Attachment and Relational Dynamics in Residential Care Contexts: Perceptions of Reference Adults and Implications for Resilience

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Thank you for the opportunity to review this work. I can tell this had a good amount of work put into it. I am just going to jump straight into it. 

The introduction provides a useful foundation for understanding the study and its research questions. However, it would benefit from a more comprehensive discussion of resilience within these settings, including additional historical and theoretical context. In addition, a dedicated contemporary literature review section appears to be missing. There is likely a broader body of recent research addressing these issues that is not reflected in the current reference list. As written, the manuscript transitions rather abruptly from the introduction to the methods section.

In the participants section, I recommend including a descriptive table summarizing the sample characteristics. Presenting the sociodemographic variables and their frequencies would make it easier for readers to understand the composition of both the overall and analytic samples. Similarly, I would like to see a breakdown of the sociodemographic data collected through the questionnaire.

I would also appreciate greater clarification regarding the "relevant life events" measure. Specifically, what criteria were used to determine whether an event was considered relevant? Are we discussing ACES or developmentally expected life events? 

The description of the Observation Checklist also raises several questions. The manuscript states that the checklist supports systematic observation of adolescents aged 11–16 years, whereas the study sample includes participants aged 12–18 years. This discrepancy should be addressed or justified. Additionally, although the instrument is described as non-diagnostic and intended for reflective monitoring, I wonder whether adults working in these facilities would have sufficient familiarity with adolescents after only two months to provide reliable assessments. Is a two-month period considered adequate for young people in these settings to become comfortable enough that their behaviors and relationships can be accurately observed? Finally, it would be helpful to describe the observer training procedures in greater detail, including whether inter-rater reliability was assessed among observers.

I believe the presentation of the results would benefit substantially from additional descriptive tables. Since many of the analyses appear to focus on frequency distributions, summarizing these findings in tables would improve readability and reduce the amount of text required. I would also recommend consolidating the numerous small subheadings into broader subsections (e.g., under Section 3.1), accompanied by summary tables where appropriate. Relatedly, I question the interpretation of the presence of a reference figure after only a two-month period as evidence of stable relational support. This conclusion may require further justification or a more cautious interpretation.

There also appears to be some inconsistency in the reporting of descriptive statistics. For example, Table 1 reports values differently (e.g., n versus percentages). I encourage the authors to adopt a consistent reporting format throughout the manuscript. Regarding the regression analyses, was sex the only control variable included? Based on the methods section, it seemed that additional sociodemographic variables were collected. If so, it would be helpful to clarify why they were not included in the analyses, or to explain the rationale for the selected covariates. For the qualitative findings, I recommend including illustrative participant quotes or examples to support the broader themes and interpretations presented. Providing representative excerpts would strengthen the credibility of the qualitative analysis. Finally, I found that portions of the Results section include discussion of implications that would be more appropriately reserved for the Discussion. Moving these interpretive comments would help distinguish the results from their interpretation and reduce some of the repetition across sections.

The Discussion and Conclusion could be streamlined, as there is some overlap and repetition between the two sections. Similar to my comments on the introduction, I would also expect a stronger integration of contemporary literature to contextualize the findings and inform recommendations for future research.

Overall, this is a well-executed manuscript that addresses an important topic. The study has a number of strengths, and the analyses appear to be thoughtfully conducted. With revisions to improve clarity, organization, consistency, and engagement with the contemporary psychological literature, I believe the manuscript would be substantially strengthened and better positioned for publication.

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted file.

Reviewer 1 comment:

Thank you for the opportunity to review this work. I can tell this had a good amount of work put into it. I am just going to jump straight into it. “The introduction provides a useful foundation for understanding the study and its research questions. However, it would benefit from a more comprehensive discussion of resilience within these settings, including additional historical and theoretical context. In addition, a dedicated contemporary literature review section appears to be missing.“

Response:
We introduced a paragraph dedicated to this specific issue, resilience in residential care, addressing what we considered as major points: the concept and indicators of resilience, factors that promote resilience from the perspective of professionals and children in care, and also the processes through which good results are accomplished.

Reviewer 1 comment:

“There is likely a broader body of recent research addressing these issues that is not reflected in the current reference list.”

Response:
The introduced paragraph is based on recent research; new references were considered and included in the references list.

Reviewer 1 comment:

“As written, the manuscript transitions rather abruptly from the introduction to the methods section.”

Response:
We introduced one more sentence citing an earlier study that points to the importance of addressing the specific issue of the use of attachment theory in residential care. We consider that this can clarify the specific gap we look to fill with this work.

Reviewer 1 comment:

“In the participants section, I recommend including a descriptive table summarizing the sample characteristics. Presenting the sociodemographic variables and their frequencies would make it easier for readers to understand the composition of both the overall and analytic samples. Similarly, I would like to see a breakdown of the sociodemographic data collected through the questionnaire.”

Response:
We have expanded the presentation of the sociodemographic characteristics of the adolescent sample by providing a descriptive breakdown of the variables collected through the sociodemographic questionnaire. A new table has been included reporting the available frequencies and percentages for categorical variables and descriptive statistics for continuous variables. This addition provides a clearer overview of the participants’ demographic, educational, and residential care characteristics and facilitates interpretation of the study sample.

Reviewer 1 comment:

I would also appreciate greater clarification regarding the "relevant life events" measure. Specifically, what criteria were used to determine whether an event was considered relevant? Are we discussing ACES or developmentally expected life events?

Response:
We are discussing ACES, and for this reason we substituted “relevant life events” by “adverse childhood events”.

Reviewer 1 comment:

The description of the Observation Checklist also raises several questions. The manuscript states that the checklist supports systematic observation of adolescents aged 11–16 years, whereas the study sample includes participants aged 12–18 years. This discrepancy should be addressed or justified.

Response:
The discrepancy is now justified in lines 320-324.

Reviewer 1 comment:

“Additionally, although the instrument is described as non-diagnostic and intended for reflective monitoring, I wonder whether adults working in these facilities would have sufficient familiarity with adolescents after only two months to provide reliable assessments. Is a two-month period considered adequate for young people in these settings to become comfortable enough that their behaviors and relationships can be accurately observed? Finally, it would be helpful to describe the observer training procedures in greater detail, including whether inter-rater reliability was assessed among observers.”

Response:
The workers received previous training concerning attachment issues in residential care and the use of the checklist. The psychologists of each residential care facility were also present in the training and were available (as the research team members) to answer any doubt that may arise. Two months was considered enough time to report a behavioural pattern in the several areas because the staff members contact daily with the adolescents, and are aware of their specific features and concerns. During the training staff was guided to talk with staff colleagues  in order to fill the checklist considering the usual behavioural pattern and not the more salient one. It was not assessed the inter-rater reliability but was promoted a collaborative posture between the team members in order to participate in the study. This explanation was included in the text (lines 372-374).

Reviewer 1 comment:

“I believe the presentation of the results would benefit substantially from additional descriptive tables. Since many of the analyses appear to focus on frequency distributions, summarizing these findings in tables would improve readability and reduce the amount of text required. I would also recommend consolidating the numerous small subheadings into broader subsections (e.g., under Section 3.1), accompanied by summary tables where appropriate.”

Response:
We have revised the Results section to improve clarity and readability. Specifically, frequency-based descriptive findings that were previously presented primarily in the text have been consolidated into summary tables, with the accompanying narrative shortened to emphasize only the main patterns observed in the data. We have also reorganized the Results section by removing several small subheadings and grouping related findings within broader subsections. In particular, the findings concerning previous residential placements, reference figures, perceived support and trust in professionals, peer relationships, family contact, and sources of support are now presented together under Section 3.1, “Relational Environment in Residential Care,” and summarized in a descriptive table. These changes reduce repetition in the text, provide a clearer overview of the frequency distributions, and facilitate comparison across the main relational and sociodemographic characteristics of the sample.

