Navigating Social Inclusion: How Social and Cognitive Factors Relate to Friendship Quality in Children with ADHD, Dyslexia, and Neurotypical Development

Abstract

Friendships contribute to children’s social inclusion at school. Children with Attention Deficit/Hyperactivity Disorder (ADHD) or dyslexia experience substantial difficulty maintaining meaningful friendships. This study aims to elucidate the direct and indirect influence of social understanding on friendship quality features and executive functions in primary school children with neurotypical development (NT), ADHD, or dyslexia. Participants were divided into three groups of 64 children each (Mage = 9.77 years, SD = 1.22). Self-report research instruments and tasks were individually administered to examine children’s friendship quality, social understanding (beliefs, empathy, emotion regulation), and executive functions (working memory, inhibition, cognitive flexibility). Results showed that in children with NT, beliefs and emotions as components of social understanding mediated the positive relationship between all executive functions and friendship quality, after controlling for gender. In children with ADHD or dyslexia, only beliefs mediated the relationship between working memory and friendship quality. These findings highlight the dynamic role of social understanding in children’s cognitive development and their friendships, leading to psychosocial adjustment and school inclusion.

1. Introduction

Research findings highlight the significant role of friendships in children’s psychosocial and emotional development, well-being, and school adjustment and performance (Eggum-Wilkens et al., 2014). Particularly, friendships have been reported to provide opportunities for children’s enhancement of sociocognitive and emotional skills, such as acknowledging others’ perspectives (theory of mind—ToM) and emotions (affective empathy), managing their emotions (emotion regulation—ER), and organizing and controlling their thoughts and actions (executive functions—EFs) (Carpendale & Lewis, 2006; Garcia-Andres et al., 2010; Sendil & Erden, 2014). On the other hand, social understanding (i.e., ToM, empathy, and ER) and EFs are often recognized as important skills required to shape children’s social functioning, enabling them to understand others’ perceptions or social situations and to respond effectively to them (O’Toole et al., 2017; Weimer et al., 2021). This study draws on the Social Information Processing model (Crick & Dodge, 1994), which posits that both cognitive and emotional regulation processes underpin children’s successful peer relationships. However, research including social understanding and EFs to explore possible differential effects on friendship relationships in neurotypical (NT) or neurodivergent (ND) children is quite limited (Kouvava et al., 2022, 2025a, 2025b; Miller et al., 2018; Neprily et al., 2025; X. Wang & Feng, 2024; Weimer et al., 2021).

As in school, the main concern of the educational community is cognitive development and academic achievement; children’s social relationships are often overlooked. Consequently, many ND children with social skills deficits face difficulties in forming and maintaining close friendships with their peers, and frequently they befriend younger or older children with similar problems or antisocial or even deviant behaviors (Bunford et al., 2018; Marton et al., 2015). In addition, numerous studies on Attention Deficit/Hyperactivity Disorder (ADHD) and dyslexia focus on the cognitive, learning or behavioral problems children encounter (A. John et al., 2022; Löytömäki et al., 2023; Nadeau et al., 2020), neglecting to scrutinize the psychosocial and emotional dimensions of each disorder. In Greece specifically, research focusing mainly on ND and NT children’s social status (Antonopoulou et al., 2019; Statiri & Andreou, 2017) or friendships (Antonopoulou et al., 2022; Avramidis et al., 2018, 2022) is quite limited. Published work on children with ADHD or dyslexia and their friendships or the sociocognitive factors affecting their friendships is scant (Kouvava et al., 2025a, 2025b, 2022; Neprily et al., 2025).

Gaining insight into the development of peer relationships in children is of great significance. Nevertheless, there has been limited research exploring the connections among NT children’s friendships, social understanding, and EFs, and none in the case of ND children. Thus, the aim of the present study is to explore the interrelationship among social understanding, EFs, and primary school children’s perceptions of their friendship quality across NT and ND groups. Specifically, we hypothesized that aspects of children’s social understanding, such as emotion understanding (including affective empathy and ER) and belief understanding (including 2nd order and advanced ToM), will mediate the relationship between EFs, including inhibitory control, cognitive flexibility and working memory, and friendship quality characteristics, both positive and negative. In addition, we sought to examine whether the hypothesized mediation pathways may differ across NT and ND children, given the distinct cognitive and socio-emotional profiles associated with ADHD and dyslexia. Comparing these groups provides insight into both shared and disorder-specific pathways affecting peer relationships.

1.1. Friendships, Social Understanding, and Executive Functions (EFs) in Children with and Without Neurodevelopmental Disorders

Friendships with peers are important interpersonal relationships, as they are voluntary, reciprocal, and egalitarian in structure (Hartup & Stevens, 1997). Children share things and enjoy similar activities with their friends and acknowledge one another’s needs and emotions, whereas adolescents seek support, reciprocity, loyalty, intimacy, and self-disclosure from their friends (Rose et al., 2022). Friendships, whether casual, close or best, based on their stability and commitment (Policarpo, 2015), can greatly vary in their quality features. Thus, some friends attribute positive qualities to their friendships, such as companionship, caring, intimacy, and support, while others attribute negative qualities, such as conflict, betrayal, jealousy, exclusiveness, and antagonism (Rokeach & Wiener, 2022). Finally, having close and stable friendships of high (positive) quality promotes children’s identity-building process and self-esteem and serves as a protective factor against loneliness and depression (Bagwell & Bukowski, 2018; Neprily et al., 2025). On the other hand, the lack of friends has been linked to poor school adjustment and deviant behaviors, such as bullying others or being victimized (Schwartz-Mette et al., 2020).

Moreover, social understanding encompasses the acknowledgement of emotions (affective empathy) and their management (emotion regulation—ER), as well as the beliefs, perspectives, and intentions of others (ToM), skills that are essential to navigate social interactions (Carpendale & Lewis, 2006; Garcia-Andres et al., 2010; Perry & Shamay-Tsoory, 2013). It has been suggested that children with close friends present higher levels of social understanding than those without friends (Peterson & Siegal, 2002), and conversely, a well-developed social understanding enhances social skills, such as empathy, cooperation, help, intimacy, self-regulation, and conflict resolution, that are prerequisites for high-quality friendships (Miller et al., 2018; Weimer et al., 2021). For example, close and high-quality friendships, even among preschool children, have been found to enhance mental-state talk related to ToM (Ruffman et al., 2002). Additionally, children with well-developed affective empathy have positive attitudes towards their friends (Meuwese et al., 2017), and their friendships are characterized by intimacy and support (Ciarrochi et al., 2017; de Wied et al., 2007; van den Bedem et al., 2019). Finally, emotionally competent children who acknowledge different reactions to the same situations based on prior experiences and beliefs and control their emotions (Gross, 2015) are more popular (Y. Wang et al., 2019) and engage in qualitative friendships (Ciarrochi et al., 2008).

