Self-Perceptions of Aging in Older Adults: A Network Analysis of Clinical and Non-Clinical Samples

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript presents an original application of network analysis to examine psychosocial representations of aging and memory in older adults. The comparison between memory clinic patients and non-clinical controls with subjective complaints is particularly novel. Overall, the manuscript is strong and offers a valuable contribution to the literature. The suggestions below are intended to help clarify several conceptual and methodological points.

Age and education differ significantly between groups. While age was appropriately controlled statistically, education was not included as a covariate. Providing a brief justification for this decision or reporting whether inclusion of education alters the key findings would strengthen group comparisons.

The Discussion is generally careful and well grounded; however, some phrasing implies causal or regulatory roles (e.g., buffering or influencing effects). Given the cross-sectional design and correlational nature of network analysis, it may be helpful to frame these interpretations more explicitly as associative. Adding a brief paragraph in the Discussion reminding readers that network centrality does not imply causal priority, and that observed structures may reflect contextual salience rather than functional mechanisms, would enhance conceptual clarity.

Using partial correlation networks rather than regularized EBIC-LASSO networks is reasonable given sample size considerations. That said, this analytic decision has implications for interpretability. Clarifying the rationale for considering N ≈ 130 and N ≈ 84 insufficient for EBIC-LASSO but acceptable for partial correlations, and briefly discussing potential implications for network density and stability, would be helpful for readers less familiar with these methods.

Several subscales were excluded due to low reliability (e.g., Personal control [Helplessness] and Treatment control), while others with more modest reliability (e.g., Consequences and Illness coherence in the patient group) were retained. Although these decisions are well justified psychometrically, they may influence construct coverage, particularly for control-related dimensions. A short discussion of how these exclusions might have shaped network structure and centrality patterns, especially the relative prominence of emotional and threat-related nodes, would help contextualize the findings.

The finding that patients show a largely unified network whereas controls show two distinct communities is interesting but alternative explanations (e.g., heightened covariance in clinical contexts inflating network density and/or measurement context effects (hospital vs laboratory) increasing shared variance) could be acknowledged.

 

With these clarifications, this manuscript is ready for publication. Its strong theoretical grounding, thoughtful use of network methodology, and clear clinical relevance contribute meaningfully to the literature on cognitive aging and psychosocial processes.

 

Author Response

This manuscript presents an original application of network analysis to examine psychosocial representations of aging and memory in older adults. The comparison between memory clinic patients and non-clinical controls with subjective complaints is particularly novel. Overall, the manuscript is strong and offers a valuable contribution to the literature. The suggestions below are intended to help clarify several conceptual and methodological points.

We thank the reviewer for this positive and constructive feedback of our manuscript. 

Comment 1 : Age and education differ significantly between groups. While age was appropriately controlled statistically, education was not included as a covariate. Providing a brief justification for this decision or reporting whether inclusion of education alters the key findings would strengthen group comparisons.

Response 1 : We agree that the treatment of educational level needed to be clarified, as it is not inherently constitutive of the clinical vs. non-clinical contrast in the same way as cognitive status or memory complaints. To address this point directly, we re-estimated the network analyses including education level as an additional covariate alongside age. These analyses yielded results that were qualitatively identical to those obtained when controlling for age only. Consequently, to maintain a clear focus on aging-related processes, we retained the more parsimonious model including age as the sole covariate in the main text. The results of education-controlled network analyses are reported in the supplementary materials. This analytical choice is now made explicit in the Materials and Methods section (2.4 Statistical analysis, page 10, lines 347-355), the Results section (page 17, lines 602-609), and its implications are discussed in the Discussion section (page 23, lines 880-896).  

Comment 2 : The Discussion is generally careful and well grounded; however, some phrasing implies causal or regulatory roles (e.g., buffering or influencing effects). Given the cross-sectional design and correlational nature of network analysis, it may be helpful to frame these interpretations more explicitly as associative. Adding a brief paragraph in the Discussion reminding readers that network centrality does not imply causal priority, and that observed structures may reflect contextual salience rather than functional mechanisms, would enhance conceptual clarity.