Reviewer 1 comment:

“Relatedly, I question the interpretation of the presence of a reference figure after only a two-month period as evidence of stable relational support. This conclusion may require further justification or a more cautious interpretation.”

Response:
Two months was the amount of time of observation of adolescent’s behaviour required to apply the checklist, in order to ensure that the observations correspond to a pattern and not to single events. Besides that, and as can be observed from the sociodemographic data, the length of stay in resident ranges from 1 month to 8 years, with a mean value of 24 months; that means that the great majority of the adolescents has a long relation with the reference figure.

Reviewer 1 comment:

“There also appears to be some inconsistency in the reporting of descriptive statistics. For example, Table 1 reports values differently (e.g., n versus percentages). I encourage the authors to adopt a consistent reporting format throughout the manuscript. Regarding the regression analyses, was sex the only control variable included? Based on the methods section, it seemed that additional sociodemographic variables were collected. If so, it would be helpful to clarify why they were not included in the analyses, or to explain the rationale for the selected covariates.”

Response:
We have revised the descriptive reporting throughout the manuscript to ensure consistency. Categorical variables are now systematically presented as n (%), whereas continuous variables are reported as M (SD) and, where informative, Min–Max. Table 1 and the additional descriptive tables have been revised accordingly, and discrepancies between frequencies and percentages in the attachment classifications have also been corrected. Regarding the regression analysis, we confirm that sex was the only sociodemographic covariate included in the reported model. We agree that the rationale for this decision was not sufficiently clear in the original manuscript. The regression was conducted using the Enter method in SPSS; therefore, all predictors specified for the model were entered simultaneously rather than selected automatically by the statistical procedure. Given the limited sample size (N = 33) and the exploratory nature of the study, we adopted a parsimonious modelling strategy to avoid overparameterization. The attachment-related predictors and sex were entered simultaneously, with sex included as a sociodemographic covariate to account for potential sex-related differences in psychological vulnerability. Although additional sociodemographic information was collected, these variables were not simultaneously included because doing so would have substantially increased model complexity relative to the available sample size. We have revised the Statistical Analysis subsection to make the Enter procedure and the rationale for covariate inclusion explicit. We have also clarified that the regression represents an exploratory, parsimonious model rather than a fully adjusted model incorporating all available sociodemographic characteristics.

Reviewer 1 comment:

“For the qualitative findings, I recommend including illustrative participant quotes or examples to support the broader themes and interpretations presented. Providing representative excerpts would strengthen the credibility of the qualitative analysis.”

Response:
We added illustrative quotes for each category described in section 3.6.

Reviewer 1 comment:

“Finally, I found that portions of the Results section include discussion of implications that would be more appropriately reserved for the Discussion. Moving these interpretive comments would help distinguish the results from their interpretation and reduce some of the repetition across sections.

The Discussion and Conclusion could be streamlined, as there is some overlap and repetition between the two sections. Similar to my comments on the introduction, I would also expect a stronger integration of contemporary literature to contextualize the findings and inform recommendations for future research.

Response:

The Discussion and Conclusion sections were reviewed in order to decrease the verified overlap between them.

Reviewer 2 Report

Comments and Suggestions for Authors

1. The introduction covers the following topics (lines 51-138): residential care, attachment theory, developmental trauma, therapeutic residential care, the role of professionals, and the research gap. A stronger introduction should answer questions such as What is already known? What remains uncertain? Why are previous findings insufficient? Why is this study needed now?
2. "Empirical research examining how residential care professionals perceive attachment-related behaviors remains limited..." (lines 122-128). Authors do not demonstrate this claim through the literature. Therefore, the gap appears asserted rather than established. The authors need to discuss which studies examined caregiver perceptions. What did those studies find? What methodological limitations existed? What questions remain unanswered?
3. Although several recent references are cited, the citation functions mainly as supportive evidence and is rarely discussed critically. For instance, Martínez-Usarralde et al. (2025) is cited as evidence that caregiver relationships matter (lines 98-101). However, the review could explain: What variables were examined? What evidence supports attachment-informed care? What limitations did that review identify? How does the present study extend those findings?
4. Objective 3 uses the verb "explore" to describe the examination of associations. Given that inferential statistical analyses (Spearman correlations and multiple linear regression) were conducted, a more precise verb such as "examine" or "investigate" would better reflect the quantitative nature of the analyses and improve consistency with the study design.
5. The author outlines its intended contribution by stating that it combines standardized attachment measures with professionals' perspectives to enhance understanding of attachment-informed practice (lines 135-138). However, the novelty and scientific contribution could be articulated more explicitly. In particular, the authors should clarify how this integrated approach advances existing knowledge beyond previous studies and explain why combining these data sources addresses an important gap in the literature.
6. Theoretical concepts are introduced but not operationalized. They include attachment theory, trauma-informed care, resilience, and internal working models. Yet authors do not clearly explain how these concepts relate to the measured variables. For example, how does "reference adults" represent attachment-informed caregiving? How is resilience measured?
7. The study combines IVIA, the observation checklist, and qualitative professional narratives. However, the authors do not explain why all three methods are necessary. (lines 132-138).
8. The manuscript does not explicitly state the study design. Authors should clearly identify the study design at the beginning of the Methods section, describe the overall design, and explain how the quantitative and qualitative components were incorporated into the study. 
9. The manuscript incorporates both quantitative and qualitative data; however, the methodological integration is insufficiently described. For the methods section, the author does not explain the qualitative component in adequate detail. Specifically, the authors should describe how the qualitative data were collected, the rationale for including open-ended responses, the thematic analysis procedures (e.g., coding process, coder(s), development of themes, measures to enhance trustworthiness), and how these qualitative findings complement the quantitative analyses. 
10. The manuscript lacks a dedicated Statistical Analysis subsection in the Methods. Although the Results report descriptive statistics, Mann–Whitney U tests, Spearman correlation analyses, multiple linear regression, and qualitative thematic analyses, the Methods do not describe the analytical strategy. The authors should specify the statistical software used, the significance threshold, the descriptive statistics reported, the statistical tests selected for each objective, the procedures for assessing test assumptions, the handling of missing data, and the calculation of effect sizes and confidence intervals. 
11. The section "3.5 Qualitative Findings provided by professionals" (line 511) states that a "thematic content analysis" was conducted. However, this analytical approach is not clearly defined. The authors should specify whether they employed thematic analysis (e.g., Braun & Clarke) or qualitative content analysis (e.g., Graneheim & Lundman or Elo & Kyngäs) and provide an appropriate methodological reference. If the term "thematic content analysis" is intentionally used, the analytical procedures should be described in sufficient detail to distinguish this approach from established qualitative methodologies.
12. The title of Section 3.6 "What Works with the Young Person" (line 550) appears stronger than the evidence presented. What Works" naturally implies that the study has identified interventions or practices that have been demonstrated to be effective. The qualitative findings reflect professionals' perceptions of effective relational practices rather than demonstrating which practices objectively "work" or improve resilience. Similarly, references to resilience or psychological well-being should be framed as perceived or theoretically inferred outcomes unless supported by direct empirical measures. The authors should ensure that the language throughout this section distinguishes participants' perspectives from evidence of intervention effectiveness.
13. Theme 5 ("Development of Socio-Emotional Competencies") appears conceptually distinct from the preceding themes (lines 584-587). While the manuscript frames the qualitative findings as professionals' perceptions of effective relational practices, this theme focuses primarily on structured socio-emotional interventions and skill-development programs rather than relational interactions between professionals and adolescents. The authors should clarify whether these interventions are considered attachment-informed relational practices or whether they represent a separate category of therapeutic or educational support.

