Factors Relating to the Presence and Modifiability of Self-Perceptual Bias among Children with ADHD

Abstract

Past research raises concerns about whether the presence of self-perceptual biases among children with attention-deficit/hyperactivity disorder (ADHD) interferes with accurate assessment and/or diminishes treatment response. Yet, it remains unclear whether self-perceptual bias is a construct that can be modified. The current study examines individual differences in how children with ADHD (n = 178) display and modify their self-perceptions of competence in the presence of an external motivator for self-perceptual accuracy. Participants were grouped based on the presence and modifiability of their self-perceptual biases across three experimental conditions. Results demonstrate that the presence and modifiability of participants’ self-perceptual biases across conditions was associated with adjustment (i.e., externalizing and internalizing problems) and cognitive functioning. Findings suggest multiple factors may be associated with self-perceptual bias (e.g., self-protection and cognitive impairment), and that these factors may differ across children. Implications for intervention, including whether assessment and treatment can be improved, are discussed.

A substantial body of literature highlights that, on average, children with attention-deficit/hyperactivity disorder (ADHD) overestimate their competence in comparison to their typically developing peers.^1–6 These positive self-perceptual biases, often referred to as “Positive Bias” (PB), occur in a subset of children with ADHD and typically are assessed by calculating the discrepancy between children’s self-reports of competence and an external measure of competence, such as a parent or teacher report. Although this perspective that positive self-perceptual biases characterize many children with ADHD is not universally accepted,^7–9 research highlighting the potential functional consequences among children with ADHD who maintain such biases supports the need for further study. For example, empirical evidence suggests that positive self-perceptual biases among children with ADHD may interfere with accurate assessment of comorbid internalizing problems¹⁰ and/or diminish treatment responsiveness.¹¹ Further, positive self-perceptual biases appear to be associated with indices of maladjustment in children with ADHD, which raises additional concern regarding its prevalence among this clinical population.^3,12,13 Despite research documenting the existence of self-perceptual biases among children with ADHD, little is known about the prevalence of different types of biases (i.e., negative, positive), the modifiability of these biases, and the factors related to whether or not children can alter their biased self-perceptions. To address this gap, this study examines individual differences in the extent to which children with ADHD display and modify biased self-perceptions when presented with external motivation for self-perceptual accuracy.

It is important to note that there is current debate in the field regarding the most appropriate method to measure self-perceptual bias.⁶ Although discrepancy scores have been widely used in past research, this approach has faced criticism for two primary reasons. First, discrepancy scores often have lower overall reliability than the reliability of each variable that comprises the discrepancy score.¹⁴ Second, discrepancy scores tend to be highly correlated with each of their components, and, as such, a significant relation between a discrepancy score and a given outcome may reflect a relation between only one of the components and the outcome.^6,15 Although these criticisms do have merit, discrepancy scores were used in the current study for several reasons. First, alternative approaches, such as using regression analysis to obtain standardized residual difference scores, have their own statistical limitations (e.g., low reliability).^6,16 Importantly, standardized scores, whether they are standardized residuals or discrepancy scores, result in metrics that are difficult to interpret in a meaningful way,¹⁶ especially in a clinical setting and in the absence of normative data. In addition, the standardized residual score approach assumes 50% of individuals will fall above the regression line (overestimators) and 50% will fall below the line (underestimators).⁶ This assumption is not necessarily valid for a clinical sample of children with ADHD for whom an unequal distribution of over- and underestimators might be expected.

Next, and most importantly, the current study focuses primarily on whether and how biased self-perceptions change across conditions. Modifiability of self-perceptual bias is the primary focus, operationalized in nominal categories based on the extent to which this bias varies across three experimental conditions. Using this approach, the second criticism of a result being driven by the relation of only one component of the discrepancy score becomes less applicable, as the independent variable is no longer a difference score, but rather a categorical variable. Finally, there remains ongoing debate regarding the methods used to categorize children as displaying positively biased self-perceptions.¹⁷ Although several studies utilize a +1SD or +1 point cutoff score to identify children as having a PB,^12,18,19 some critics have argued there is no empirical basis for this categorization method.¹⁷ Thus, in order to create groups without reliance on a cutoff score, the present study utilizes Latent Profile Analysis (LPA)²⁰ to derive statistically based groups based on the presence and modifiability of self-perceptual bias, measured continuously, across a series of experimental conditions. After groups were created, we reference a +1 SD cutoff score to help describe the groups that emerge and place them in the context of the broader self-perceptual bias literature. However, reference to the cutoff score in the present study was solely for the purpose of profile description and not for the purpose of group classification.