In the present study, affective empathy and ER were combined into a single variable, termed “Emotions,” due to the strong conceptual and empirical overlap between these constructs. Statistically, measures of affective empathy and ER exhibited a significant and positive overall association (r = 0.19, p < 0.001), suggesting that they were capturing a shared underlying construct (Yavuz et al., 2024). Theoretically, we argue that affective empathy, while involving the initial experience of another’s emotions, necessitates emotion regulation to prevent emotional overwhelm and promote adaptive social responding. Thus, while acknowledging the potential loss of information regarding the independent contributions of each construct, this approach allowed us to examine the mediating role of emotions and their management with regard to the executive functions and the friendship quality, particularly in ND children, where deficits in both affective empathy and emotion regulation are frequently observed (Berenguer Forner et al., 2017; Kouvava et al., 2022).

Furthermore, Baggetta and Alexander (2016), in a review of 106 studies assessing behavioral manifestations of EFs in an individual’s daily life, found that the most commonly mentioned EFs were inhibition (68%), working memory (35%), and cognitive flexibility (31%). Research associates children’s EFs with their psychosocial adjustment (Benavides-Nieto et al., 2017), the quality and the stability of their interpersonal relationships with adults and peers (Holmes et al., 2016; Woltering et al., 2016), and their communication skills (Harris & Orth, 2020; X. Wang & Feng, 2024). Specifically, inhibition helps children avoid setting revenge goals when feeling wronged by their friends, and cognitive flexibility facilitates the mitigation of interpretations of friends’ actions (Miller et al., 2020).

While working memory, inhibitory control, and cognitive flexibility are often discussed together under the umbrella of EFs (Diamond, 2013; Miyake et al., 2000), treating them as distinct but interrelated constructs provides a more nuanced understanding of child development (Sambol et al., 2025). In the present study, inhibition, working memory, and cognitive flexibility were consolidated into two composite variables based on the scoring metrics employed. Specifically, working memory was assessed primarily through error rates, reflecting the accuracy of maintaining and manipulating information. The remaining EFs, inhibition and cognitive flexibility, were combined into a single variable characterized by response time, thus capturing the efficiency with which participants could suppress prepotent responses and adapt to changing task demands. This stratification aimed to provide a parsimonious yet informative representation of EF performance, distinguishing between the accuracy-driven domain of working memory and the efficiency-focused domains of inhibition/cognitive flexibility (Ger & Roebers, 2023). This approach acknowledges the distinct cognitive demands assessed by each composite variable while also streamlining statistical analyses.

Neurodivergent (ND) children, such as children with ADHD or dyslexia, have difficulties in having close and mutual friends (Leseyane et al., 2018; Spender et al., 2023). It has been suggested that the severity of their symptoms and the difficulty regulating their behaviors and emotions have a significant impact on the number of their friends and the stability of their friendships (Normand et al., 2024). For example, children with ADHD often complain that they have few and not very stable friendships characterized by poor quality (McQuade et al., 2021), with more conflicts and less connectedness (Hoza et al., 2005; Normand et al., 2011). Moreover, children with dyslexia also face challenges in their friendships. Research indicates that children with dyslexia often feel insecure, anxious or embarrassed at school; derive little satisfaction from life; and have reduced mental resilience (Mammarella et al., 2016; Undheim et al., 2011). Thus, they tend to express fewer positive emotions, such as happiness, resulting in having difficulties in forming stable friendships at school, characterized by joy, respect, and intimacy (Kalka & Lockiewicz, 2018).

Children with ADHD have low tolerance levels and experience challenges in understanding the mental states of others (Berenguer Forner et al., 2017; Mary et al., 2016), exhibiting behaviors that result in their rejection from their peer group and low-quality friendships (Mrug et al., 2009; Parker et al., 2006; Rielly et al., 2006), characterized by conflicts and conflict resolution difficulties (Berenguer Forner et al., 2017). However, there are researchers who address the need for a thorough investigation of the impact of social understanding on these children’s social skills and friendships (Papadopoulos et al., 2005; Soltani et al., 2013). Findings regarding social understanding in children with dyslexia are few and controversial, with certain studies linking reading difficulties to low performance in ToM tasks (Gabay et al., 2016; Kidd & Castano, 2013), and some stating that children with dyslexia perform equally well in ToM tasks as NT children (Cardillo et al., 2018). Nonetheless, it has been found that children with dyslexia show great difficulties in more advanced ToM tasks, where figures of speech, such as humor, irony, etc., are evaluated (Eyuboglu et al., 2018), and thus, they prefer to befriend children who share similar difficulties and would not judge their problems (Wiener & Schneider, 2002). Finally, adolescents with dyslexia were also reported to face great challenges in their social functioning and interpersonal relationships (Özyurt et al., 2024). Regarding the empathy of children with ADHD, some researchers support the view that these children encounter great difficulties in or a total lack of affective empathy skills (Gvirts & Perlmutter, 2020; Maoz et al., 2019), while others postulate that children with ADHD have well-developed affective empathy for positive emotions (Gambin & Sharp, 2016). Although published work on the empathy of children with dyslexia is limited, it highlights the link among phonological awareness, reading ability, and empathy (Gabay et al., 2016), positively associating their difficulties in affective empathy with the problems that they face in their friendships (Eyuboglu et al., 2018; Gabay et al., 2016). A comparative study among NT children and children with ADHD or dyslexia showed that children with ADHD performed significantly lower across all positive characteristics of friendship quality, affective empathy, second-order ToM and advanced ToM tasks than those with dyslexia, who in turn scored significantly lower than NT children (Kouvava et al., 2025b). In addition, children with ADHD or dyslexia scored significantly higher on the conflict and betrayal scale when compared to NT children, while ToM and affective empathy could predict friendship quality characteristics in all groups of children (Kouvava et al., 2025b). Finally, children with either ADHD or dyslexia have both been found to encounter problems in their ER abilities (Bauminger et al., 2005; Parke et al., 2021), leading to anxious, impulsive or antisocial behaviors, which often result in peer rejection and poor friendship quality (Doikou-Avlidou, 2015; Kouvava et al., 2022; Mikami & Hinshaw, 2003). Children with ADHD are reported to have friendships of low quality and worse ER skills compared to NT children or children with dyslexia, while children with dyslexia performed worse in ER tasks using mainly negative emotion regulation strategies and had friendships of lower qualitative features than NT children (Kouvava et al., 2022). This research highlighted that ER could predict friendship quality in both NT and ND children (Kouvava et al., 2022).