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted file.

Comment 1:

“The introduction covers the following topics (lines 51-138): residential care, attachment theory, developmental trauma, therapeutic residential care, the role of professionals, and the research gap. A stronger introduction should answer questions such as What is already known? What remains uncertain? Why are previous findings insufficient? Why is this study needed now?

Response:
We improved the introduction including more references about the resilience topic and addressing the gap in research that justifies the present study.

Comment 2:

“Empirical research examining how residential care professionals perceive attachment-related behaviors remains limited..." (lines 122-128). Authors do not demonstrate this claim through the literature. Therefore, the gap appears asserted rather than established. The authors need to discuss which studies examined caregiver perceptions. What did those studies find? What methodological limitations existed? What questions remain unanswered?

Response:
We introduced discussion about studies with caregivers in residential care and addressed the questions that justifies the present study.

Comment 3:

“Although several recent references are cited, the citation functions mainly as supportive evidence and is rarely discussed critically. For instance, Martínez-Usarralde et al. (2025) is cited as evidence that caregiver relationships matter (lines 98-101). However, the review could explain: What variables were examined? What evidence supports attachment-informed care? What limitations did that review identify? How does the present study extend those findings?”

Response:

We acknowledge the relevance of the reviewer's comment. We introduced a sentence discussing an earlier study that points to the importance of addressing the specific issue of the use of attachment theory in residential care. We believe this can help clarify the specific gap we aim to fill with this work (lines 158-161).

Comment 4:

“Objective 3 uses the verb "explore" to describe the examination of associations. Given that inferential statistical analyses (Spearman correlations and multiple linear regression) were conducted, a more precise verb such as "examine" or "investigate" would better reflect the quantitative nature of the analyses and improve consistency with the study design.”

Response:
We substituted the verb “explore” by “examine”.

Comment 5:

“The author outlines its intended contribution by stating that it combines standardized attachment measures with professionals' perspectives to enhance understanding of attachment-informed practice (lines 135-138). However, the novelty and scientific contribution could be articulated more explicitly. In particular, the authors should clarify how this integrated approach advances existing knowledge beyond previous studies and explain why combining these data sources addresses an important gap in the literature.”

Response:
We introduced a more detailed explanation of the importance of the combination of data sources, and the specific contribute expected from this work. (lines 176-192)

Comment 6:

“Theoretical concepts are introduced but not operationalized. They include attachment theory, trauma-informed care, resilience, and internal working models. Yet authors do not clearly explain how these concepts relate to the measured variables. For example, how does "reference adults" represent attachment-informed caregiving? How is resilience measured?”

Response:
The measured variables are directly related with attachment theory, as they were developed in order to assess behavioural signs of the internal working models. In introduction is explained that professionals, as carers, have a major role contributing to repair the non secure attachment representations that many children have when they enter in residential care. Resilience is not a measured construct; it is addressed in this study through the implications of the qualitative data and the interpretation of qualitative results.

Comment 7:

“The study combines IVIA, the observation checklist, and qualitative professional narratives. However, the authors do not explain why all three methods are necessary. (lines 132–138).”

Response:
We agree that the rationale for combining the three sources of information was not sufficiently explicit in the original manuscript. We have therefore revised the Introduction to clarify the purpose of the multi-method approach. The three methods were intended to capture complementary dimensions of attachment-related functioning rather than provide redundant assessments. Specifically, the IVIA assesses adolescents’ self-reported attachment representations and therefore reflects their subjective relational perceptions. The Observation Checklist provides a complementary behavioral perspective by assessing attachment-related behaviors and socioemotional functioning observed by professionals in everyday residential care contexts. Finally, the qualitative professional narratives provide contextual and interpretative information regarding adolescents’ relational and emotional functioning and the caregiving strategies perceived as effective. The revised manuscript now explicitly explains that combining these sources allows attachment-related functioning to be examined across internal representations, observable behaviors, and relational-contextual experiences. This distinction is particularly relevant in the present study, as the findings showed limited convergence between adolescents’ self-reported attachment representations and professionals’ observational assessments, further illustrating the value of considering multiple sources of information.

Comment 8:

“The manuscript does not explicitly state the study design. Authors should clearly identify the study design at the beginning of the Methods section, describe the overall design, and explain how the quantitative and qualitative components were incorporated into the study.”

Response:
We agree that the study design was not explicitly identified in the original version of the manuscript. We have therefore added a new “Study Design” subsection at the beginning of the Materials and Methods section. The study is now explicitly described as an exploratory, cross-sectional, multi-method study incorporating quantitative and qualitative sources of information. The revised subsection clarifies that the quantitative component included adolescents’ self-reported attachment representations, sociodemographic information, and professionals’ observational assessments of attachment-related behaviors and socioemotional functioning. The qualitative component consisted of professionals’ narrative records concerning adolescents’ psychosocial and relational functioning and caregiving strategies perceived as effective. We have also clarified how these components were incorporated into the study. Quantitative and qualitative data were analyzed using methods appropriate to each type of data and were subsequently considered together at the interpretative level to provide complementary perspectives on adolescents’ internal attachment representations, observable relational behaviors, and relational experiences within residential care. This information has now been included at the beginning of the Methods section to make the overall methodological framework and the relationship between the quantitative and qualitative components explicit.

Comment 9:

Specifically, the authors should describe how the qualitative data were collected, the rationale for including open-ended responses, the thematic analysis procedures (e.g., coding process, coder(s), development of themes, measures to enhance trustworthiness), and how these qualitative findings complement the quantitative analyses.

Response:
The qualitative data are included in the Observation checklist and are provided as evidence of the chosen options. This information was included in the description of the Observation Checklist. The specific procedures adopted in the analysis of the qualitative data were described in section 2.5

Comment 10:

“The manuscript lacks a dedicated Statistical Analysis subsection in the Methods. Although the Results report descriptive statistics, Mann–Whitney U tests, Spearman correlation analyses, multiple linear regression, and qualitative thematic analyses, the Methods do not describe the analytical strategy. The authors should specify the statistical software used, the significance threshold, the descriptive statistics reported, the statistical tests selected for each objective, the procedures for assessing test assumptions, the handling of missing data, and the calculation of effect sizes and confidence intervals.”

Response:
We agree that the analytical strategy was insufficiently described in the original Methods section. We have therefore added a dedicated “Statistical and Qualitative Data Analysis” subsection providing a comprehensive description of the procedures used in the study. The revised subsection now specifies that quantitative analyses were conducted using IBM SPSS Statistics, version 29.0, with statistical significance set at p < .05. Categorical variables are reported as n (%), whereas continuous variables are summarized using M (SD) and, where informative, minimum and maximum values. We have also clarified the inferential analytical strategy. Mann–Whitney U tests were used to examine differences between secure and non-secure attachment groups, Spearman’s rank-order correlations were used to examine associations among the main study variables, and multiple linear regression using the Enter method was conducted to examine predictors of mental health risk. The revised subsection also specifies the procedures used to assess regression assumptions. We have additionally clarified the handling of missing data. There were no missing data for the quantitative variables included in the statistical analyses; therefore, no imputation or other missing-data procedures were required. For the qualitative component, blank narrative entries were treated as missing qualitative data rather than as evidence of the absence of a theme. Finally, the revised subsection describes the thematic content analysis applied to the professional narratives and clarifies the exploratory nature of the qualitative findings. These additions provide a more transparent and reproducible account of the analytical strategy used for both the quantitative and qualitative components of the study.