Positively Biased Self-Perceptions and Maladjustment

Previous research links positively biased self-perceptions with externalizing behaviors, such as oppositionality³ and aggression^5,13, and deficits in social functioning, such as low peer preference and overall psychosocial adjustment.^3,5,12 Though few studies shed light on the directionality of effects due to their cross-sectional design, some longitudinal investigations suggest positive self-perceptual bias is a unique predictor of later maladjustment. For example, recent findings from a short-term longitudinal study (2 weeks) indicate positive self-perceptual bias as a predictor of peer preference and oppositionality, specifically among children with ADHD as compared to typically developing peers.³ Yet, another longitudinal investigation over a longer span of time (6 years) provides evidence for a bidirectional association between self-perceptual bias and externalizing behavior among children with ADHD, with increases in positive behavioral self-perceptual bias associated with greater aggression and vice versa.¹³ Interestingly, cross-sectional research has documented that biased self-perceptions are greatest in the domains in which children with ADHD are the least competent.^1,2 Additionally, research examining self-perceptions of competence and internalizing behaviors suggests that higher levels of depression are related to underestimations of competence among children with ADHD.^5,21,22 Together, these findings support one of the most widely discussed explanations of the self-perceptual bias phenomenon: the self-protective hypothesis. This hypothesis asserts that when children with ADHD are faced with their own incompetence and feelings of inadequacy, they exaggerate reports of self-competence to help preserve their self-esteem.²³ Thus, according to this hypothesis, children with the greatest deficits in functioning would be expected to employ self-protective tendencies to extents greater than children with higher levels of functioning.

Positively Biased Self-Perceptions and Treatment

Not only is self-perceptual bias related to maladjustment among individuals with ADHD, but evidence suggests that lack of insight into impairment, a defining feature of positive self-perceptual bias, interferes with treatment responsiveness.^6,11,24 For example, in the context of an 8-week summer treatment program for children with ADHD, high baseline self-perceptual bias predicted poorer treatment response as measured by observed conduct problems and peer sociometric ratings.¹¹ Recent work also suggests that self-perceptual bias may interfere with the accurate assessment of comorbid internalizing problems among children with ADHD. Specifically, Martin et al.¹⁰ examined self-perceptual bias as a suppressor variable, highlighting that the association between ADHD status and internalizing problems was stronger when self-perceptual bias was included in the model, as compared to when it was not. These findings suggest that compared to control children, social self-perceptual biases among children with ADHD may mask the severity of self-reported loneliness, and anxious and depressive symptoms. Given that positively biased self-perceptions could be a barrier to effective assessment and/or treatment for some individuals with ADHD, it is critical that research examine whether self-perceptual bias is a characteristic that is amenable to change.

Importantly, some argue that poor insight into impairment may result from a cognitive deficit rather than self-protective tendencies. Indeed, the neuropsychological deficit hypothesis posits that self-perceptual bias stems from an executive functioning (EF) deficit and an associated lack of personal self-awareness. Only a handful of studies have directly examined the association between cognitive functioning and self-perceptual bias in children with ADHD^18,19,25,26. First, multiple cross-sectional studies have linked EF deficits with positive self-perceptual biases in the academic and social domains.^25,26 In addition, McQuade et al.¹⁹ found cognitive deficits to partially mediate the relation between ADHD status (i.e., ADHD versus control) and positively biased self-perceptions across multiple domains. Hence, research to date suggests that cognitive deficits may indeed play some role in the positively biased self-perceptions of children with ADHD.

Manipulating Self-Perceptual Bias in a Lab Setting

In light of the potential negative consequences of displaying a positive self-perceptual bias, a question of great clinical importance is whether these biases are amenable to change. Several lab-based experimental studies provide initial evidence for the modifiability of self-perceptual bias.^23,27–29 For example, children with ADHD who were provided positive feedback about performance on a maze task improved the accuracy of their social competence self-appraisals, though positive feedback did not have the same effect for self-appraisals of academic performance.²⁹ More recently, Emeh and Mikami²⁷ asked parents to observe their children’s social interactions with peers and provide feedback aimed at improving their relationships. Parental warmth during feedback was associated with more accurate self-perceptions of social competence among children with ADHD, whereas parental criticism was related to less accurate perceptions of social competence. Together, these findings suggest that, under some circumstances, self-perceptual bias is a characteristic that is amenable to change.

In one of the most comprehensive studies of modifiability of self-perceptual bias, Hoza et al.²⁸ examined whether the presence of an external motivator could facilitate more accurate ratings of self-competence (see method section below for details). Findings indicated that positively biased self-perceptions among children with ADHD could in fact be reduced with a monetary incentive in the academic and behavioral domains. On average, however, even with this incentive, children with ADHD were not able to rate their self-competence as accurately as the comparison group. Interestingly, there were no reductions in social self-perceptual biases for children with or without ADHD across the conditions.

To further build our understanding of the modifiability of self-perceptual bias, the current study utilizes this same sample of children with ADHD to examine the individual differences in those who are (or are not) able to modify social and behavioral self-perceptual bias. Specifically, we investigated whether self-perceptual bias change status relates in a meaningful way to externalizing symptoms (i.e., ADHD symptoms, Oppositional Defiant Disorder (ODD) symptoms, internalizing symptoms (i.e., depression and anxiety), and cognitive functioning. Results will inform potential interventions designed to directly address biased self-perceptions of competence among children with ADHD.

In line with our study aims, we utilized LPA to group children based on the presence and modifiability of self-perceptual bias across a series of experimental conditions. To our knowledge, only two studies to date have utilized LPA to examine self-perceptual bias among children with ADHD,^17,30 and none have used LPA for the purposes of categorizing children based on the modifiability of their bias. Hence, no specific hypotheses were made regarding the number of groups that would result. However, based on overall evidence for the modifiability of self-perceptual bias in lab settings,^23,27,28 we expected at least three types of groups to emerge: 1) children with little or no self-perceptual bias across conditions, 2) children with consistent, large, positive self-perceptual bias scores, and 3) children who initially have large positive self-perceptual bias scores but who are motivated to be more accurate by an extrinsic motivator.