Research on EFs either suggests these skills are quite intact (Wåhlstedt et al., 2009) or that children with ADHD show significant deficits (Adler et al., 2017). Specifically, their working memory has been found to present the gravest difficulties, followed by their cognitive flexibility skills (Crisci et al., 2021). In addition, positive associations have been reported between working memory and self-regulation difficulties (Groves et al., 2022) and between inhibitory control and behavioral problems (Mary et al., 2016) or antisocial behaviors (Silverstein et al., 2020), resulting in children with ADHD often being rejected or isolated by their peers and having few friendships (Spender et al., 2023). In children with dyslexia, inhibitory control and working memory present the greatest deficits (Booth et al., 2010; Hulme & Snowling, 2009), followed by difficulties in their cognitive flexibility (Cutting et al., 2009). Examples of EF-related difficulties in the everyday lives of children with dyslexia could refer to their inability to be flexible when planning and organizing or to remember previous experiences and inhibit their immediate responses during a conflict (Lonergan et al., 2019). Finally, recent findings comparing NT and ND children’s performance in EFs reported that NT children were significantly better in all EFs, while children with dyslexia scored better than children with ADHD only in working memory and cognitive flexibility tasks, skills that were predictive of friendship stability and quality in both NT and ND children (Kouvava et al., 2025a).

1.2. The Interplay Among Friendships and Social Understanding, EFs and ER in NT and ND Children

Although the development of social understanding and EFs requires time, experience, and maturation to be maximized (Kuhnert et al., 2017), with their onset in early childhood and continuing to evolve into adolescence and even into adulthood (Diamond, 2014; Milligan et al., 2007; Wellman, 2014), some researchers note that to date there have not been substantial findings of their interrelationships or their connection with peer relationships (Miller et al., 2018). Recently, some studies have begun to examine the association among some of these factors to NT children’s interpersonal relationships. For example, X. Wang and Feng (2024), in their study on Chinese preschool children, found that both EU and EFs (but not just EU) positively predicted later peer acceptance, after accounting for covariates such as age, gender, family socioeconomic status, and vocabulary. These findings yield the fundamental impact of EFs on peer relationships, as they help children face challenges and adapt to social environments (Caporaso et al., 2019). Positive correlations have also been reported between social understanding and prosocial behaviors, conflict resolution among friends, and peer acceptance in NT children (Fink et al., 2014; Slaughter, 2015). Particularly, it has been indicated that all the EFs are instrumental in successfully completing various ToM tasks, as children need to remember, inhibit, and switch between representations to distinguish between true and false beliefs (Carlson & Moses, 2001), while both skills, and especially ToM, have a significant impact on children’s friendship quality (Miller et al., 2018, 2020). In addition, success in EU and advanced ToM tasks has been positively linked to social adjustment (Hughes & Devine, 2015), sociability, and high-quality friendships (Banerjee et al., 2011; Caputi et al., 2012; Miller et al., 2020; Rose & Asher, 2004). On the other hand, low performance in social understanding has been associated with moral disengagement, deficits in social skills and interpersonal relationships (Crick & Dodge, 1994), as well as aggressive and bullying behaviors in children and adolescents (Gomez-Garibello & Talwar, 2015; Kokkinos et al., 2016). Meanwhile, there are findings demonstrating positive correlations between social understanding and aggression, as children with developed social understanding frequently engage in aggressive or bullying behaviors in their interpersonal relationships (Caravita et al., 2010), such as social exclusion, rumor spreading, gossiping, and manipulating friends (Gomez-Garibello & Talwar, 2015; Renouf et al., 2010).

Recently, Weimer et al. (2021) synthesized the relevant research and presented a dynamic developmental framework showcasing significant interrelations among social behavior (e.g., prosociality, aggression, peer relations), academic achievement (e.g., reading, writing, comprehension, math, science), and emotional self-regulation (e.g., effortful control, executive functions) of NT children, with ToM having a mediating role, and all variables embedded within the influence of sociocultural, linguistic, and contextual factors, and neurodevelopmental cascades. However, all the aforementioned studies examining the interrelationships among social understanding, EFs, and friendships focused solely on NT children, while relevant research evidence on children with ADHD or dyslexia is nonexistent. Thus, it is essential to examine whether similar patterns of associations also apply to ND children and whether their distinct sociocognitive challenges may alter or moderate these relationships.

Based on the above, the study aims to explore the dynamic interplay among social understanding, EFs, and the characteristics of friendships among primary schoolers, contrasting neurotypical (NT) children with those who have been diagnosed with ADHD or dyslexia (ND group). More specifically, the research hypotheses of the study are as follows:

Research Hypothesis 1 (Mediation): Aspects of children’s social understanding will mediate the relationship between Executive Functions and perceived friendship quality. The independent variables include Executive Functions (EFs) comprising Inhibitory Control/Cognitive Flexibility and Working Memory, while the mediator variables involve facets of Social Understanding such as Emotion Understanding (Affective Empathy and Emotion Regulation (ER)) and Belief Understanding (2nd order and advanced ToM). The dependent variables are Perceived Friendship Quality Characteristics, including Positive qualities and Negative qualities.

Research Hypothesis 2 (Group Differences): The above hypothesized mediation pathways will differ between NT children and ND children. The strength, nature, or even existence of the mediation relationships is expected to vary when comparing the NT group to the ND groups (ADHD and dyslexia), reflecting their distinct cognitive and socio-emotional profiles, the goal being to gain an understanding of the ways peer relationships are affected, distinguishing between pathways shared across groups and those unique to specific disorders.

2. Materials and Methods

2.1. Participants

Sixty-four children with ADHD, 64 children with dyslexia, and 64 children with neurotypical development (NT), between 8 and 12 years old (Mage = 9.77 years, sd = 1.21), took part in the study. The parents of all participants signed consent forms, allowing their children’s participation. Children with ADHD or dyslexia had an official diagnosis from national diagnostic centers, stating either ADHD or dyslexia with no comorbid disorders, and thus, no additional diagnostic procedures were performed. NT children’s selection was based on their teachers’ reports regarding their academic achievement and on their school grades (deduced by monthly exams and the grades at the end of the semesters) in the subjects of Language, Mathematics, Physics, and History.

Table 1. Demographics by Group.

		NT Children (n = 64)		Children with Dyslexia (n = 64)		Children with ADHD (n = 64)		Differences
		f	%	f	%	f	%	χ2	p
Gender	Boy	31	48.4	33	51.6	32	50	0.13	0.94
	Girl	33	51.6	31	48.4	32	50
Grade level	3rd	16	25	16	25	16	25	0.21	0.99
	4th	17	26.6	15	23.4	16	25
	5th	16	25	17	26.6	16	25
	6th	15	23.4	16	25	16	25
		M	SD	M	SD	M	SD	F(2,104)	p
Age (in years)		9.67	1.15	9.82	1.15	9.81	1.16	0.29	0.75

Note: NT = Neurotypical children, ADHD = children with Attention Deficit/Hyperactivity Disorder.

presents demographic information for each group of the participating students.