Response 2 : Thank you for this important reminder. In line with your suggestion, we added a new paragraph explicitly reminding readers that the network analyses are cross-sectional and correlational. In particular, we clarify that centrality indices should not be considered as proof of causality, as they offer insight into self-representations of aging of clinical vs.  non-clinical samples at a specific time and context. We also added that the networks provide knowledge on how the psychosocial variables relate to each other but do not explain functional mechanisms. This is now discussed in the Discussion section (4.6 Limitations, page 23, lines 862-871).

Comment 3 : Using partial correlation networks rather than regularized EBIC-LASSO networks is reasonable given sample size considerations. That said, this analytic decision has implications for interpretability. Clarifying the rationale for considering N ≈ 130 and N ≈ 84 insufficient for EBIC-LASSO but acceptable for partial correlations, and briefly discussing potential implications for network density and stability, would be helpful for readers less familiar with these methods.

Response 3: We agree that this analytic choice has important implications for interpretability and required further clarification. EBIC-LASSO networks were initially estimated. However, due to small samples in the patient (N=129) and the control (N=84) groups, the EBIC-LASSO penalization procedure produced networks that were excessively sparse, with most edges set to zero, leading to a crucial loss of information for this exploratory study. In contrast, partial correlation networks allowed an interpretable descriptive characterization of the overall organization of associations among psychosocial variables, which was more consistent with our research question.  Accordingly, this methodological rationale is now explicitly clarified in Materials and Methods section (2.4 Statistical analysis, page 10, lines 372-376). In addition, we acknowledge in the Discussion that larger sample sizes would be required to support more robust network estimation, such as EBIC-LASSO (4.6 Limitations, page 22, line 838-839). 

Comment 4 : Several subscales were excluded due to low reliability (e.g., Personal control [Helplessness] and Treatment control), while others with more modest reliability (e.g., Consequences and Illness coherence in the patient group) were retained. Although these decisions are well justified psychometrically, they may influence construct coverage, particularly for control-related dimensions. A short discussion of how these exclusions might have shaped network structure and centrality patterns, especially the relative prominence of emotional and threat-related nodes, would help contextualize the findings.

Response 4: We agree that this point required clarification. While we excluded subscales with clearly insufficient reliability (e.g., Personal control and Treatment control), we retained subscales showing more modest internal consistency in the patient group (notably Consequences and Illness coherence) to preserve construct coverage of two dimensions that are theoretically central to our research question: perceived impact of memory difficulties (Consequences) and perceived understanding/insight regarding these difficulties (Illness coherence). Importantly, both subscales showed acceptable reliability in the control group and removing them would have disproportionately reduced coverage of illness representations in the patient network. We now explicitly acknowledge that these measurement decisions may have shaped the resulting network structure and centrality patterns. This trade-off and its implications are now discussed in the revised manuscript in the Discussion section (4.6 Limitations; page 22, lines 841–854), with the corresponding clarification in Materials and Methods (2.5 Data preparation; page 11, lines 419-420, 422-426).

Comment 5 : The finding that patients show a largely unified network whereas controls show two distinct communities is interesting but alternative explanations (e.g., heightened covariance in clinical contexts inflating network density and/or measurement context effects (hospital vs laboratory) increasing shared variance) could be acknowledged.

Response 5: Thank you for highlighting this point. We acknowledge that the unified network observed in our patients’ sample may also reflect contextual effects related to the clinical setting. Heightened stress or perceived diagnostic threat in a memory clinical context may increase covariance between variables, thereby artificially inflating network density and reducing the emergence of distinct communities. Additionally, differences in assessment context (hospital vs. laboratory) may influence response patterns, with the clinical setting potentially promoting a more emotionally charged or threat-focused perspective. These context-related effects are now explicitly acknowledged as alternative explanations that may have contributed to the more unified network structure observed in patients compared to controls. It is now discussed in the Discussion section (4.4 Community structure of representational networks; page 21, lines 796–801).