Based on past research indicating a positive relation between self-perceptual bias and externalizing problems^3,13, and in line with the self-protective hypothesis, we hypothesized a stepwise pattern in which children who show no biases would have lower levels of externalizing symptoms as compared to those who modify positively biased self-perceptions, who would in turn have lower levels of externalizing symptoms than children showing consistent large positive self-perceptual biases. In addition, in light of literature linking greater depression with more accurate self-perceptions of competence,^5,21,22 and given that depression and anxiety are highly correlated,³¹ we expected that individuals showing consistent large positive self-perceptual biases would display fewer internalizing symptoms than those who modify their self-perceptual bias, who, in turn, would have fewer internalizing symptoms than children who show a consistent ‘no bias.’ Finally, given that cognitive deficits appear to partially mediate the relation between ADHD status and biased self-perceptions,¹⁹ we examined if there were differences in cognitive functioning across self-perceptual bias subgroups. If the neuropsychological deficit hypothesis were supported, we would expect individuals with a consistent large positive self-perceptual bias to have lower cognitive functioning skills than children who modify their self-perceptual bias or children showing no bias.

Method

Participants

The current study was conducted in compliance with the Institutional Review Board, with informed consent obtained from each participant’s parent or guardian. Participants were recruited as part of a larger study of self-perceptual bias among children with ADHD and typically developing (TD) peers (see Hoza et al.²⁸). The current study focused only on those children with ADHD, including 178 participants between the ages of 7.68 and 11.42 (M_age = 9.23, SD_age = .94; 80% male; 84% white). Although TD children were not included in primary analyses, self-perceptual bias scores for TD children were used as a reference for PB cutoffs described in a later section. The sample of TD children included 86 participants without an ADHD diagnosis who were between the ages of 7.84 and 10.87 (M_age = 9.26, SD_age = .86; 72% male; 79% white).

Participants were recruited from two separate sites, spanning three geographic locations, using identical recruitment procedures (i.e., referrals from local medical and mental health professionals, news advertisements, area schools, and ADHD treatment services). Participants in the ADHD sample were required to meet the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text revision (DSM-IV-TR)³² criteria for ADHD, combined or predominantly hyperactive/impulsive type. TD participants could have no current or previous ADHD diagnosis but were not excluded if they met criteria for ODD, CD, or internalizing problems in order to maintain a representative sample. At each site, two separate doctoral-level diagnosticians arrived at diagnoses by reviewing all assessment data for each child, including the National Institute of Mental Health Diagnostic Interview Schedule for Children - Version IV³³, the Disruptive Behavior Disorders Rating Scale³⁴, the parent Child Behavior Checklist, and the Teacher Report Form³⁵ (see Hoza et al.²⁸ for detailed information on the assessment procedure). Additional study exclusions for both the ADHD and TD groups included: 1) IQ below 77 (1.5 SD below mean); 2) history of seizures; 3) treatment with medication that could not be withdrawn for testing (e.g., anti-depressants); or 4) a history or concurrent diagnosis of pervasive developmental disorder, schizophrenia, other psychotic disorders, sexual disorder, or an organic mental disorder. Individuals with ADHD predominantly inattentive type were excluded, as positively biased self-perceptions do not appear to be prevalent among these individuals.³⁶

Parents completed all measures as part of the eligibility screening process for the study. Once a child was deemed eligible for inclusion in the study, additional rating forms were obtained by the designated classroom teacher. All children completed the self-report measures and cognitive tasks during the child intake session (“baseline”). Children completed two additional ratings of self-concept on a separate day (see experimental procedure below).

Medication status of participants.

All children who typically took ADHD medication were unmedicated during on-site testing to eliminate any confounding effects produced by medication. Each child remained off medication for the minimum amount of time necessary for the study or following the guidelines of the child’s prescribing physician. Parents and teachers were directed to rate the children’s behavior off medication; however, for a subset of children (n = 44), teachers reported competence ratings about medicated behavior or did not identify whether their ratings were on or off medication.

Measures

Self- and teacher-reported competence.

The social acceptance and behavioral competence subscales from the Self-Perception Profile for Children (SPPC)³⁷ were completed by participants. Each six-item subscale was rated on a 4-point scale, with higher scores indicating higher levels of self-perceived competence. The 15-item teacher report version of the SPCC³⁷ was expanded for the current study to mirror the child report version. Thus, for the current study, teachers reported on participants’ competence using the social (5-item) and behavioral (6-item) domain subscales. All items were rated on a 4-point scale, with higher scores indicating higher levels of teacher-reported competence. Previous analyses using this sample have demonstrated adequate reliability, with coefficient alphas for child reports ranging from .76 to .92 and for teacher reports ranging from .90 to .97 (see Hoza et al.²⁸). Domain-specific discrepancy scores were calculated to provide an index of self-perceptual bias. Teacher ratings of child competence were subtracted from the child self-perception scores in each of the three conditions (baseline, matching, and money) and separately for the social and behavioral domains.

Child behavior problems.