The primary exclusion criteria applied at the participant screening stage included the presence of comorbid conditions other than ADHD/Dyslexia and age outside the target age range.

2.2. Measures

2.2.1. Demographics Information

The first author collected information about the participants’ gender, age, and school grade and administered individually all the following tasks.

2.2.2. Friendship Quality

The Friendship Quality Questionnaire (FQQ; Parker & Asher, 1993), adapted in Greek by Kouvava et al. (2023), examined children’s perceptions about the qualitative features of their best friendships. The questionnaire includes forty items in six subscales, assessing the positive (i.e., validation and caring, help and guidance, companionship and recreation, conflict resolution, and intimate exchange) and negative (i.e., conflict and betrayal) qualitative features of children’s best friendships. Participants respond with reference to a nominated best friend on a 5-point Likert-type scale, ranging from 0 “not at all true” to 4 “really true”. Only item 21 is reverse-scored. Each subscale score was calculated at the mean of children’s ratings. In the present study, Cronbach’s alpha was 0.90 for validation and caring, 0.90 for help and guidance, 0.86 for intimate exchange, 0.84 for conflict and betrayal, 0.75 for companionship and recreation, and 0.73 for conflict resolution.

2.2.3. Social Understanding Tasks

Second-order false belief (ToM) tasks assessed participants’ ability to understand the beliefs of one person about the beliefs of another, using three stories (Perner & Wimmer, 1985). Each story had a script and 4 pictures. Although the children were able to see the stories, they were read to them to avoid possible misunderstandings due to either their young age or their reading difficulties. Participants had to answer the following types of questions: (a) a reality question and a memory question, to confirm that they have understood the story; (b) an experimental question, scored with 0 or 1 points, based on its correctness to identify the second-order belief; and (c) a documentation question, which examined the understanding of the second order of belief and was scored with 0 points when the documentation was incorrect, with 1 point when the documentation was correct using natural explanations, and with 2 points when the documentation was correct using mental situations. The maximum score for each story was 3.

Advanced ToM tasks, such as the Strange Stories (Happé, 1994), were also administered to all participants to evaluate their advanced mentalizing ability. Strange Stories were presented in short vignettes featuring non-literal statements (e.g., pretense, joke, lie, white lie, misunderstanding, irony, forgetting, double bluff, contradicting emotions, persuasion, and figure of speech). All stories were read to the students, who then had to answer two questions: a comprehension question, with the correct answer being scored with 1 point, and a documentation question, where answers were scored with either 0 if they were incorrect or with 1 (correct—physical documentation) or 2 (correct—mental state documentation) (Freed et al., 2015).

The Bryant’s Index of Empathy for Children and Adolescents (Bryant, 1982), validated (Mitsopoulou & Giovazolias, 2013) and adapted (Vassilopoulos et al., 2021) in Greek, was used to examine participants’ affective empathy. The Index has 22 items, with “Yes” or “No” as possible answers. All items were read to the children. Positive answers were scored with 1 point and negative answers with 0, with the total score ranging from 0 to 22. Sentences 2, 3, 9, 10, 15, 16, 17, 18, 20, 21, and 22 were reverse-scored. Higher performance implies greater affective empathy. In the present study, although Cronbach’s alpha for Bryant’s Index of Empathy was relatively low (α = 0.64), it aligned with other studies stating that the scale’s reliability estimates ranged from 0.54 for the first graders to 0.68 for the fourth graders and 0.79 for the seventh graders (Huang & Tran-Chi, 2020).

The Emotion Regulation Questionnaire for Children and Adolescents (ERQ-CA; Gullone & Taffe, 2012) adapted in Greek by Kokkinos and Voulgaridou (2017), is a 10-item tool assessing students’ strategies to regulate their emotions. Cognitive reappraisal (CR) (emotionally charged situation and the deliberate modification of the automatic emotional response) was evaluated with 6 items and expressive suppression (ES) (inhibition of an ongoing, positive or negative emotion) (O. P. John & Gross, 2004) with 4 items, using a 5-point Likert-type response scale (1 “strongly disagree” and 5 “strongly agree”). Higher mean scores correspond to the preference of the corresponding ER strategy. In the present study, Cronbach’s α was 0.93 for CR and 0.81 for ES.

2.2.4. Executive Functions Tasks

The Stroop Color and Word Test (SCWT; Stroop, 1935) was used to examine children’s inhibitory control. Three different tables with words referring to colors were presented to the participants, who were asked to read fast the printed words in black ink (first table) or colored ink (second table). In the second table, the meaning of the words and the color in which they are printed were identical. However, in the third table (color-word condition), participants had to name the color of the ink that the color words were printed in instead of reading the words. In this table, there were inconsistencies between the color of the ink and the meaning of the written words (e.g., the word “black” was printed in yellow ink). At the beginning, children were asked whether they could recognize and name the 12 colors used. Performance was scored based on the time (in seconds) required by the students to complete the task in the color-word condition (van der Elst et al., 2006).

Participants’ cognitive flexibility was assessed by using the Trail Making Test (TMT; Armitage, 1946), which has also been adapted in Greek (Vlahou & Kosmidis, 2002). TMT has been administered to both adults and adolescents with and without neurodevelopmental disorders (Ladopoulos, 2019; Vlahou & Kosmidis, 2002). In the second part (part B) of the test, where cognitive flexibility is assessed, participants are asked to connect 25 encircled numbers and letters, arranged in a semi-random fixed order, in alternating numerical and alphabetical order (Vlahou & Kosmidis, 2002). Scoring considers the total time (in seconds) required to accurately complete the connections during the task (Vlahou & Kosmidis, 2002).

Working memory was assessed using subscale 12 of the Greek version (Georgas et al., 1997) of the Wechsler Intelligence Scale for Children (WISC-III; Wechsler, 1991). Children had to repeat a series of digits progressively increasing in length in reverse order. Seven pairs of digit sequences, each containing the same number of digits, were read aloud to each child at a rate of one digit per second. Two points were awarded if both attempts for an item were correct, 1 point if only one attempt was correct, and 0 points if both attempts were incorrect. The assessment ended when no correct repetition was observed in both attempts for a given sequence length. Total scores reflected the sum of correct responses across trials.

2.3. Procedure

After permission was granted by the Institute for Educational Policy of the Greek Ministry of Education, Religious Affairs, and Sports for 84 randomly selected inclusive mainstream primary schools from various parts of Attica, Greece, to participate in this study, the head teachers and teachers of these schools were approached and informed about the purpose of the study. Fifty-eight schools agreed to participate in the research, enabling the researchers to send a letter to the parents of children with and without ADHD or dyslexia explaining the aims of the study and asking them to sign a consent form permitting their children’s participation in the study and granting access to their children’s school clinical records to verify ADHD or dyslexia diagnoses. Children’s participation was voluntary. All tasks were administered individually to each child by the first author and in a specific order (i.e., friendship, ToM, and EF tasks). Participants were given detailed instructions and were informed about the voluntary and confidential nature of the study. In addition, all items in every task were read aloud to support children with difficulties in reading. The evaluation procedure, which lasted about one hour, took place in a quiet area in children’s schools, where their anonymity was guaranteed. Children exhibiting signs of fatigue or discomfort were allowed to stop briefly, and, if necessary, they could continue the assessment on another day. The present study had a cross-sectional and comparative design, as the data were collected in a single-point measurement, and comparisons among groups were performed.