Comment 11:

The section "3.5 Qualitative Findings provided by professionals" (line 511) states that a "thematic content analysis" was conducted. However, this analytical approach is not clearly defined. The authors should specify whether they employed thematic analysis (e.g., Braun & Clarke) or qualitative content analysis (e.g., Graneheim & Lundman or Elo & Kyngäs) and provide an appropriate methodological reference. If the term "thematic content analysis" is intentionally used, the analytical procedures should be described in sufficient detail to distinguish this approach from established qualitative methodologies.

Response:
We agree that the term “thematic content analysis” used in the previous version of the manuscript was insufficiently precise and could create ambiguity regarding the qualitative analytical approach. We have therefore revised the manuscript to identify the approach consistently as inductive qualitative content analysis. The revised Methods section now provides a detailed description of the qualitative analytical process. Specifically, the analysis followed the three broad phases of preparation, organization, and reporting. The professional narratives were first read repeatedly to obtain an overall understanding of the material. Relevant meaning units were then identified and manually coded according to their manifest content. Codes expressing similar meanings were compared and inductively grouped into broader categories, which were subsequently reviewed and refined to represent recurrent patterns across the narratives. We have also clarified the researcher involvement in the analytical process. Coding and categorization were conducted manually by two researchers. Initial coding was performed by the last author and subsequently reviewed by the first author. The researchers compared and discussed the codes and emerging categories, with differences in interpretation resolved through discussion and consensus. No qualitative data analysis software was used. In addition, terminology has been standardized throughout the manuscript. The previous references to “thematic content analysis” have been replaced by “inductive qualitative content analysis,” and the findings are now consistently described in terms of qualitative categories rather than themes. Blank narrative fields are explicitly treated as missing qualitative data, and, given the brevity and heterogeneity of the narrative material and the substantial number of blank responses, the qualitative findings are characterized as exploratory. We believe these revisions clarify the methodological framework and provide sufficient detail to distinguish the analytical approach from thematic analysis and other qualitative methodologies.

Comment 12

The title of Section 3.6 "What Works with the Young Person" (line 550) appears stronger than the evidence presented. What Works" naturally implies that the study has identified interventions or practices that have been demonstrated to be effective. The qualitative findings reflect professionals' perceptions of effective relational practices rather than demonstrating which practices objectively "work" or improve resilience. Similarly, references to resilience or psychological well-being should be framed as perceived or theoretically inferred outcomes unless supported by direct empirical measures. The authors should ensure that the language throughout this section distinguishes participants' perspectives from evidence of intervention effectiveness.

Response:
The title of the section (now 3.7) was changed to “3.7. Descriptive Summary of Professionals’ Perceptions of Effective Support Strategies” and the language was revised considering the above comment of reviewer 2.

Comment 13

Theme 5 ("Development of Socio-Emotional Competencies") appears conceptually distinct from the preceding themes (lines 584-587). While the manuscript frames the qualitative findings as professionals' perceptions of effective relational practices, this theme focuses primarily on structured socio-emotional interventions and skill-development programs rather than relational interactions between professionals and adolescents. The authors should clarify whether these interventions are considered attachment-informed relational practices or whether they represent a separate category of therapeutic or educational support.

Response:
We agree that the category “Development of Socio-Emotional Competencies” differs from the preceding categories in that it primarily concerns structured developmental and educational interventions rather than relational practices embedded in everyday caregiver–adolescent interactions. We have therefore revised the manuscript to make this distinction explicit. The category has been renamed “Structured Socio-Emotional Skill Development” to better reflect its content. We now clarify that, whereas the preceding categories primarily represent relational caregiving practices—such as trust-building, emotional validation, consistent boundaries, and emotionally supportive care—this category represents a complementary form of structured developmental and educational support aimed at strengthening emotional regulation, social understanding, and interpersonal competencies. We have also clarified that these structured interventions are not considered attachment-informed relational practices in themselves. Rather, they may complement attachment-informed caregiving by supporting socio-emotional capacities that are relevant to adolescents’ relational and emotional functioning. The Results and Discussion have been revised accordingly to distinguish relational caregiving practices from structured socio-emotional interventions. In addition, the wording of the relevant study objective has been broadened from “effective relational practices” to “effective caregiving and supportive practices” to more accurately reflect the range of professional strategies captured by the qualitative data. We appreciate this comment, as it helped us clarify the conceptual organization of the qualitative findings and avoid conflating attachment-informed relational practices with broader developmental or educational interventions.

 

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript addresses an important topic, and I appreciate the difficulty of conducting research with adolescents in residential care. Combining adolescent self-reports with professional observations is potentially valuable. However, there are substantial concerns about the sample size, measurement, and analysis.

 

The study includes only 33 adolescents. This may be sufficient for an exploratory or pilot study, but it is very limited for a regression model with four predictors. The estimates are likely to be unstable and sensitive to a small number of cases. In addition, the adolescents came from five institutions, and 20 professionals rated 33 participants. Some observations may therefore be clustered by institution or evaluator. The authors should provide a power or sensitivity analysis and explain how this dependence was considered. If it cannot be addressed, the regression findings should be treated as preliminary.

 

The outcome described as “mental health risk” is not clearly defined. The Methods should specify which items were included, how the score was calculated, its possible range, and its reliability in this sample. The Observation Checklist is presented as a reflective and monitoring tool rather than a diagnostic instrument. Its use as a quantitative mental-health outcome therefore needs stronger justification.

 

The attachment classification procedure also requires clarification. It appears that participants were assigned to a category according to their highest IVIA dimension. The authors should confirm that this approach is part of the validated scoring procedure. Reliability coefficients for all IVIA dimensions should be reported. The acknowledged low reliability of the ambivalent dimension is especially important because this variable is used in several analyses.

 

I am also concerned about the Mann–Whitney comparisons. If the secure and non-secure groups were created from the same attachment scores that were later compared between the groups, the analysis is circular. The observed differences would largely follow from the classification rule itself. The authors should explain exactly how the groups were defined and remove these comparisons if the same variables were used for both classification and testing.

 

Several numerical inconsistencies need to be corrected. In Table 1, five avoidant cases correspond to 15.2%, not 11.9%. Two unclassified cases correspond to 6.1%, not 4.8%. The percentages currently do not total 100%. The minimum and maximum values for secure and ambivalent checklist scores in Table 2 also differ from those reported in the text. All tables and calculations should be checked against the original data.

 

The regression result for ambivalent behavior should be interpreted with caution. Its zero-order correlation with mental health risk is positive, but its regression coefficient becomes strongly negative. This may reflect suppression, overlap among predictors, or instability due to the small sample. The suggestion that adolescents who express attachment needs receive more support is plausible, but it was not tested in this study. It should be presented as a possible explanation rather than a finding.

 

The qualitative component is based on limited information. The paper states that 76.2% of the relevant entries were blank. The authors should report the exact number of usable responses and explain who coded the material, how the themes were developed, and whether the coding was independently reviewed. Representative quotations are also needed. Given the amount of missing material, this section may be more appropriately presented as a descriptive summary rather than a full thematic analysis.

 

Resilience is central to the title, abstract, and conclusion, but it was not directly measured. The study also cannot show that reference adults caused improvements in resilience or mental health. These claims should either be removed or clearly identified as theoretical interpretations.

 

The reference list needs further work. The number of references is not necessarily the main problem, but the coverage is limited for a paper addressing attachment, trauma, resilience, and residential care. More recent empirical research should be considered. The IVIA source cited in the Methods—Carvalho, Soares, and Baptista (2006)—is missing from the reference list. Graham is cited as 2006 in the text but listed as 2005. The relevance of Figueiredo (2023) to the claim it supports should also be checked. Reference formatting is inconsistent throughout.