Symptoms of ADHD and disruptive behavior disorders, including ADHD – Hyperactive/Impulsive (HI), ADHD – Inattentive (IA), and ODD were assessed with the Disruptive Behavior Disorder Rating Scale (DBD)³⁴, a 45-item questionnaire administered to both parents and teachers. Respondents indicated on a 4-point scale the degree to which a statement described the child’s behavior (0 = Not at All; 3 = Very Much). For the present study, a combined measure of parent and teacher symptom severity was created by averaging parent and teacher ratings together for each item and calculating an overall mean score for each subscale (HI, IA, ODD). Thus, subscale scores provide an index of overall symptom severity across settings (i.e., home, school), with higher scores indicating higher symptoms. Coefficient alphas for each of the combined parent/teacher subscales ranged from .94 to .97.

Child depressive symptoms.

Children’s depressive symptoms were assessed using the Children’s Depression Inventory (CDI),³⁸ a well-established self-report measure assessing cognitive, affective, and behavioral symptoms of depression. The CDI consists of 27 items, with three responses to choose from for each item (0 indicating least severe; 2 indicating most severe). Children were instructed to select the statement that best describes their feelings from the past two weeks. As recommended by previous research,³⁹ seven items were removed (Items 5, 15, 21, 22, 23, 26, 27) that overlap with behavioral symptoms of ADHD. Using the remaining 20 items, a mean score was calculated to provide an index of symptom severity (Cronbach’s α = 0.84).

Child anxiety symptoms.

Children’s anxiety symptoms were assessed with the Multidimensional Anxiety Scale for Children (MASC),⁴⁰ a self-report questionnaire that assesses four areas of anxiety: physical symptoms, harm avoidance, social anxiety, and separation/panic. Participants rated themselves on 39 items using a 4-point scale (0 = never true about me, 3 = often true about me). A mean score was calculated for an overall index of symptom severity, with higher scores indicating higher anxiety symptoms (Cronbach’s α = 0.89).

Child cognitive functioning.

Children’s cognitive functioning was assessed using the Executive Processes, Broad Attention, Working Memory, and Cognitive Fluency clusters of the Woodcock-Johnson Test of Cognitive Abilities, Third Edition (WJ-III COG).⁴¹ The Executive Processes cluster consists of three subtests used to assess the ability to shift one’s mental set, planning and decision-making ability, and interference control. The Broad Attention cluster consists of four subtests used to assess an array of attention indices, such as attentional capacity, sustained attention, divided attention, and selective attention. The Working Memory cluster is comprised of two subtests and measures one’s ability to mentally manipulate information held in immediate memory. Lastly, the Cognitive Fluency cluster is comprised of three subtests and measures ease and speed in performing cognitive tasks. For each cluster, standardized scores were calculated based on an age-equivalent normative sample.

Experimental Procedure and Manipulation

The procedure consisted of three experimental conditions. All children completed all three conditions. During the first condition, children were asked to complete the SPPC to provide baseline ratings of social and behavioral competence. On a separate day, children were informed that their teachers also had completed the competence ratings; they then were asked to complete the questionnaire for a second time, attempting to match the ratings they thought their teachers had given them. The research assistant and the child then engaged in a distractor task (Go Fish) for 5 minutes. After finishing the distractor task, children were told a mistake had been made and the research assistant forgot to inform them that they could earn money for each correct match to their teacher’s responses. The children were told that due to this oversight, they could fill out the questionnaire again and earn one dollar for each correct match with their teacher’s ratings. Before starting again, the research assistant had a short conversation with each child about what he or she would do with a possible $18 to make the money a more salient motivator. Children then completed the questionnaire for a third time. Children were not allowed to see their teacher’s ratings. Regardless of whether they earned it, all children were given a minimum of $5 for the task. Children received up to $18 if they matched the teacher exactly on all items.

Data Analytic Plan

Creation of Bias Subgroups.

LPA was used to examine patterns of change in domain-specific self-perceptual bias scores across the baseline, match, and money conditions. LPA is a multivariate statistical tool that estimates an underlying grouping variable (latent profile) that is not observed, but is inferred from a set of continuous dependent variables. For the present study, each estimated latent profile was based on the magnitude of the bias and pattern of change in bias across conditions (i.e., baseline, match, money). For each domain of competence, three models (2, 3, and 4 profiles) were estimated with profiles added iteratively to determine the best-fitting model. Following the guidelines outlined by Geiser⁴², models were evaluated using multiple statistical indicators: Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), Adjusted BIC (aBIC), and the Vuong-Lo-Mendell-Rubin ratio test (VLMR). The model with the smallest AIC, BIC, and aBIC values is considered to be the best fit to the data. The VLMR ratio test statistically compares the indicated model to a model with one fewer classes. The resulting p-value is used to determine which model demonstrates the best fit. Finally, each model was evaluated on its interpretability to determine whether each estimated profile has sufficient theoretical grounding to make valid interpretations.

