LeaderObtained funding, study designDautricourt Sophie, MD,PhDInstitut National de la Santé et de la Recherche Médicale, Caen, FrancePhD studentAcquisition, analysis, or interpretation of datade Flores Robin, PhDInstitut National de la Santé et de la Recherche Médicale, Caen, FrancePostdoctoral researcherAcquisition, analysis, or interpretation of data; administrative, technical, or material supportde la Sayette Vincent, MD, PhDCentre Hospitalier Universitaire de Caen, Caen, FranceNeurologistPrincipal Investigating doctorDelarue Marion, MScInstitut National de la Santé et de la Recherche Médicale, Caen, FranceNeuropsychologistAcquisition, analysis, or interpretation of dataDemnitz‐King Harriet, MScUniversity College London, United KingdomPhD studentAcquisition, analysis, or interpretation of dataEgret Stéphanie, MScInstitut National de la Santé et de la Recherche Médicale, Caen, FranceNeuropsychologistAcquisition, analysis, or interpretation of dataEl Sadawy Rawda, MScInstitut National de la Santé et de la Recherche Médicale, Caen, FranceNeuropsychologistAcquisition, analysis, or interpretation of dataEspérou Hélène, MDInstitut National de la Santé et de la Recherche Médicale, Paris, FranceGroup leaderSponsorFauvel Séverine, BAInstitut National de la Santé et de la Recherche Médicale, Caen, FranceTechnicianAcquisition, analysis, or Interpretation of dataFelisatti Francesca, MScInstitut National de la Santé et de la Recherche Médicale, Caen, FrancePhD studentAcquisition, analysis, or interpretation of dataFerment Victor, MScInstitut National de la Santé et de la Recherche Médicale, Caen, FranceTechnicianAcquisition, analysis, or interpretation of dataFerrand Devouge Eglantine, MD, MScInstitut National de la Santé et de la Recherche Médicale, Caen, FranceNeurologistInvestigating doctorFrison Eric, MD, PhDEUCLID/F‐CRIN Clinical Trials, Platform, Bordeaux, FranceMethodologistAcquisition, analysis, or interpretation of dataGonneaud Julie, PhDInstitut National de la Santé et de la Recherche Médicale, Caen, FranceWork Package LeaderObtained funding, study designHamel Anaïs, MScInstitut National de la Santé et de la Recherche Médicale, Caen, FrancePhD studentAcquisition, analysis, or interpretation of dataHaudry Sacha, MScInstitut National de la Santé et de la Recherche Médicale, Caen, FrancePhD studentAcquisition, analysis, or interpretation of dataHébert Oriane, MScInstitut National de la Santé et de la Recherche Médicale, Caen, FranceNeuropsychologistAcquisition, analysis, or interpretation of dataHeidmann Marc, MScInstitut National de la Santé et de la Recherche Médicale, Lyon, FrancePhD studentAcquisition, analysis, or interpretation of dataKuhn Elizabeth, PhDInstitut National de la Santé et de la Recherche Médicale, Caen, FrancePhD studentAcquisition, analysis, or interpretation of dataKlimecki Olga, PhDUniversity of Geneva, Geneva, SwitzerlandWork Package LeaderObtained funding; study designLandeau Brigitte, MScInstitut National de la Santé et de la Recherche Médicale, Caen, FranceNeuroimaging developer (engineer)Acquisition, analysis, or interpretation of dataLedu Gwendoline, MScInstitut National de la Santé et de la Recherche Médicale, Caen, FranceClinical research assistantAcquisition, analysis, or interpretation of data; administrative, technical, or material supportLefranc Valérie, BAInstitut National de la Santé et de la Recherche Médicale, Caen, FranceClinical research assistantAcquisition, analysis, or interpretation of data; Administrative, technical, or material supportLutz Antoine, PhDInstitut National de la Santé et de la Recherche Médicale, Lyon, FranceWork Package LeaderObtained funding; study designMarchant Natalie, PhDUniversity College London, United KingdomWork Package LeaderObtained funding; study designMézenge Florence, BAInstitut National de la Santé et de la Recherche Médicale, Caen, FranceNeuroimaging engineer assistantAcquisition, analysis, or interpretation of dataMoulinet Inès, PhDInstitut National de la Santé et de la Recherche Médicale, Caen, FrancePhD studentAcquisition, analysis or interpretation of dataOurry Valentin, PhDInstitut National de la Santé et de la Recherche Médicale, Caen, FrancePhD studentAcquisition, analysis, or interpretation of