 

Overall, the manuscript contains useful exploratory observations, but the conclusions are stronger than the current evidence allows. The study would need to be substantially reframed as a small pilot study, with simpler analyses and more cautious interpretation. A larger sample would be needed to support the present regression model and the broader claims.

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted file.

Comment: This manuscript addresses an important topic, and I appreciate the difficulty of conducting research with adolescents in residential care. Combining adolescent self-reports with professional observations is potentially valuable. However, there are substantial concerns about the sample size, measurement, and analysis.

Response:

We have substantially revised the manuscript in response to these concerns and have further tempered the interpretation of the findings.

Comment: The study includes only 33 adolescents. This may be sufficient for an exploratory or pilot study, but it is very limited for a regression model with four predictors. The estimates are likely to be unstable and sensitive to a small number of cases. In addition, the adolescents came from five institutions, and 20 professionals rated 33 participants. Some observations may therefore be clustered by institution or evaluator. The authors should provide a power or sensitivity analysis and explain how this dependence was considered. If it cannot be addressed, the regression findings should be treated as preliminary.

Response: Sample size, sensitivity, and dependence of observations. We agree that the sample of 33 adolescents is small for a four-predictor regression model. The regression was intended as an exploratory analysis, and this is now stated more explicitly throughout the Methods, Results, Discussion, and Limitations. We have additionally included a sensitivity analysis showing that, with N = 33, four predictors, α = .05, and 80% power, the study was adequately powered only to detect a relatively large overall effect (approximately Cohen’s f² = 0.43). We therefore acknowledge that smaller effects may have remained undetected and that individual regression coefficients may be unstable. We also agree that the nested structure of the data represents an important limitation. Participants were recruited from five residential care institutions and were assessed by 20 professionals, meaning that some observations may share institutional and/or evaluator-level influences. Given the small number of institutions and the limited total sample, reliable multilevel or cluster-adjusted modeling was not feasible. This limitation is now explicitly acknowledged, and the regression findings are described as preliminary and requiring replication in larger, appropriately nested samples.

Comment: The outcome described as “mental health risk” is not clearly defined. The Methods should specify which items were included, how the score was calculated, its possible range, and its reliability in this sample. The Observation Checklist is presented as a reflective and monitoring tool rather than a diagnostic instrument. Its use as a quantitative mental-health outcome therefore needs stronger justification.

Response: Mental health risk indicator. We agree that the term “mental health risk” was insufficiently defined and could be interpreted as referring to a validated diagnostic or psychometric measure. We have therefore clarified the terminology and description of this variable in the Methods. The variable was constructed as an exploratory observational composite indicator based on the frequency of risk-related behaviors recorded in the Observation Checklist, including behaviors such as substance use, self-harm, and addictive or other potentially harmful behaviors. Higher scores therefore represent a greater frequency of observed risk indicators rather than a diagnosis or a direct measure of psychopathology. We have also clarified how the composite score was calculated and its possible range. Because the included items represent heterogeneous risk behaviors rather than interchangeable indicators of a single underlying construct, internal consistency was not considered an appropriate reliability criterion. The manuscript now explicitly acknowledges that this composite indicator is exploratory and should not be interpreted as a validated mental-health measure. In summary, we have revised the Methods to provide a clearer definition of the variable previously referred to as “mental health risk,” including the items included in the score, scoring procedure, possible range, and internal consistency in the present sample. Importantly, we have clarified that the Observation Checklist is not a diagnostic instrument. Accordingly, this variable is now explicitly described as an observational indicator of psychological vulnerability derived from the checklist and is not interpreted as a clinical diagnosis or validated diagnostic mental health outcome. The purpose of its use in the present exploratory study was to examine whether attachment-related representations and observed behaviors were associated with variation in professionals’ observations of psychological vulnerability. The terminology and interpretation have been revised throughout the manuscript to reflect this distinction.

Comment: The attachment classification procedure also requires clarification. It appears that participants were assigned to a category according to their highest IVIA dimension. The authors should confirm that this approach is part of the validated scoring procedure. Reliability coefficients for all IVIA dimensions should be reported. The acknowledged low reliability of the ambivalent dimension is especially important because this variable is used in several analyses.

Response: We have expanded the description of IVIA scoring and now report reliability coefficients for all three attachment dimensions in the present sample. Particular caution is explicitly noted for the ambivalent dimension because of its lower internal consistency. The assignment to a category according to their highest IVIA dimension is part of the validated scoring procedure.

Comment: I am also concerned about the Mann–Whitney comparisons. If the secure and non-secure groups were created from the same attachment scores that were later compared between the groups, the analysis is circular. The observed differences would largely follow from the classification rule itself. The authors should explain exactly how the groups were defined and remove these comparisons if the same variables were used for both classification and testing. 

Response: Mann–Whitney analyses. Thank you for raising this important point. We agree that comparing secure and non-secure groups on the same IVIA dimensions used to derive the classification would constitute a circular analysis. We apologize that the previous wording did not make the analytical procedure sufficiently clear. The secure versus non-secure grouping was derived from adolescents’ self-reported IVIA profiles, whereas the variables compared using the Mann–Whitney U tests were the independently obtained observational attachment-related behavior scores from the professional-completed Observation Checklist. Thus, the variables used as outcomes in these comparisons were not the same variables used to establish group membership. We considered these comparisons relevant because they provide an exploratory examination of whether adolescents classified according to their self-reported attachment representations also differed in attachment-related behaviors observed independently by residential-care professionals. Nevertheless, given the small sample size and the exploratory nature of the study, we agree that these analyses should not be interpreted as validation of the IVIA classification or as confirmatory evidence of distinct attachment groups. We have revised the Methods, Results, and table labels to make the distinction between IVIA-derived groups and Observation Checklist scores explicit and have tempered the interpretation accordingly.

Comment: Several numerical inconsistencies need to be corrected. In Table 1, five avoidant cases correspond to 15.2%, not 11.9%. Two unclassified cases correspond to 6.1%, not 4.8%. The percentages currently do not total 100%. The minimum and maximum values for secure and ambivalent checklist scores in Table 2 also differ from those reported in the text. All tables and calculations should be checked against the original data. 

Response: Numerical inconsistencies. We thank the reviewer for identifying these discrepancies. All descriptive statistics and tables have been rechecked against the original data. Table 1 has been corrected so that five avoidant cases are reported as 15.2% and two unclassified cases as 6.1%. The minimum and maximum values reported for the observational attachment dimensions in Table 2 have also been checked against the original dataset and corrected to ensure consistency between the table and the accompanying text. We additionally conducted a broader check of numerical reporting throughout the manuscript.

Comment: The regression result for ambivalent behavior should be interpreted with caution. Its zero-order correlation with mental health risk is positive, but its regression coefficient becomes strongly negative. This may reflect suppression, overlap among predictors, or instability due to the small sample. The suggestion that adolescents who express attachment needs receive more support is plausible, but it was not tested in this study. It should be presented as a possible explanation rather than a finding.

Response: Regression coefficient for ambivalent behavior. We agree that the negative regression coefficient for observed ambivalent behavior requires particular caution because its direction contrasts with the positive zero-order association with the psychological vulnerability indicator. The Results now explicitly note this sign reversal and acknowledge that it may reflect suppression, shared variance among predictors, or instability associated with the small sample. We have also revised the Discussion to clarify that the possibility that overt expression of attachment needs may elicit greater caregiver responsiveness is a possible explanatory hypothesis only; caregiver responsiveness to these behaviors was not directly tested in the present study and is no longer presented as an empirical finding.