After the best-fitting model was selected for each domain of competence, repeated measures analyses of variance (ANOVAs) were conducted to examine change in bias scores across conditions for each latent profile that emerged across models. Following a significant omnibus test, pairwise comparisons were used to understand significant change across the three conditions. To further aid in describing self-perceptual bias groups, reference was made to a bias cutoff score calculated using the TD sample. Specifically, following recent work by McQuade et al.¹⁸, we calculated values representing one SD above and below the mean of the TD sample across each domain of competence to approximate PB and Negative Bias (NB) designations (PB Cutoff: behavioral domain = 0.63; social domain = 0.75; NB Cutoff: behavioral domain = −1.01; social domain = −0.96). Given our use of LPA to reduce reliance on the cutoff score approach, these values were not intended to be used for strict categorization purposes, but instead as a reference point to help interpret the magnitude of self-perceptual bias scores emerging from the LPA and describe resulting bias subgroups.

Comparison of Bias Subgroups.

Univariate one-way ANOVAs were utilized to compare LPA-derived bias subgroups for each externalizing, internalizing, and cognitive functioning outcome. Post hoc comparisons using the Greenhouse-Geisser correction were conducted to understand significant group differences and to account for violations of the homogeneity of variance assumption. Because a primary aim of this study was to examine differences between children who are able to modify their self-perceptual bias versus children who are not able to modify their bias, group comparisons were only conducted when LPA revealed a bias change group within a given domain of competence (behavioral, social).

Results

Latent Profile Analysis Results

Model fit indices for each LPA model (i.e., 2, 3, and 4 profiles) are reported in Table 1. In the behavioral domain, the 4-profile solution revealed slightly lower AIC, BIC, and aBIC values, however, the VLMR indicated no significant differences between the 3- and 4-profile solutions (p = 0.59). Moreover, given that the 3-profile solution had greater parsimony and was deemed to be more consistent with expected group designations within the context of previous work, the 3-profile solution was deemed the best fitting model in the behavioral domain. In the social domain, the 4-profile solution revealed slightly lower AIC, BIC, and aBIC values. The VLMR also indicated the 4-profile solution fit significantly better than the 3-profile model (p = 0.04). Thus, the 4-profile solution was considered the best fit for the data in the social domain.

Table 1.

Fit Indices for Latent Profile Analysis Models, Separately Reported for the Behavioral and Social Domains

No. of Profiles	2	3	4
Behavioral Domain
AIC	1345.59	1219.62	1196.48
BIC	1377.41	1264.16	1253.75
ABIC	1345.74	1219.83	1196.75
VLMR	−749.431	−662.80*	−595.81
Social Domain
AIC	1301.26	1207.15	1185.76
BIC	1333.02	1251.62	1242.93
ABIC	1301.46	1207.28	1185.93
VLMR	−736.27*	−640.63*	−589.57*

Note. Bold entries reflect the selected model. AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion; ABIC = Adjusted BIC; VLMR = Vuong-Lo-Mendell-Rubin Ratio Test.

^*p < .05

Description of profiles.

Latent profile group means for the best-fitting models across each domain of competence are presented in Table 2 and the proportion of participants in each group is displayed in Figure 1. In the behavioral domain (3-profile solution), results from the repeated-measures ANOVAs indicated significant mean-level variability in bias scores across conditions (baseline, match, money) for Profile 1, F (2, 204) = 26.07, p < .001, and Profile 3, F (2, 90) = 5.26, p = .01, but not for Profile 2. For Profile 1, pairwise comparisons revealed mean bias scores at baseline were significantly higher than bias scores in the match condition, t (102) = 4.13, p < .001, and the money condition, t (102) = 6.20, p < .001. Bias scores in the match condition were significantly higher than scores in the money condition, t (102) = 3.78, p < .001. For Profile 3, pairwise comparisons revealed that mean bias scores at baseline were significantly higher than scores in the match, t (45) = 2.21, p = .032, and money conditions, t (45) = 2.64, p = .011, but mean bias scores did not differ between the match and money conditions. Profile 1 was assigned the descriptive label “PB Change” given this group began at baseline with a mean bias score above the PB cutoff value described above and changed to reduce bias over the course of the match and money conditions to a level no longer meeting the PB designation. Profile 2 was given the descriptive label “Consistent Small NB” due to the presence of negative self-perceptual biases across each condition. Because mean bias scores across each condition for Profile 2 did not reach the NB cutoff, the word “small” was added to the profile description. Finally, Profile 3 was assigned the descriptive label “Consistent Large PB” due to the presence of a large positive self-perceptual bias across each condition. Although Profile 3 demonstrated a significant reduction in bias from the baseline to match condition, mean self-perceptual bias scores across each condition were at least two SDs above mean bias scores of the TD sample, suggesting that the biases of individuals belonging to this group did not ‘normalize.’

Table 2.

Latent Profile Group Means on Self-Perceptual Bias Scores Across Conditions for the Behavioral, Social, and Academic Domains

	Assigned Group Label	Bias Score at Baseline Condition		Bias Score at Match Condition		Bias Score at Money Condition
		M	SD	M	SD	M	SD
Behavioral (n)
Profile 1 (103)	“PB Change”	0.86a	0.62	0.57b	0.45	0.38c	0.53
Profile 2 (29)	“Consistent Small NB”	−0.64	0.65	−0.70	0.51	−0.84	0.62
Profile 3 (46)	“Consistent Large PB”	2.00a	0.54	1.81b	0.45	1.64b	0.75
Social (n)
Profile 1 (45)	“Consistent No Bias”	−0.21	0.67	−0.28	0.30	−0.22	0.34
Profile 2 (17)	“Consistent NB”	−0.88	0.52	−1.18	0.26	−1.06	0.37
Profile 3 (37)	“Consistent Large PB”	1.62	0.69	1.83	0.50	1.65	0.49
Profile 4 (78)	“Consistent Small PB”	0.69	0.70	0.66	0.46	0.63	0.41

Note. Significant post hoc comparisons are identified by superscripts a, b and c. Means that differ significantly within each domain of competence are identified by different superscripts, means that do not differ between groups are identified by the same superscript.