The data of this study have also been used in separate publications (Kouvava et al., 2025a, 2025b, 2022). However, respecting the ethical research guidelines, each publication focused on distinct and non-overlapping research objectives, and the multiple use of data, which were collected according to rigorous ethical protocols ensuring the confidentiality and anonymity of participants’ responses, was transparently disclosed. The decision to use the data in multiple studies was made with careful consideration of ethical standards to maintain the integrity and ethical conduct of the research.

2.4. Data Analyses

Descriptive statistics were calculated for all scales used in the study to examine their distributions and to assess the suitability of the data for further analyses. Assumption checks were conducted prior to modeling. Skewness and kurtosis values for all observed variables fell within the recommended ranges (|skew| < 2, |kurtosis| < 7; West et al., 1995), and multicollinearity diagnostics indicated no problematic intercorrelations (all r < 0.85) or variance inflation factors (all VIF < 5). Given these results, the data were deemed suitable for SEM. Gender differences in the variables of interest were explored using independent t-tests within each group, with Bonferroni correction applied to adjust for multiple comparisons (adjusted α = 0.0042). After correction, no gender differences reached statistical significance in any group. Consequently, gender was not included as a covariate in the SEM analyses.

Structural equation modeling (SEM) with latent variables was employed to test the hypothesized mediation models. Specifically, the models examined whether emotions and beliefs mediated the relationships between (1) inhibitory control/cognitive flexibility and the dimensions of friendship (positive quality and conflict/betrayal), and (2) working memory and the dimensions of friendship. The analyses were conducted separately for each of the three participant groups: NT children, children with dyslexia, and children with ADHD, resulting in a total of six models. While working memory, inhibitory control, and cognitive flexibility are often discussed together under the umbrella of EFs, in the present study, inhibitory control and cognitive flexibility were combined into one latent construct in Model 1 because both were assessed through response-time tasks tapping overlapping cognitive control processes and were highly correlated in our data. This combination was chosen over parceling or alternative model simplification strategies because the small sample size within each group limited the feasibility of more complex specifications, and parceling could have obscured construct-level distinctions that were central to our hypotheses. Nonetheless, their conceptual distinction is acknowledged, and the combination may have contributed to poorer fit indices observed in some groups. (Sambol et al., 2025).

Following the two-step approach recommended by Anderson and Gerbing (1988), the measurement models were first evaluated to confirm the fit and reliability of the latent variables and their indicators, including EFs (inhibitory control/cognitive flexibility, working memory), social understanding (emotions: affective empathy and ER; beliefs: 2nd-order and advanced ToM), and friendship quality dimensions (positive and negative). Affective empathy and ER were modeled as indicators of a single “emotions” latent factor because both reflect the affective and regulatory components of social–emotional competence and showed strong shared variance in preliminary analyses. This aggregation is consistent with theoretical frameworks that view empathy and regulation as interdependent in shaping prosocial and relational outcomes in childhood, although they can also be studied as distinct constructs. Once the measurement models showed acceptable model fit, the structural models were tested using robust maximum likelihood estimation (MLR) in Mplus v8.6 (Muthén & Muthén, 2017). To examine whether structural paths differed significantly between groups, multi-group measurement invariance testing was conducted for each model following the sequence of configural, metric, and scalar invariance (Vandenberg & Lance, 2000). Invariance was evaluated using changes in CFI (ΔCFI ≤ 0.01) and RMSEA (ΔRMSEA ≤ 0.015) (Chen, 2007). Where metric invariance was supported, Wald χ² tests were used to compare specific path coefficients across groups.

Model fit was assessed using multiple indices recommended by Hu and Bentler (1999), including standardized root-mean-square residual (SRMR) less than 0.08, root-mean-square error of approximation (RMSEA) between 0.05 and 0.08, Tucker-Lewis Index (TLI) above 0.90, and comparative fit index (CFI) above 0.90 (van de Schoot et al., 2012). To evaluate mediation effects, bias-corrected bootstrapped 95% confidence intervals (B = 10,000) were used, as this method has been shown to account for the non-normality of indirect effects (MacKinnon et al., 2013). Mediation was considered significant if the confidence intervals did not include zero. Effect sizes for all structural paths are presented as standardized β coefficients and interpreted using conventional benchmarks (Cohen, 1988), with β ≈ 0.10 considered small, β ≈ 0.30 medium, and β ≥ 0.50 large, to inform both statistical and practical significance.

3. Results

3.1. Preliminary Analyses

Table 2. Descriptive Statistics and Correlations for the ADHD Group.

Variable	M	SD	1	2	3	4	5
1. Beliefs (2nd order & advanced ToM)	18.25	5.32
2. Emotions (affective empathy & ER)	55.39	5.97	0.03
3. Inhibitory control/cognitive flexibility (response time) (EFs)	225.50	52.28	0.08	−0.17
4. Working Memory (EFs)	2.86	1.10	0.45 **	0.00	0.18
5. Pos. Friendship Quality	2.84	0.84	0.36 **	0.15	−0.07	0.30 *
6. Conflict/Betrayal	2.51	0.71	−0.31 *	0.09	−0.22	−0.20	−0.09

Note. M = mean; SD = standard deviation. Correlations marked with p < 0.01 (**) are significant at the 0.01 level (two-tailed), and correlations marked with p < 0.05 (*) are significant at the 0.05 level (two-tailed).

,

Table 3. Descriptive Statistics and Correlations for the Dyslexia Group.

Variable	M	SD	1	2	3	4	5
1. Beliefs (2nd order & advanced ToM)	21.62	6.46
2. Emotions (affective empathy & ER)	56.19	7.34	0.12
3. Inhibitory control/cognitive flexibility (response time) (EFs)	220.81	68.26	−0.56 **	−0.18
4. Working Memory (EFs)	3.50	1.14	0.33 **	−0.03	−0.22
5. Pos. Friendship Quality	3.26	0.84	0.40 **	0.21	−0.30 *	−0.05
6. Conflict/Betrayal	2.37	0.67	−0.23	−0.01	0.32 **	0.04	−0.24

Note. M = mean; SD = standard deviation. Correlations marked with p < 0.01 (**) are significant at the 0.01 level (two-tailed), and correlations marked with p < 0.05 (*) are significant at the 0.05 level (two-tailed).

and

Table 4. Descriptive Statistics and Correlations for the NT Group.