Comment: The qualitative component is based on limited information. The paper states that 76.2% of the relevant entries were blank. The authors should report the exact number of usable responses and explain who coded the material, how the themes were developed, and whether the coding was independently reviewed. Representative quotations are also needed. Given the amount of missing material, this section may be more appropriately presented as a descriptive summary rather than a full thematic analysis.

Response: Qualitative component. The qualitative Methods had already been expanded to specify that the narrative material was analyzed manually using inductive qualitative content analysis. Coding and categorization were conducted by two researchers: initial coding was performed by the last author and subsequently reviewed by the first author, with codes and categories discussed and differences in interpretation resolved through consensus. No qualitative analysis software was used. We nevertheless agree that the substantial proportion of blank responses limits the depth of the qualitative material. We have therefore further revised the presentation of these findings. The manuscript now reports the exact number of usable and blank responses and presents the findings as an exploratory descriptive qualitative summary derived from inductive content analysis, rather than implying an in-depth qualitative analysis. Representative anonymized quotations have also been added to illustrate the categories where sufficient narrative material was available. The limitations associated with the amount and heterogeneity of the qualitative material are now explicitly acknowledged. 

Comment: Resilience is central to the title, abstract, and conclusion, but it was not directly measured. The study also cannot show that reference adults caused improvements in resilience or mental health. These claims should either be removed or clearly identified as theoretical interpretations.

Response: Resilience construct: We thank the reviewer for the comment. In fact the resilience construct is not measured; qualitative data includes the professionals perspectives about effective relational practices and the assertions about resilience reflect this specific component of the study, as they may contribute to promote resilience and do not intend to express that the study evidences its effectiveness. The language was revised in order to clarify this aspect.

Comment: The reference list needs further work. The number of references is not necessarily the main problem, but the coverage is limited for a paper addressing attachment, trauma, resilience, and residential care. More recent empirical research should be considered. The IVIA source cited in the Methods—Carvalho, Soares, and Baptista (2006)—is missing from the reference list. Graham is cited as 2006 in the text but listed as 2005. The relevance of Figueiredo (2023) to the claim it supports should also be checked. Reference formatting is inconsistent throughout.

Response: we added more recent references in text and in references list. The list formatting and content was also revised.

Comment: Overall, the manuscript contains useful exploratory observations, but the conclusions are stronger than the current evidence allows. The study would need to be substantially reframed as a small pilot study, with simpler analyses and more cautious interpretation. A larger sample would be needed to support the present regression model and the broader claims.

Response: These revisions substantially temper the claims made from the quantitative and qualitative findings and more clearly position the study as an exploratory investigation whose results require replication in larger, independently structured samples.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I appreciate the thought and care the authors have taken in addressing my concerns. I believe the paper is stronger than in the initial submission. Especially given the addition of the tables and overall organization. I also saw the engagement with recent scholarship and clarifications on previously unclear ideas or concepts related to care. Thank you for the oppertunity to review this work. 

Author Response

Comment 1:

I appreciate the thought and care the authors have taken in addressing my concerns. I believe the paper is stronger than in the initial submission. Especially given the addition of the tables and overall organization. I also saw the engagement with recent scholarship and clarifications on previously unclear ideas or concepts related to care. Thank you for the oppertunity to review this work.

Response

Thank you very much for your thoughful comments and the time you dedicated to our work.

Reviewer 2 Report

Comments and Suggestions for Authors

1. The introduction needs stronger rationale for answering the following questions: What is already known? What remains uncertain? Why are previous findings insufficient? Why is this study needed now? The introduction still reads predominantly as a narrative accumulation of supportive literature rather than a sharply structured argument showing exactly what prior studies have failed to establish.
2. A research gap regarding professionals’ perceptions was asserted rather than demonstrated. The original request was broader: discuss which studies examined caregiver perceptions, their findings, methodological limitations, and remaining questions. That comparative critical synthesis is still limited.
3. Martínez-Usarralde et al. is still mainly used as supporting evidence that sensitive caregivers improve outcomes rather than critically discussing its variables, evidence, limitations, or how the current study extends it. Literature should be critically discussed.
4. The response correctly acknowledges that resilience is not directly measured. However, the manuscript still shows some statements that sound as though resilience itself was empirically demonstrated. For example, the discussion states that reference adults “appear to constitute a central protective mechanism” through which adolescents develop greater resilience. More importantly, the abstract states that “resilience…is strengthened through stable, emotionally attuned relationships,” despite resilience not being measured. 
5. The authors added software, α, descriptive statistics, inferential tests, regression assumptions, missing-data handling, and sensitivity analysis. However, the requested reporting/calculation of effect sizes and confidence intervals has not been adequately addressed according to Comment 10. Their response itself does not explain why this part of the comment was omitted.

Author Response

Comment 1

The introduction needs stronger rationale for answering the following questions: What is already known? What remains uncertain? Why are previous findings insufficient? Why is this study needed now? The introduction still reads predominantly as a narrative accumulation of supportive literature rather than a sharply structured argument showing exactly what prior studies have failed to establish.

Response 1

We understand the revisors’ questions and we acknowledge its importance. The present exploratory study introduces a new approach to how attachment can be assessed in residential care contexts in order to address the gap between attachment theory and attachment based practices; the introduction of the article justifies why its important an observational measure focused in daily behaviour of the young persons, based on items that professionals can observe in their practices  and are used to do it. Prior studies have not failed to establish; prior studies did not address specifically this issue. However a specific reference (Vorria et al., 2015) was included to justify the importance of the approach and other reference of a recent scoping review in out-of-home care was also introduced (Shein, 2025), as it shows very clearly that this approach (an observational measure used by carers in residential setting) was not used yet. This is the reason why the introduction is focused in showing how and why this new stance can be supported. This is done considering the associations between residential care, attachment theory and the importance of the relational work done by professionals with adolescents. We introduced new in lines 164-174 in order to clarify the revisor’s questions.

Comment 2

A research gap regarding professionals’ perceptions was asserted rather than demonstrated. The original request was broader: discuss which studies examined caregiver perceptions, their findings, methodological limitations, and remaining questions. That comparative critical synthesis is still limited.

Response 2

We included new references in order to demonstrate the research gap. It is not possible to make a comparative critical synthesis as there's not research with adolescents in residential care using observational tools.

Comment 3

Martínez-Usarralde et al. is still mainly used as supporting evidence that sensitive caregivers improve outcomes rather than critically discussing its variables, evidence, limitations, or how the current study extends it. Literature should be critically discussed.

Response 3

The study of Martinez-Usarralde is a systematic review that supports the importance of the relations between carers and the adolescents in residential care. That is its main conclusion; the systematic review don’t consider how attachment is assessed in residential care and for that reason we consider that, in this article, it won’t be adequate to beside this specific point. We introduced the nature of the work in the text  (lines 129-130) to clarify the importance of this reference.

Comment 4

The response correctly acknowledges that resilience is not directly measured. However, the manuscript still shows some statements that sound as though resilience itself was empirically demonstrated. For example, the discussion states that reference adults “appear to constitute a central protective mechanism” through which adolescents develop greater resilience. More importantly, the abstract states that “resilience…is strengthened through stable, emotionally attuned relationships,” despite resilience not being measured. 

Response 4

We corrected the abstract and the text to avoid references to resilience as it was a measured construct, and acknowledging the source of the assertions done about resilience (only qualitative data).

Comment 5

The authors added software, α, descriptive statistics, inferential tests, regression assumptions, missing-data handling, and sensitivity analysis. However, the requested reporting/calculation of effect sizes and confidence intervals has not been adequately addressed according to Comment 10. Their response itself does not explain why this part of the comment was omitted.