Figure 1.

Proportion of participants across bias subgroups for behavioral and social domains.

In the social domain (4-profile solution), results indicated no mean-level variability across conditions for any of the four profiles. Profile 1 was assigned the descriptive label “Consistent No Bias” given that this group had a mean bias score close to zero across all three conditions. Profile 2 was given the label “Consistent NB” due to the presence of negative bias scores across the three conditions. Of note, the mean bias score at baseline did not meet the NB designation, but scores at the match and money conditions did meet the NB criteria. Profile 3 was assigned the descriptive label “Consistent Large PB” given that this group showed a stable pattern of a large PB across the three conditions. Finally, Profile 4 was provided the label “Consistent Small PB” given that this group showed a stable but low pattern of positive self-perceptual biases across the three conditions. Of note, mean bias scores across each condition were slightly below the PB cutoff score and as such, the word “small” was included in the profile description. Because LPA revealed no group of participants demonstrating a significant change in self-perceptual bias scores across the baseline, match, and money conditions, follow-up analyses comparing groups on externalizing, internalizing, and cognitive functioning outcomes were not examined within the social domain.

Bias Subgroup Comparisons

Missing data.

Most study variables had less than 5% missing data, which is unlikely to bias findings.⁴³ There were three variables missing more than 5% of data: Executive Processes: 6%; Working Memory: 33%; and Broad Attention: 33%. These data were missing due to a test administration error at one data collection site and missingness was not believed to be related to a systematic bias. To confirm this, patterns of missingness were examined using Little’s Missing Completely at Random (MCAR) test. Little’s MCAR test examines the null hypothesis that missing data are missing completely at random. The MCAR test was not significant for Executive Processes, Working Memory, or Broad Attention (χ²(12) = 11.40, p > .05), indicating that the data were unlikely to bias parameter estimates.

Teacher- and parent-reports of externalizing problems.

Differences in average teacher/parent-reported symptoms of HI, IA, and ODD were examined across bias subgroups in the behavioral domain. Results indicated mean-level variability across groups in symptoms of HI: F (2, 174) = 22.78, p < .001, IA: F (2, 174) = 3.02, p = .05, and ODD: F (2, 174) = 5.28, p = .01. Children in the Consistent Large PB group had higher levels of HI (p < .001; Cohen’s d = 1.71), IA (p = .03; Cohen’s d = 0.64), and ODD (p = .003; Cohen’s d = 0.81) symptoms than children in the Consistent Small NB group; they also had higher levels of HI symptoms than those in the PB Change group (p < .001; Cohen’s d = 0.70). Moreover, children in the PB Change group had higher parent/teacher reported levels of HI (p < .001; Cohen’s d = 0.90) and ODD (p = .002; Cohen’s d = 0.68) symptoms than children in the Consistent Small NB group.

Self-report of internalizing problems.

Differences in self-reported symptoms of depression and anxiety were examined across bias subgroups in the behavioral domain. Results indicated mean-level variability in depressive symptoms across groups, F (2, 175) = 3.66, p = .03. Post hoc comparisons revealed marginally significant differences such that depressive symptoms for children in the Consistent Large PB group were lower than those in the PB Change (p = .06; Cohen’s d = 0.40,) and Consistent Small NB groups (p = .06; Cohen’s d = 0.59). There were no significant differences across bias subgroups for symptoms of anxiety.

Child cognitive functioning.

Differences in child cognitive functioning were examined across bias subgroups in the behavioral domain. Results indicated that mean scores for Broad Attention, F (2, 116) = 8.40, p < .001, and Working Memory, F (2, 117) = 5.08, p = .01, varied across groups. Individuals in the Consistent Large PB group had lower mean scores in Broad Attention (p = .002; Cohen’s d = 0.79) and Working Memory (p = .004; Cohen’s d = 0.69) than those in the PB Change group. Individuals in the Consistent Small NB group also had significantly lower Broad Attention scores than those in the PB Change group (p = .01; Cohen’s d = 0.74). There were no significant differences in Working Memory scores between those in the Consistent Small NB and PB Change groups.

Discussion

This study examined if externalizing symptoms, internalizing symptoms, and cognitive functioning among children with ADHD differed based on the extent to which self-perceptual bias was present and modifiable in the context of a potent motivator for self-perceptual accuracy. Findings from Hoza et al.²⁸ showed that, on average, children with ADHD could be motivated to reduce their positive self-perceptual biases in some domains, but did not rate their self-competence as accurately as their typically developing peers. This study demonstrates the existence of a subgroup of children with ADHD who initially present with a positive self-perceptual bias in the behavioral domain but who can be motivated to improve the accuracy of their self-perceptions to a level that is no longer considered biased. Further, across both competency domains, there were groups of children with either consistent negative biases or no biases, highlighting that many children with ADHD do not hold positively biased views of self-competence. Although it is important to have an overall understanding of self-perceptual bias in children with ADHD, studies that look only at overall average levels of bias do not tell the whole story. This study highlights the important individual variability from child to child in both the presence of biases and the ability to modify self-perceptions.