With these clarifications, this manuscript is ready for publication. Its strong theoretical grounding, thoughtful use of network methodology, and clear clinical relevance contribute meaningfully to the literature on cognitive aging and psychosocial processes

Thank you again for your reviews and kind comments about the paper.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

Overall Comment

The manuscript addresses an important and underexplored topic by examining the organization of psychosocial representations related to aging and memory in clinical and non-clinical contexts. The conceptual focus on the structure of self-perceptions—rather than isolated mean-level differences—is both original and theoretically meaningful. The application of network analysis to this question is creative, timely, and well aligned with the study’s objectives. Overall, the study is thoughtfully designed and highly encouraging, and it makes a valuable contribution to the literature on cognitive aging. That said, a few issues remain that, if addressed, would further strengthen the clarity, rigor, and interpretability of the manuscript. These concerns are primarily related to framing and interpretation rather than to the quality of the data or analyses.

Major Issue 1: Group differences are not sufficiently framed as a core feature of the study design

The clinical and control groups differ in several respects, including age, educational level, and, most importantly, recruitment context (memory clinic versus laboratory-based research setting). Although age was statistically controlled, the manuscript does not sufficiently emphasize that these contextual differences are an inherent part of the study design rather than a methodological limitation. As a result, readers may question whether the observed differences reflect sampling bias rather than meaningful contextual effects.

The authors would benefit from explicitly clarifying that the purpose of the comparison is not to determine which group performs better or worse, but rather to examine how psychosocial representations are structured differently across clinical and non-clinical contexts.

Major Issue 2: Limited discussion of the practical and clinical significance of the main findings

The manuscript identifies distinct central variables within the two networks, with subjective age emerging as the most central component in the clinical group and aging stereotypes occupying a central role in the non-clinical group. While these findings are clearly reported and well supported by the analyses, the Discussion remains largely descriptive and does not sufficiently elaborate on why these results are important from a practical or clinical perspective. As a result, readers may have difficulty understanding how the identified network structures could inform clinical interpretation, communication with patients, or the development of psychosocial interventions, particularly in the context of a first memory clinic assessment.

The authors would benefit from more explicitly addressing the potential implications of these findings in the Discussion. For instance, clarifying why subjective age may become particularly salient in a clinical diagnostic context, and how this variable could represent a meaningful psychological leverage point for psychoeducation, clinician–patient communication, or early psychosocial support, would substantially strengthen the applied relevance of the study. These interpretations can be presented cautiously, without implying causal relationships, while still highlighting the broader significance of the results.

Author Response

Overall Comment

The manuscript addresses an important and underexplored topic by examining the organization of psychosocial representations related to aging and memory in clinical and non-clinical contexts. The conceptual focus on the structure of self-perceptions—rather than isolated mean-level differences—is both original and theoretically meaningful. The application of network analysis to this question is creative, timely, and well aligned with the study’s objectives. Overall, the study is thoughtfully designed and highly encouraging, and it makes a valuable contribution to the literature on cognitive aging. That said, a few issues remain that, if addressed, would further strengthen the clarity, rigor, and interpretability of the manuscript. These concerns are primarily related to framing and interpretation rather than to the quality of the data or analyses.

We thank the reviewer for this positive feedback of our manuscript. 

Comment 1 : Major Issue 1: Group differences are not sufficiently framed as a core feature of the study design

The clinical and control groups differ in several respects, including age, educational level, and, most importantly, recruitment context (memory clinic versus laboratory-based research setting). Although age was statistically controlled, the manuscript does not sufficiently emphasize that these contextual differences are an inherent part of the study design rather than a methodological limitation. As a result, readers may question whether the observed differences reflect sampling bias rather than meaningful contextual effects.

The authors would benefit from explicitly clarifying that the purpose of the comparison is not to determine which group performs better or worse, but rather to examine how psychosocial representations are structured differently across clinical and non-clinical contexts.

Response 1 : We agree that the recruitment context could be more emphasized, as it is in fact a fundamental component of our study rather than a bias.  Following your suggestion, changes were made in the Discussion (pages 17-18, lines 615-619). In addition, we agree that the treatment of educational level needed to be clarified, as it is not inherently constitutive of the clinical vs. non-clinical contrast in the same way as cognitive status or memory complaints. Regarding education level, network analysis was estimated with education as a covariate, alongside age, and yielded similar results. To maintain focus on our study, age was kept as the sole covariate. Results of network analysis with education years can be found in our supplementary material. This analytical choice is now made explicit in the Materials and Methods section (2.4 Statistical analysis, page 10, lines 347-355), the Results section (page 17, lines 602-609), and its implications are discussed in the Discussion section (page 23, lines 880-896).   