Response 5

We agree that the previous revision did not adequately address the reviewer’s request concerning the reporting of effect sizes and confidence intervals. We have now revised the Statistical Analysis and Results sections to explicitly report effect-size estimates and 95% confidence intervals for the relevant inferential analyses. For the Mann–Whitney U comparisons, effect sizes are reported as r, together with their 95% confidence intervals. For the multiple linear regression, unstandardized regression coefficients (B) are reported with their corresponding 95% confidence intervals, while standardized coefficients (β) are retained to facilitate interpretation of the relative magnitude and direction of associations. We have also added an explicit statement to the Statistical Analysis subsection specifying how effect sizes and confidence intervals are reported. Given the small sample size and exploratory nature of the study, effect-size estimates and their confidence intervals are interpreted cautiously, with emphasis on estimation and precision rather than statistical significance alone. These revisions have been incorporated into the Statistical Analysis subsection, and the multiple regression results/Table 6.

Reviewer 3 Report

Comments and Suggestions for Authors

Thank you for the extensive revision. The paper is much clearer than the version I previously reviewed, and the authors have taken a few of my concerns seriously. The numerical inconsistencies have been corrected, the IVIA-based classification is now described more cautiously, and the qualitative procedures are explained in greater detail. I also appreciate the more open discussion of the small sample and the possible instability of the regression results.

 

The paper is now easier to see as an exploratory study, and I would be willing to reconsider it on that basis. Several substantial issues remain, however, particularly in relation to the regression analysis and the way the findings are interpreted.

 

My main concern is still the regression. The model includes four predictors but is based on only 33 adolescents drawn from five institutions and assessed by 20 professionals. Under these conditions, the estimates may be strongly influenced by one institution, one rater, or a few unusual observations. The power calculation added in the revision is informative, but it addresses the size of the overall effect that could be detected; it does not establish that the individual coefficients are reliable. The authors should examine influential cases and report confidence intervals for the coefficients. It would also be useful to re-estimate the model after removing each institution in turn. If the direction or magnitude of the main coefficients changes substantially, I do not think the regression should remain central to the paper.

 

More information is needed about the observational risk score. The revised Methods state that five items were summed to produce a score from 0 to 15, but the individual items and their coding are not fully reported. Readers need to know exactly what contributes to this outcome. Its distribution is also important. A score based on self-harm, substance use, and other relatively uncommon behaviors may contain many zeros and may be quite skewed. If so, ordinary linear regression may not be suitable. The authors should report the score distribution and show that the main result is not an artifact of the chosen model or of a small number of high-risk cases.

 

I was also unable to determine how the predictors in the regression were selected. The study includes several IVIA and checklist dimensions, but the final model contains avoidant IVIA, secure IVIA, observed ambivalent behavior, and sex. Were these variables specified before the analyses were conducted, or were they chosen after examining the correlations? This distinction matters in such a small sample. If the model was data-driven, the p values should not be interpreted in the same way as those from a prespecified model. A simpler analysis based on a clear theoretical rationale may be more appropriate here.

 

The structure of the professional ratings is still unclear. Both the risk score and the observed attachment dimensions come from the same checklist, apparently completed by the same group of professionals. This raises the possibility of shared-rater or common-method effects. The manuscript should state how many adolescents were rated by each professional, whether each adolescent was rated by one person or by a staff team, and how participants were distributed across institutions.

 

Resilience remains too prominent given that it was not measured. The additional literature is useful, but it does not change what the present data can establish. The study can discuss the possible relevance of stable relationships for resilience, but it cannot conclude that resilience was strengthened or that reference adults reorganized adolescents’ internal working models. Several statements in the Abstract, Discussion, practice implications, and Conclusion still read as findings of the study when they are more accurately theoretical interpretations. The title may also need to be reconsidered if resilience continues to be presented as one of the study’s main outcomes.

 

The IVIA results need another careful check. The revised manuscript reports an alpha of .71 for the secure dimension and .91 for “the other dimensions,” whereas the earlier version described the internal consistency of the ambivalent dimension as low. Since the sample and questionnaire do not appear to have changed, this difference is difficult to understand. Alpha should be reported separately for all three dimensions, and the authors should explain why the assessment of the ambivalent scale changed between versions.

 

The qualitative section has improved, particularly through the addition of quotations. Even so, the evidence concerning effective support strategies is based on only eight usable responses, from which five categories are derived. The number of responses contributing to each category should be reported. It is also not clear whether Sections 3.6 and 3.7 draw on the same narrative material or on separate questions. The relevant number of responses should be stated for each analysis. Given the small amount of material, this part is better described as a descriptive summary of professional observations than as a developed qualitative analysis.

 

The Observation Checklist was developed for young people aged 11–16, whereas the present sample includes participants aged 17 and 18. The argument that the original authors did not explicitly prohibit its use with older adolescents is not sufficient evidence of validity. The authors should report how many participants were older than 16 and examine whether excluding them changes the findings. At the very least, this needs to be acknowledged as a measurement limitation.

Author Response

Comment

Thank you for the extensive revision. The paper is much clearer than the version I previously reviewed, and the authors have taken a few of my concerns seriously. The numerical inconsistencies have been corrected, the IVIA-based classification is now described more cautiously, and the qualitative procedures are explained in greater detail. I also appreciate the more open discussion of the small sample and the possible instability of the regression results.

The paper is now easier to see as an exploratory study, and I would be willing to reconsider it on that basis. Several substantial issues remain, however, particularly in relation to the regression analysis and the way the findings are interpreted.

Response

We thank the revisor for the detailed revision of our work.

Comment 1

My main concern is still the regression. The model includes four predictors but is based on only 33 adolescents drawn from five institutions and assessed by 20 professionals. Under these conditions, the estimates may be strongly influenced by one institution, one rater, or a few unusual observations. The power calculation added in the revision is informative, but it addresses the size of the overall effect that could be detected; it does not establish that the individual coefficients are reliable. The authors should examine influential cases and report confidence intervals for the coefficients. It would also be useful to re-estimate the model after removing each institution in turn. If the direction or magnitude of the main coefficients changes substantially, I do not think the regression should remain central to the paper.

 Response 1

We agree that the small sample size and the clustering of participants across institutions and professionals require particular caution when interpreting individual regression coefficients. We therefore conducted additional sensitivity analyses. Influence diagnostics identified one observation exceeding the conventional Cook’s distance threshold, although no observation exceeded the leverage threshold. Excluding this observation did not change the substantive pattern of the results. We have also reported 95% confidence intervals for all regression coefficients. As suggested, we additionally re-estimated the regression after excluding each institution in turn. The direction of all regression coefficients remained unchanged across the five reduced-sample models, with R² values ranging from .328 to .530. However, the magnitude and statistical significance of individual coefficients varied, particularly when either of the two largest institutions was excluded (N = 23). The Secure Checklist coefficient remained negative across all models, whereas the Avoidant IVIA coefficient remained positive but was not statistically significant in all reduced samples. These analyses suggest that the overall directional pattern was not driven by a single institution or influential observation, but they also confirm that the precision of individual coefficients is sensitive to sample composition. We have therefore revised the Results and Limitations to report these sensitivity analyses and further emphasize that the regression is exploratory and that individual coefficients should be interpreted cautiously.

Comment 2

More information is needed about the observational risk score. The revised Methods state that five items were summed to produce a score from 0 to 15, but the individual items and their coding are not fully reported. Readers need to know exactly what contributes to this outcome. Its distribution is also important. A score based on self-harm, substance use, and other relatively uncommon behaviors may contain many zeros and may be quite skewed. If so, ordinary linear regression may not be suitable. The authors should report the score distribution and show that the main result is not an artifact of the chosen model or of a small number of high-risk cases.