One strength of the current study was the use of a novel approach to categorizing children based on the modifiability of their self-perceptual biases. Children were categorized based on the pattern and magnitude of their biases across the three experimental conditions (baseline, match, money) using a statistical technique that estimates group membership based on a set of continuous dependent variables (i.e., self-perceptual bias discrepancy scores). The use of LPA to create bias subgroups addresses a concern that PB designations using a cutoff score might be arbitrary.¹⁷ Although the use of a +1.0 SD or + 1.0 point cutoff score to designate children as having a PB is used in the literature and is arguably a useful approach when examining PB at one time point, it presents a challenge when examining change in PB over a short span of time. For example, with the use of a 1 point cutoff score, a child who only changes by a small amount but crosses the cutoff score threshold (e.g., 1.1 to 0.9) would be placed in the PB Change group, whereas, a child who changes by a much larger amount but who stays above the cutoff score threshold (e.g., 2.2 to 1.1) would not be categorized into the change group using this approach. Using LPA, we address the potential for arbitrary categorizations by estimating statistically-derived groups that do not rely on a predetermined cutoff value to determine group designations. Nonetheless, in order to describe and interpret the statistically-derived groups, it remained necessary to consider the bias scores in the context of the broader PB literature. To this end, the present study referenced a +/− 1 SD cutoff score as described by McQuade et al.¹⁸ to help assign descriptive labels to the LPA-derived groups and to aid in the interpretation of findings.

Results in the behavioral domain partially aligned with the a priori hypothesis. LPA revealed a Consistent Large PB and PB Change group, as was predicted. Rather than the ‘No Bias’ group that was expected, the third group in the behavioral domain represented children with a consistent negative bias (i.e., Consistent Small NB). Crucially, like the PB Change group, the Consistent Large PB group also showed a significant reduction in self-perceptual bias from the baseline to match condition, though not from the match to money condition. However, mean bias scores at each condition were more than two SDs above the mean of the control group, suggesting that the self-perceptual biases among children in the Consistent Large PB group did not ‘normalize’ in response to an external motivator for self-perceptual accuracy. In contrast, the PB Change group evidenced an average self-perceptual bias score of approximately half a SD above the mean of the control group in the final ‘money’ condition, which supports the notion that children in this group normalized in their ratings of self-competence.

In the social domain, four bias subgroups emerged, including two consistent positive self-perceptual bias groups (i.e., Consistent Large PB, Consistent Small PB), a Consistent No Bias group, and a Consistent NB group. Unlike in the behavioral domain, there was no evidence for a group with a modifiable self-perceptual bias in the social domain. Within the context of previous research using this sample,²⁸ the lack of a change group in the social domain is perhaps unsurprising. Specifically, Hoza et al.²⁸ reported no significant reductions in social bias for the ADHD group. The present findings also suggest there is no meaningful subset of children with ADHD who are able to reduce positive self-perceptual biases in the social domain. As noted by Hoza et al.²⁸, these findings might stem from multiple factors. For example, children might receive less direct feedback from teachers regarding social functioning, as compared to academic or behavioral functioning. Also, it is possible that acknowledging low social competence in itself may be a particularly threatening experience, thus leading to a more change-resistant social positive self-perceptual bias.

The current study further extends prior work by identifying factors related to whether or not children may be able to modify their behavioral self-perceptual biases to be more accurate. When differences in internalizing and externalizing symptoms emerged across the three bias subgroups in the behavioral domain, children with a consistent positive self-perceptual bias (i.e., ‘Consistent Large PB’) had higher levels of externalizing symptoms and lower levels of internalizing symptoms as compared to children in the Consistent Small NB group. This overall pattern of results remains consistent with prior work linking positively biased self-perceptions of competence with lower levels of depressive symptoms^5,22 and higher levels of externalizing behaviors, such as aggression, oppositionality, and ADHD symptoms.^3,13 Of note, bias subgroups did not differ in self-reported anxiety.

Interestingly, children who started with a positive self-perceptual bias at baseline but reduced to more normative levels of bias in either the match or money conditions (i.e., ‘PB Change’) varied in the degree to which they differed from the Consistent Large PB and Consistent Small NB groups, depending on the outcome of interest. In full support of the a priori hypothesis, children in the Consistent Small NB group had lower levels of HI symptoms as compared to children in the PB Change group, who in turn, had lower levels of HI symptoms than children in the Consistent Large PB group. These findings highlight that severity of HI symptoms may be one primary difference between children who are motivated to change their reports of self-competence and those who remain positively biased when presented with external motivation to change. In addition, results demonstrated a marginally significant difference between bias subgroups for symptoms of depression, such that children in the Consistent Large PB group had lower levels of depressive symptoms than children in the PB Change group. Although this finding warrants replication, it suggests that depressive symptoms may also be an important factor related to the modifiability of positively biased self-perceptions of competence.