Comment 2 : Major Issue 2: Limited discussion of the practical and clinical significance of the main findings

The manuscript identifies distinct central variables within the two networks, with subjective age emerging as the most central component in the clinical group and aging stereotypes occupying a central role in the non-clinical group. While these findings are clearly reported and well supported by the analyses, the Discussion remains largely descriptive and does not sufficiently elaborate on why these results are important from a practical or clinical perspective. As a result, readers may have difficulty understanding how the identified network structures could inform clinical interpretation, communication with patients, or the development of psychosocial interventions, particularly in the context of a first memory clinic assessment.

The authors would benefit from more explicitly addressing the potential implications of these findings in the Discussion. For instance, clarifying why subjective age may become particularly salient in a clinical diagnostic context, and how this variable could represent a meaningful psychological leverage point for psychoeducation, clinician–patient communication, or early psychosocial support, would substantially strengthen the applied relevance of the study. These interpretations can be presented cautiously, without implying causal relationships, while still highlighting the broader significance of the results.

Response 2 : Thank you for this important suggestion. We have added a paragraph addressing these points in a new paragraph in the Discussion section (4.5 Clinical impact of subjective age and illness coherence in patients, pages 21-22, lines 803-831), and revised the end of our abstract to mention the potential clinical relevance of our findings (page 2, Lines 35-39).  

 

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript is well designed, clearly written, and addresses a relevant and original research question. The following comments are intended to improve interpretative clarity and transparency.

 

Comments:

1. The patient and control groups differ not only in clinical context but also in age, education level, cognitive performance, and severity of cognitive complaints. Although age was statistically controlled, other group differences may still contribute to the observed network structures. This potential confounding should be explicitly acknowledged in the Discussion or Limitations section.

 

2. The sample is drawn from a specific cultural and healthcare context and includes older adults with preserved functional autonomy and mild cognitive complaints. These factors may limit extension of the findings to other populations.

Author Response

This manuscript is well designed, clearly written, and addresses a relevant and original research question. The following comments are intended to improve interpretative clarity and transparency.

We thank the reviewer for this positive feedback of our manuscript. 

Comment 1 :The patient and control groups differ not only in clinical context but also in age, education level, cognitive performance, and severity of cognitive complaints. Although age was statistically controlled, other group differences may still contribute to the observed network structures. This potential confounding should be explicitly acknowledged in the Discussion or Limitations section.

Response 1 : Thank you for this suggestion. We agree that, beyond age, other group differences should be more detailed in the article. In the revised manuscript, we clarified that not only do our groups differ in the assessment context but also on education years, cognitive resources (MMSE) and cognitive complaints (QCC).  

MMSE and QCC were not included as covariates because they reflect core clinical features that define the contrast between individuals referred for a first diagnostic evaluation in a memory clinic and non-clinical community-dwelling older adults. Partialing out these variables would have altered the very clinical distinction the study aimed to characterize. Education level, which is not inherently constitutive of the clinical versus non-clinical contrast, was examined in supplementary network analyses as an additional covariate alongside age. These analyses yielded results that were qualitatively identical to those obtained when controlling for age only. To maintain a clear focus on aging-related processes, age was therefore retained as the sole covariate in the main analyses, while education-adjusted results are reported in the Supplementary Materials. This rationale is now made explicit in the Materials and Methods section (2.4 Statistical analysis, page 10, lines 347-355), the Results section (page 17, lines 602-609), and its implications are discussed in the Discussion section (page 23, lines 880-896).   

Comment 2 : The sample is drawn from a specific cultural and healthcare context and includes older adults with preserved functional autonomy and mild cognitive complaints. These factors may limit extension of the findings to other populations.

Response 2 : Thank you for this insightful suggestion. We agree with this statement and have added a paragraph addressing those points in our manuscript. These changes were made in the Discussion section (4.6 Limitations, page 23, lines 897-903). 