 Response 2

We have revised the Methods to report the five items contributing to the observational mental health risk indicator, their coding (A = 4 to E = 0), and the calculation of the composite score. We also examined the outcome distribution. Only one participant had a score of zero, skewness was 0.32, and the Shapiro–Wilk test was non-significant (p = .131), indicating no substantial zero inflation or marked departure from normality. To assess whether the regression findings were driven by influential observations, we examined Cook's distance, leverage, and studentized residuals. One observation exceeded the conventional 4/N Cook's distance threshold (D = .189). We therefore repeated the regression excluding this case. The substantive findings remained unchanged and became somewhat stronger (R² = .499): Avoidant IVIA remained positively associated with mental health risk (B = 1.39, p = .004), Secure Checklist remained negatively associated with risk (B = −0.104, p < .001), and sex remained significant (B = 0.91, p = .042). Thus, the main findings were not attributable to a small number of high-risk cases. These additional methodological details and sensitivity analyses have been added to the revised manuscript.

Comment 3

I was also unable to determine how the predictors in the regression were selected. The study includes several IVIA and checklist dimensions, but the final model contains avoidant IVIA, secure IVIA, observed ambivalent behavior, and sex. Were these variables specified before the analyses were conducted, or were they chosen after examining the correlations? This distinction matters in such a small sample. If the model was data-driven, the p values should not be interpreted in the same way as those from a prespecified model. A simpler analysis based on a clear theoretical rationale may be more appropriate here.

Response 3

We agree that the basis for predictor selection was not sufficiently clear in the previous version. The regression model was not specified in a preregistered analysis plan and should therefore be regarded as exploratory rather than confirmatory. Given the small sample (N = 33), we intentionally restricted the model to a parsimonious set of attachment-related variables and sex. Avoidant and Secure IVIA scores represented adolescents' self-reported insecure and secure attachment-related representations, respectively, while the Secure Checklist score provided an independent observational indicator of secure relational behavior. Sex was included as a basic demographic covariate. Additional attachment dimensions and sociodemographic variables were not entered to avoid overparameterization. We also identified and corrected a labeling error in the regression table: the observational predictor included in the model was the Secure Checklist dimension, not the Ambivalent Checklist dimension. We have revised the Statistical Analysis section to make this rationale explicit and now consistently describe the regression as exploratory. We have also tempered the interpretation of individual p values and emphasize the direction and magnitude of associations rather than treating them as confirmatory evidence.

Comment 4

The structure of the professional ratings is still unclear. Both the risk score and the observed attachment dimensions come from the same checklist, apparently completed by the same group of professionals. This raises the possibility of shared-rater or common-method effects. The manuscript should state how many adolescents were rated by each professional, whether each adolescent was rated by one person or by a staff team, and how participants were distributed across institutions.

Response 4

We have revised the manuscript to clarify the structure of the professional ratings. Each adolescent was rated by one reference professional rather than by a staff team. Across the sample, 13 professionals rated one adolescent each, six rated two adolescents each, and one rated eight adolescents. We have also reported the distribution of adolescents and professionals across the five participating institutions. We further clarified that the observed attachment dimensions and the observational mental health risk indicator were derived from the same Observation Checklist completed by the same professional for each adolescent. We therefore explicitly acknowledge shared-rater/common-method variance and potential within-evaluator and within-institution dependence as limitations of the study, and these findings are interpreted cautiously.

Comment 5

Resilience remains too prominent given that it was not measured. The additional literature is useful, but it does not change what the present data can establish. The study can discuss the possible relevance of stable relationships for resilience, but it cannot conclude that resilience was strengthened or that reference adults reorganized adolescents’ internal working models. Several statements in the Abstract, Discussion, practice implications, and Conclusion still read as findings of the study when they are more accurately theoretical interpretations. The title may also need to be reconsidered if resilience continues to be presented as one of the study’s main outcomes.

Response 5

We corrected the abstract and the text to avoid references to resilience as it was a measured construct, and acknowledging the qualitative data as the source of assertions about resilience, and also its limitations.

Comment 6

The IVIA results need another careful check. The revised manuscript reports an alpha of .71 for the secure dimension and .91 for “the other dimensions,” whereas the earlier version described the internal consistency of the ambivalent dimension as low. Since the sample and questionnaire do not appear to have changed, this difference is difficult to understand. Alpha should be reported separately for all three dimensions, and the authors should explain why the assessment of the ambivalent scale changed between versions.

Response 6

In the previous version, Cronbach’s alpha for the IVIA dimensions was calculated/reported incorrectly, which led to an inaccurate assessment of the internal consistency of the ambivalent dimension. We apologize for this error. In the revised manuscript, the reliability analysis was checked and the coefficients are now reported separately for the Secure, Ambivalent, and Avoidant dimensions to ensure clarity and consistency. The corrected Cronbach’s alpha coefficients were .71 for the Secure dimension, α = .91 for the Ambivalent dimension, and α = .91 for the Avoidant dimension.

Comment 7

The qualitative section has improved, particularly through the addition of quotations. Even so, the evidence concerning effective support strategies is based on only eight usable responses, from which five categories are derived. The number of responses contributing to each category should be reported. It is also not clear whether Sections 3.6 and 3.7 draw on the same narrative material or on separate questions. The relevant number of responses should be stated for each analysis. Given the small amount of material, this part is better described as a descriptive summary of professional observations than as a developed qualitative analysis.

Response 7

We have further clarified the qualitative component. Sections 3.6 and 3.7 draw on separate narrative material: Section 3.6 is based on the descriptive evidence fields accompanying the Observation Checklist, for which 17 of the 33 records contained at least one usable narrative entry, whereas Section 3.7 is based on a separate open-ended question concerning effective support strategies, for which 9 of the 33 records contained a usable response. For Section 3.7, we have also added the number of responses contributing to each descriptive category: Supportive and Trusting Relationships (n = 6), Emotional Validation and Empathic Communication (n = 5), Clear Structure, Boundaries, and Guidance (n = 5), Emotional Support and Attachment-Based Care (n = 5), and Structured Socio-Emotional Skill Development (n = 2). As individual responses could contribute to more than one category, these frequencies are not mutually exclusive. Given the limited amount of narrative material, we have retained the revised designation of Section 3.7 as a “Descriptive Summary of Professionals’ Perceptions of Effective Support Strategies” and explicitly frame these findings as exploratory and descriptive rather than as a developed qualitative analysis.

Comment 8

The Observation Checklist was developed for young people aged 11–16, whereas the present sample includes participants aged 17 and 18. The argument that the original authors did not explicitly prohibit its use with older adolescents is not sufficient evidence of validity. The authors should report how many participants were older than 16 and examine whether excluding them changes the findings. At the very least, this needs to be acknowledged as a measurement limitation.

Response 8

We agree that the absence of an explicit restriction above age 16 cannot be considered evidence of validity for older adolescents. Nine participants in our sample were older than 16 (eight aged 17 and one aged 18), and we have now explicitly acknowledged this as a measurement limitation. As suggested, we repeated the regression analysis excluding these nine participants (N = 24). The model remained statistically significant (R² = .438, p = .022), and the direction of all coefficients was unchanged. However, some individual coefficients were less precise and no longer statistically significant in the reduced sample. The Methods, Results, and Limitations have been revised accordingly. We now explicitly state that the validity of the Observation Checklist for adolescents aged 17–18 has not been established and that findings involving this age group should therefore be interpreted with caution.