Variable	M	SD	1	2	3	4	5
1. Beliefs (2nd order & advanced ToM)	33.39	6.38
2. Emotions (affective empathy & ER)	62.08	5.57	0.49 **
3. Inhibitory control/cognitive flexibility (response time) (EFs)	82.48	40.02	−0.56 **	−0.44 **
4. Working Memory (EFs)	6.39	1.64	0.52 **	0.32 *	−0.52 **
5. Pos. Friendship Quality	4.36	0.55	0.54 **	0.35 **	−0.72 **	0.44 **
6. Conflict/Betrayal	1.82	0.43	−0.25 *	−0.10	0.50 **	−0.33 **	−0.59 **

Note. M = mean; SD = standard deviation. Correlations marked with p < 0.01 (**) are significant at the 0.01 level (two-tailed), and correlations marked with p < 0.05 (*) are significant at the 0.05 level (two-tailed).

present means, standard deviations, and correlations for all study variables, separately for each group of children (ADHD, dyslexia and NT). Skewness and kurtosis values for all observed variables were within acceptable ranges (|skew| < 2, |kurtosis| < 7; West et al., 1995), and multicollinearity diagnostics indicated no problematic intercorrelations (all r < 0.85) or variance inflation factors (all VIF < 5). Gender was not included as a covariate in the subsequent SEM analyses, as no gender differences reached statistical significance in any group.

For the ADHD group, beliefs were positively associated with working memory scores (r = 0.45, p < 0.01) and positive friendship quality (r = 0.36, p < 0.01). Additionally, working memory showed significant positive associations with positive friendship quality (r = 0.30, p < 0.05). A negative correlation emerged between beliefs and conflict/betrayal (r = −0.31, p < 0.05) (Table 2).

Within the dyslexia group, beliefs were positively correlated with working memory (r = 0.33, p < 0.01) and positive friendship quality (r = 0.40, p < 0.01). Inhibitory control/cognitive flexibility scores were negatively correlated with beliefs (r = −0.56, p < 0.01) (Table 3).

For NT children, beliefs were positively correlated with working memory (r = 0.52, p < 0.01) and positive friendship quality (r = 0.54, p < 0.01). Working memory was also positively correlated with positive friendship quality (r = 0.44, p < 0.01). Inhibitory control/cognitive flexibility were negatively associated with beliefs (r = −0.56, p < 0.01) and positive friendship quality (r = −0.72, p < 0.01) (Table 4).

3.2. Measurement Models

The study tested two statistical models to see if beliefs and emotions explain how executive functions relate to friendship quality (positive quality and conflict) in ADHD, Dyslexia and NT children. Both models fit the data well for the ADHD and Dyslexia groups. However, neither model fit the data well for the NT group, suggesting that these specific models—one based on Inhibitory Control/Cognitive Flexibility (Model 1) and one based on Working Memory (Model 2)—may not accurately capture the complex relationships in NT children. This poor fit might be due to combining related but distinct executive function skills into single factors (see Appendix A for the models’ statistics). Multi-group measurement invariance testing was conducted for both models across the ADHD, dyslexia, and NT groups, following the sequence of configural, metric, and scalar invariance (Vandenberg & Lance, 2000). Invariance was evaluated using changes in CFI (ΔCFI ≤ 0.01) and RMSEA (ΔRMSEA ≤ 0.015) as recommended by Chen (2007). Where at least metric invariance was supported, Wald χ² tests were used to compare specific structural path coefficients across groups.

3.3. Structural Models

Model 1: For the ADHD group, inhibitory control/cognitive flexibility had no significant direct effect on either friendship dimension (friendship positive quality: β = −0.071, p = 0.545; conflict/betrayal: β = −0.182, p = 0.117), but its effect was indirectly significant via beliefs, which themselves had a significant positive effect on friendship quality (β = 0.027, p = 0.002) and a negative effect on conflict/betrayal (β = −0.055, p = 0.007). Conversely, emotions had no significant direct or indirect effects on either friendship outcome, a pattern observed across all models and groups, potentially because combining empathy and regulation into a single “emotions” factor diluted its specific effects. The total effect on friendship positive quality was non-significant (β = −0.071, p = 0.545), while the effect on conflict/betrayal only approached significance (β = −0.237, p = 0.062) (Figure 1).

For the dyslexia group, inhibitory control/cognitive flexibility had a large, negative effect on beliefs (β = −0.555, p < 0.001), no direct effect on emotions (β = −0.181, p = 0.133), a significant positive direct effect on friendship conflict/betrayal (β = 0.290, p = 0.037) but not on positive quality (β = −0.084, p = 0.539). Beliefs significantly and positively predicted friendship positive quality (β = 0.338, p = 0.009) but not conflict/betrayal (β = −0.076, p = 0.589). Notably, emotions did not significantly predict either friendship outcome. Significant indirect effects of inhibitory control/cognitive flexibility on friendship positive quality (β = −0.027, p = 0.023) were observed via beliefs, while total effects on both friendship dimensions were statistically significant but negligible in magnitude (both β < 0.01) (Figure 2).

For the (NT) group the analysis revealed that inhibitory control/cognitive flexibility had a large, negative effect on both beliefs (β = −0.560, p < 0.001) and emotions (β = −0.439, p < 0.001), a large, negative direct effect on friendship positive quality (β = −0.606, p < 0.001) and a large, positive direct effect on conflict/betrayal (β = 0.556, p < 0.001). Beliefs had a marginally significant positive effect on friendship quality (β = 0.206, p = 0.054) but no effect on conflict/betrayal (β = −0.005, p = 0.970). Consistent with other groups, emotions did not significantly predict either friendship outcome. The indirect effects through the mediators were largely non-significant, and the total effects of inhibitory control/cognitive flexibility on both friendship dimensions were statistically significant (friendship positive quality: β = −0.010, p < 0.001; conflict/betrayal: β = 0.005, p < 0.001) but trivial in magnitude (Figure 3).

The core conclusion from these results of Model 1 is that, across all groups (ADHD, Dyslexia, and NT), the way a child’s Inhibitory Control/Cognitive Flexibility relates to their friendships is highly complex and primarily indirect. In all three groups, beliefs were a significant factor, consistently and positively linked to higher positive friendship quality. Additionally, for all groups, the combined emotions factor (empathy and regulation) was not a significant predictor of either positive friendship quality or conflict/betrayal. For the ADHD group, the effect of inhibitory control was entirely indirect, working through beliefs to benefit friendships. For the Dyslexia and NT groups, better inhibitory control/flexibility was surprisingly linked to worse beliefs and, for the NT group, worse direct friendship outcomes. However, the overall total effects of inhibitory control on friendship quality were generally very small or non-significant across the groups. In simplest terms, beliefs, not emotions, are the most reliable link to positive friendships, and a child’s ability to control impulses (inhibitory control) has an indirect, varied, and often minor overall impact on their friendships.