Author Response File: Author Response.docx

Reviewer 4 Report

Comments and Suggestions for Authors

This manuscript addresses an important and underexplored aspect of cognitive aging by examining the structural organization of psychosocial self-perceptions in clinical and non-clinical contexts using network analysis. The study is theoretically well grounded, methodologically sound, and offers a novel contribution to the literature by highlighting how subjective age and aging stereotypes occupy different central roles depending on context.

To further strengthen the manuscript, the authors may consider:

Providing a more explicit justification for the choice of non-regularized partial correlation networks over EBIC-lasso networks, particularly in relation to sample size considerations.

Clarifying the conceptual implications of excluding certain subscales due to low internal consistency and how this may affect the interpretation of the network structures.

Expanding slightly on the clinical and practical implications of the findings, particularly how insights into subjective age and illness coherence could inform interventions in memory clinic settings.

Overall, this is a high-quality and original manuscript that is suitable for publication after minor methodological clarifications.

Author Response

This manuscript addresses an important and underexplored aspect of cognitive aging by examining the structural organization of psychosocial self-perceptions in clinical and non-clinical contexts using network analysis. The study is theoretically well grounded, methodologically sound, and offers a novel contribution to the literature by highlighting how subjective age and aging stereotypes occupy different central roles depending on context.

We thank the reviewer for this positive feedback of our manuscript. 

To further strengthen the manuscript, the authors may consider:

Comment 1 : Providing a more explicit justification for the choice of non-regularized partial correlation networks over EBIC-lasso networks, particularly in relation to sample size considerations.

Response 1 : Thank you for your comment. We agree that a more explicit justification was needed. EBIC-LASSO networks were initially estimated. However, due to small samples in the patient (N=129) and the control (N=84) groups, the EBIC-LASSO penalization procedure produced networks that were excessively sparse, with most edges set to zero, leading to a crucial loss of information for this exploratory study. In contrast, partial correlation networks allowed a more informative and interpretable descriptive characterization of the overall organization of associations among psychosocial variables at the current sample sizes. This approach was therefore better suited to our aim of descriptively comparing network structures across clinical and non-clinical contexts, rather than identifying a minimal set of highly robust edges.  Accordingly, this methodological rationale is now explicitly clarified in Materials and  Methods section (2.4 Statistical analysis, page 10, lines 372-376). In addition, we acknowledge in the Discussion that larger sample sizes would be required to support more robust network estimation, such as EBIC-LASSO (4.6 Limitations, page 22, line 838-839). 

Comment 2 : Clarifying the conceptual implications of excluding certain subscales due to low internal consistency and how this may affect the interpretation of the network structures.

Response 2 : We agree that this point required clarification. While we excluded subscales with clearly insufficient reliability (e.g., Personal control and Treatment control), we retained subscales showing more modest internal consistency in the patient group (notably Consequences and Illness coherence) to preserve construct coverage of two dimensions that are theoretically central to our research question: perceived impact of memory difficulties (Consequences) and perceived understanding/insight regarding these difficulties (Illness coherence). Importantly, both subscales showed acceptable reliability in the control group and removing them would have disproportionately reduced coverage of illness representations in the patient network. 
We now explicitly acknowledge that these measurement decisions may have shaped the resulting network structure and centrality patterns. This trade-off and its implications are now discussed in the revised manuscript in the Discussion section (4.6 Limitations; page 22, lines 841–854), with the corresponding clarification in Materials and Methods (2.5 Data preparation; page 11, lines 419-420, 422–426).

Comment 3 : Expanding slightly on the clinical and practical implications of the findings, particularly how insights into subjective age and illness coherence could inform interventions in memory clinic settings.

Response 3 :  Thank you for raising this important point. We have added a paragraph addressing these points in the Discussion section (4.5 Clinical impact of subjective age and illness coherence in patients, pages 21-22, lines 803-831) and revised the end of our abstract to mention the potential clinical relevance of our findings (page 2, Lines 35-37). 

Overall, this is a high-quality and original manuscript that is suitable for publication after minor methodological clarifications.
Thank you again for your reviews and kind comments

Author Response File: Author Response.docx