dataPalix Cassandre, MScInstitut National de la Santé et de la Recherche Médicale, Caen, FrancePhD studentAcquisition, analysis, or interpretation of dataPaly Léo, MScInstitut National de la Santé et de la Recherche Médicale, Caen, FranceNeuropsychologistAcquisition, analysis, or interpretation of dataPoisnel Géraldine, PhDInstitut National de la Santé et de la Recherche Médicale, Caen, FranceWork Package LeaderObtained funding; study designQuillard Anne, MDInstitut National de la Santé et de la Recherche Médicale, Caen, FranceGeriatric doctorInvestigating doctorRauchs Géraldine, PhDInstitut National de la Santé et de la Recherche Médicale, Caen, FranceResearcherStudy designRehel Stéphane, PhDInstitut National de la Santé et de la Recherche Médicale, Caen, FrancePhD studentAcquisition, analysis, or interpretation of dataRequier Florence, MScUniversity of Liege, Liege, BelgiumPhD studentAcquisition, analysis, or interpretation of dataSalmon Eric, MD,PhDUniversity of Liege, Liege, BelgiumGroup leaderObtained funding; study designSmith Rhonda, MScMinerva Health & Care Communications Ltd, Andover, United KingdomGroup leaderCommunication, disseminationTomadesso Clémence, PhDInstitut National de la Santé et de la Recherche Médicale, Caen, FrancePhD studentAcquisition, analysis, or interpretation of dataTouron Edelweiss, PhDInstitut National de la Santé et de la Recherche Médicale, Caen, FrancePhD studentAcquisition, analysis, or interpretation of dataTurpin Anne‐Laure, MScInstitut National de la Santé et de la Recherche Médicale, Caen, FrancePhD studentAcquisition, analysis, or interpretation of dataVuilleumier Patrik, MDUniversity of Geneva, Geneva, SwitzerlandGroup leaderObtained funding; study designWhitfield Tim, PhDUnited Kingdom University College London,PhD studentAcquisition, analysis, or interpretation of dataWirth Miranka, PhDDeutsches Zentrum für Neurodegenerative Erkrankungen, Dresden, GermanyGroup leaderStudy design Turpin A‐L, Ware C, Landeau B, et al. Impact of an 18‐month non‐native language intervention on episodic memory and the underlying neural mechanisms in cognitively unimpaired older adults. Alzheimer's Dement. 2026;18:e70381. 10.1002/dad2.70381 The full list of the Medit‐Ageing Research Group is provided at the end of the article in the Appendix: Collaborators. DATA AVAILABILITY STATEMENT For the Age‐Well RCT, the data underlying this report are made available on request following a formal data sharing agreement and approval by the consortium and executive committee (https://silversantestudy.eu/2020/09/25/datasharing). The material can be mobilized, under the conditions and modalities defined in the Medit‐Ageing Charter, by any research team belonging to an academic institution for carrying out a scientific research project relating to the scientific theme of mental health and well‐being in older people. The material may also be mobilized by non‐academic third parties, under conditions, in particular financial, which will be established by separate agreement between INSERM and by the said third party. Data sharing policies described in the Medit‐Ageing Charter are following our ethics approval and guidelines from our funding body. REFERENCES Associated Data Supplementary Materials Supporting Information: dad270381‐sup‐0001‐SupMat.docx Supporting Information: dad270381‐sup‐0002‐ICMJE.pdf Data Availability Statement For the Age‐Well RCT, the data underlying this report are made available on request following a formal data sharing agreement and approval by the consortium and executive committee (https://silversantestudy.eu/2020/09/25/datasharing). The material can be mobilized, under the conditions and modalities defined in the Medit‐Ageing Charter, by any research team belonging to an academic institution for carrying out a scientific research project relating to the scientific theme of mental health and well‐being in older people. The material may also be mobilized by non‐academic third parties, under conditions, in particular financial, which will be established by separate agreement between INSERM and by the said third party. Data sharing policies described in the Medit‐Ageing Charter are following our ethics approval and guidelines from our funding body.