Model 2 results for the ADHD group showed that working memory had a significant positive effect on beliefs (β = 0.447, p < 0.001) but no effect on emotions (β = 0.001, p = 0.992). Beliefs significantly influenced friendship, showing a positive effect on friendship positive quality (β = 0.273, p = 0.028) and a negative effect on conflict/betrayal (β = −0.285, p = 0.025). Consistent with previous findings, emotions did not significantly affect either friendship outcome. Working memory had a significant indirect effect on positive friendship quality (β = 0.094, p = 0.059) through beliefs and a significant total effect on positive friendship quality (β = 0.227, p = 0.013), but no significant total effect on conflict/betrayal (β = −0.126, p = 0.110) (Figure 4).

For the dyslexia group, working memory played a significant role by positively influencing beliefs (β = 0.334, p = 0.003). However, working memory had no significant effect on emotions (β = −0.027, p = 0.832). Beliefs were a significant predictor of friendship quality. They had a large, positive effect on friendship quality (β = 0.453, p < 0.001) and a medium, negative effect on conflict/betrayal (β = −0.280, p = 0.024). Emotions had no significant effect on either positive quality (β = 0.156, p = 0.164) or conflict/betrayal (β = 0.023, p = 0.851). Working memory’s influence on positive friendship quality was significantly mediated through beliefs (β = 0.110, p = 0.023). The indirect effect of working memory on conflict/betrayal through beliefs was not significant (β = −0.055, p = 0.085). The total effect of working memory (direct plus indirect) on both positive quality (β = −0.035, p = 0.702) and conflict/betrayal (β = 0.025, p = 0.731) was non-significant (Figure 5).

For the NT group, working memory was a strong positive factor, significantly influencing both beliefs and emotions. Working memory’s positive influence on friendship quality was primarily channeled through beliefs, which, in turn, predicted better, more positive friendships (β = 0.382, p = 0.002). Neither beliefs nor emotions significantly affected friendship conflict. The total effect of working memory was significant, as it led to better overall friendship quality (β = 0.150, p < 0.001) and also resulted in slightly less conflict/betrayal (β = −0.086, p = 0.005) (Figure 6). In short, better working memory helps NT individuals form positive beliefs, leading to better friendships.

The overall conclusion from Model 2 comparing the ADHD, Dyslexia, and NT groups is that beliefs are the consistent key factor linking cognitive ability to better friendships. In all three groups, stronger working memory significantly led to better understanding of false beliefs. False belief understanding, in turn, was the primary driver of higher positive friendship quality and, for the ADHD and Dyslexia groups, lower conflict/betrayal. The role of emotions was consistently insignificant in predicting friendship quality across all groups. While working memory had a significant total positive impact on friendship quality for the ADHD and NT groups, its influence was primarily indirect (through beliefs). For the Dyslexia group, the total effect of working memory on friendship quality was not significant, meaning its positive effect was completely reliant on the positive shift it created in beliefs. In essence, for children in all three groups, better working memory helps them develop better belief understanding, and it is beliefs—not their emotions—that lead to better, healthier friendships.

4. Discussion

The study aimed to elucidate the complex interplay between social understanding outcomes such as belief and emotion comprehension (through ToM tasks, affective empathy and ER), EFs including inhibitory control, cognitive flexibility and working memory, and friendship quality dimensions (positive quality and conflict/betrayal) across different neurodevelopmental groups (ADHD, dyslexia, and NT). These findings can be interpreted through the lens of the SIP model, which emphasizes how children’s interpretations of social cues and expectations shape their responses and the quality of their peer interactions (Crick & Dodge, 1994). The results revealed intriguing group-specific patterns and some consistent findings.

4.1. Inhibitory Control and Cognitive Flexibility (Model 1)

The influence of Executive Functions (EFs) on friendship quality is significantly mediated by beliefs, with a consistently minor role for emotions, but the pathways differ significantly across groups. Surprisingly, in the ADHD group, inhibitory control and cognitive flexibility showed no significant direct effects on either positive friendship quality or conflict/betrayal. Their influence was almost entirely indirect, operating through beliefs (Groves et al., 2022; X. Wang & Feng, 2024). Beliefs were a strong predictor of both positive friendship quality (positive relationship) and conflict/betrayal (negative relationship), highlighting their crucial role in shaping friendships (Normand et al., 2011). This pattern suggests that EF challenges subtly influence social beliefs, which then directly impact friendship outcomes (Silverstein et al., 2020; Spender et al., 2023). The non-significant total effect of EFs reinforces that their impact is primarily mediated rather than direct. Emotions (as measured) were non-significant as both predictors and mediators, contrasting with common assumptions about emotional dysregulation in ADHD impacting social relationships (Kouvava et al., 2022; Mikami & Hinshaw, 2003). For the dyslexia group, the relationship was more direct and pervasive. EF difficulties showed significant negative effects on beliefs and significant positive effects on conflict/betrayal (Lonergan et al., 2019). This suggests that EF challenges may directly contribute to increased conflict. Beliefs successfully mediated the negative total effect of EFs on positive friendship quality (Eyuboglu et al., 2018; Doikou-Avlidou, 2015) but did not significantly predict conflict/betrayal or mediate the EF–conflict relationship. This indicates that beliefs are instrumental in fostering positive friendship experiences, while direct EF difficulties are more strongly tied to conflict. The significant total effects of EFs on both positive quality (negative relationship) and conflict/betrayal (positive relationship) point to a more pervasive influence of broad EF abilities on friendship outcomes compared to the ADHD group. Consistent with the ADHD group, emotions were also non-significant as predictors or mediators. The proposed structural model proved a poor fit for the NT group, severely limiting the reliability of specific path interpretations. This suggests that a different theoretical model or different mediating variables are necessary to understand the relationship between EFs and friendship in NT individuals. Cautious interpretation of the strong total effects suggests a strong overall influence of EFs on friendship, though the specific mechanism remains uncaptured by this model. As with the ND groups, emotions were consistently non-significant.

The results underscore the heterogeneity of mediation pathways across neurodevelopmental groups, particularly regarding the role of beliefs in predicting conflict/betrayal. However, beliefs consistently emerged as a key factor—especially for positive friendship quality—across ADHD and dyslexia. The consistent non-significance of emotions across all groups is a particularly striking finding that warrants further investigation, suggesting either limitations in the measures used or that emotional processes influence friendship more indirectly. This pattern highlights the potential value of interventions that target social cognition and cognitive reappraisal (e.g., challenging negative friendship schemas) for children with ADHD and dyslexia.

4.2. Working Memory (Model 2)

The mediation model examining working memory and friendship quality across groups reveals that beliefs are the primary cognitive link to positive friendship outcomes, while emotions consistently play a non-significant role. In children with ADHD, working memory significantly and positively influenced beliefs (Carlson & Moses, 2001; Silverstein et al., 2020). Beliefs were strong, significant predictors of both higher positive friendship quality and lower conflict/betrayal. Crucially, beliefs significantly mediated the effect of working memory on positive quality (Miller et al., 2018), implying that working memory difficulties impact friendships by shaping these social interpretations. Working memory’s significant total effect on positive quality (Spender et al., 2023) further highlights its importance, largely operating through beliefs. Similar to the previous model, emotions were non-significant as predictors or mediators. For the dyslexia group, working memory significantly influenced beliefs, but not emotions. Beliefs had a positive effect on positive quality and a negative effect on conflict/betrayal (Eyuboglu et al., 2018). Most notably, beliefs significantly mediated the relationship between working memory and positive quality (Cardillo et al., 2018), reinforcing their centrality for positive friendship in this group. Despite the significant indirect effect, the total effect of working memory on both friendship dimensions was non-significant, suggesting working memory’s influence is entirely mediated. Emotions were again non-significant. The NT group uniquely showed that working memory significantly impacted both beliefs and emotions. However, only beliefs significantly predicted positive friendship quality (Miller et al., 2018; Weimer et al., 2021) and served as a significant mediator for the working memory-positive quality link. Despite working memory’s influence on them, emotions did not significantly predict friendship dimensions or mediate working memory’s effects. The significant total effects of working memory on both positive quality (positive) and conflict/betrayal (negative) indicate its substantial overall impact on NT friendships (Holmes et al., 2016; Woltering et al., 2016).

The most robust finding is the significant mediating role of beliefs in the relationship between working memory and positive friendship quality across all groups, highlighting that how children interpret social interactions, influenced by their working memory capacity, is critical. Conversely, the consistent non-significance of emotions across all groups and models warrants critical reflection, suggesting that the measured emotional constructs may be too broad or that emotions have more distal effects not captured by this specific model. These results strongly suggest that cognitive–behavioral interventions targeting social beliefs and interpretations could be highly effective in enhancing friendship quality, especially for children with ADHD and dyslexia (Caporaso et al., 2019). Furthermore, training working memory may offer indirect benefits by improving underlying belief systems.

The present study suffers from certain limitations. Initially, it is cross-sectional. A longitudinal design might be more adept at understanding the developmental trajectories of the relationships among variables and better establish causality. For example, do EF deficits lead to certain beliefs, or do negative friendship experiences shape beliefs and then impact EFs? Certain measurement issues should also be addressed, such as the risk of Type I errors from multiple testing or the consolidation of EFs into two composite variables based on the scoring metrics and the affective empathy and ER into a single variable (“emotions”) based on the empirical overlap between these constructs. The combination of affective empathy and ER, and of inhibitory control and cognitive flexibility into single latent variables, may have obscured important nuances contributing to poor model fit into a single latent variable, as these constructs, while highly correlated in our data, are conceptually distinct. Thus, in future research, more granular, context-specific assessments and a nuanced analysis of the EFs and “Emotions” are required to draw safer conclusions. Incorporating qualitative data (e.g., interviews with participants about their friendship experiences, beliefs, and emotional responses) could provide rich, nuanced insights that quantitative models alone cannot capture, especially regarding the non-significant role of emotions. Friendships are dyadic. Future research could explore peer perceptions and include data from friends to get a more complete picture of the friendship dynamics. Moreover, the models do not account for environmental, sociocultural, or contextual factors (e.g., family support, school environment, peer group characteristics) that undoubtedly influence friendship quality. Another limitation concerns the absence of an a priori power analysis. Given the modest sample size and the lack of formal estimation of the minimum required participants, the statistical power of some analyses cannot be guaranteed. This limitation should be considered when interpreting and generalizing the findings, and future studies should include an a priori power analysis to ensure adequate statistical power. Finally, certain demographic factors such as the socioeconomic status of the family, parental education, and language background were not examined. These variables can influence both social understanding and executive function and should at least be acknowledged, if not controlled for, in future studies.

Overall, this study provides valuable insights into the pathways through which sociocognitive factors influence friendships. Beliefs emerged as a consistent mediator, particularly in relation to positive friendship quality. The persistent non-significance of emotions and the poor model fit for the NT group in Model 1 present critical areas for future investigation and model refinement. Understanding these group-specific nuances is crucial for developing targeted and effective interventions to support positive social development.

5. Conclusions

Children with neurodevelopmental conditions like ADHD and dyslexia often face significant challenges in forming and maintaining high-quality friendships in mainstream schools, a difficulty historically linked to deficits in both social understanding (e.g., ToM and perspective-taking) and Executive Functions (EFs) (e.g., impulse control, flexibility). However, this study indicated that the relationship between EFs and friendship quality is largely indirect, operating primarily through cognitive social skills, specifically belief understanding (a key component of ToM), rather than through emotions like empathy or regulation. Across all groups—ADHD, dyslexia, and NT—belief understanding and advanced ToM emerged as the most significant mediators, leading to higher positive friendship quality and, in ND groups, less conflict and betrayal, while emotions proved insignificant. This finding redirects the focus for intervention from broad emotional regulation to cognitively based social skills.

The study reveals that the impact of specific EFs on friendship quality is complex and varies by group, underscoring the inadequacy of a “one-size-fits-all” approach to EF training. Working memory acts as a strong, consistent positive force across all groups, enhancing beliefs and leading to better overall friendship quality for the ADHD and NT groups, and indirectly benefiting the dyslexia cohort. Conversely, the effect of inhibitory control is more nuanced: its beneficial impact in the ADHD group is entirely indirect (via beliefs), but in the dyslexia and NT groups, better inhibitory control was unexpectedly linked to poorer beliefs and a slight increase in conflict, suggesting potentially different mechanisms of social processing.

Given these findings, school-based interventions and programs should prioritize challenging negative friendship schemas and explicitly fostering perspective-taking, social competence and peer intentions, as this is the primary mechanism through which EFs translate into successful friendships. To achieve this, a multi-faceted approach embedded within a school culture that values neurodiversity is paramount. This includes implementing explicit social skills instruction, utilizing strategies like social stories, role-playing, and discussions that promote perspective-taking and understanding of social cues. Simultaneously, practices targeting EF support should provide explicit instruction, scaffolding, and environmental modifications to address difficulties in areas like working memory, planning, and organization. Program design must be tailored: for children with ADHD, interventions should emphasize bolstering working memory to strengthen beliefs while recognizing that improving inhibitory control may require highly specific social skill training. Finally, fostering positive peer interactions through structured social skills groups, collaborative learning activities, and peer mentoring programs is essential to cultivate inclusive environments where qualitative friendships can flourish, enhancing the socio-emotional well-being of all students.