With regard to cognitive functioning, results in the behavioral domain showed partial alignment with the a priori hypothesis. As predicted, participants who held a consistent positive self-perceptual bias across experimental conditions demonstrated lower levels of cognitive functioning in some areas as compared to children in the PB Change group; however, only in one domain of cognitive functioning (i.e., Broad Attention) did the PB Change group show higher levels of cognitive functioning than the Consistent Small NB group. These findings suggest that although cognitive ability may not fully distinguish those with and without a positive self-perceptual bias, it may play an essential role in the ability to modify this type of bias.

Results from this study help inform the theoretical explanations for self-perceptual bias. In past work, researchers have concluded that the self-protection hypothesis (i.e., the idea that children with ADHD exaggerate reports of self-competence to help preserve their self-esteem) cannot account for the self-perceptual bias phenomenon in its entirety given that, on average, children with ADHD who demonstrate positively biased self-perceptions are not able to rate their self-competency as accurately as those of their typically developing peers, even when provided with external motivation to do so.²⁸ Our study supports this conclusion in part, given the presence of subgroups of children who were not able to modify their social or behavioral self-perceptions to a level considered normalized, even when confronted with a potent motivator. However, the self-protection hypothesis may be a viable theory for explaining the positive self-perceptual biases of the children who were able to modify their self-perceptions of behavioral competence to a level on par with TD peers. The fact that these children have the self-awareness to provide more accurate reports of their self-competence, even if they do not provide these accurate reports initially, highlights the possibility that they are using inflated self-views to protect their self-esteem.

An alternative theoretical perspective to the self-protection hypothesis is clearly warranted to comprehensively account for the potential factors that influence the development of self-perceptual bias, particularly for children who remain positively biased even when presented with an external motivator to change. Hoza et al.²⁸ consider the neuropsychological deficit hypothesis (i.e., the idea that positively biased self-perceptions stems from a cognitive functioning deficit, which is thought to result in a lack of personal self-awareness) as a possible explanation for this finding. Indeed, McQuade and colleagues¹⁹ demonstrated that across multiple domains, cognitive deficits partially mediated the relation between ADHD status and positively biased self-perceptions. Our findings extend this work by demonstrating that children with a consistent behavioral PB had lower mean scores on Broad Attention and Working Memory than children in the PB Change group. In line with the neuropsychological deficit hypothesis, our results suggest that a cognitive functioning deficit may play a role in the positively biased self-perceptions of the Consistent Large PB subgroup. Importantly, our work suggests that the self-protection hypothesis and the neuropsychological deficit hypothesis are not necessarily mutually exclusive, and that the factors that influence the development of self-perceptual bias may not be the same across all children with ADHD who display such biases.

This study has important implications for the treatment of children with ADHD. Given that prior work suggests that positive self-perceptual biases interfere with the assessment of comorbid internalizing problems¹⁰ and treatment responsiveness,^6,11,24 an intervention targeting self-perceptual bias should be considered. Our results indicate that such an intervention may need to vary depending on the extent to which self-perceptual biases in the behavioral domain are modifiable. Importantly, even if a child appears to be positively biased upon initial presentation, our results suggest that some children may in fact have some level of insight into their impairments, although this may not be true for all domains of competence (i.e., social). Determining how to best identify children who have a modifiable self-perceptual bias and how to elicit enough self-awareness to promote treatment response is a critical area of future research. Moreover, our results suggest that, in the behavioral domain, children who have the most severe symptoms of externalizing behaviors, lower depressive symptoms, and have a lower degree of cognitive functioning may be those most likely to display a bias that is difficult to modify to a level consider normalized. This finding is particularly concerning, as these are the children in greatest need of intervention and yet, they appear to be at greatest risk for a bias that interferes with treatment. As such, interventions created to modify positive self-perceptual bias will need to target this group of children and determine a way to promote enough insight into impairments that treatment is feasible. Finally, when considering an intervention that specifically targets positive self-perceptual bias, it is critical that those administering the intervention simultaneously focus on increasing actual competence and daily functioning so as to prevent a potential increase in depression that may coincide with a decrease in positive self-perceptual biases.²²

This study has several important limitations. First, given that our sample included only children with ADHD – Hyperactive-Impulsive or Combined presentations, findings cannot be generalized to children with the Inattentive presentation. Next, given that the present sample consisted of only 7–10-year-old children, more research is needed to examine the modifiability of self-perceptual bias in adolescent and adult populations. Most importantly, given the cross-sectional nature of the study, we are not able to address the issue of directionality in our findings. Future research should examine the modifiability of self-perceptual bias longitudinally to address this issue.

Summary

In sum, the present study demonstrates that multiple factors may be associated with self-perceptual bias (e.g., self-protection and cognitive impairment), and that these factors may differ across children. Notably, many children with ADHD did not demonstrate positively biased self-views of competence. Among those who did demonstrate a positive self-perceptual bias, a number of children in both the social and behavioral domains remained consistently biased even when presented with a potent motivator for change. Only in the behavioral domain was there evidence for a group of children who were able to modify their positively biased self-perceptions to be more accurate when presented with an external motivator for self-perceptual accuracy. It is critical that treatment for ADHD be flexible enough to identify children who have a modifiable positive self-perceptual bias, in order to promote a self-awareness that may improve response to intervention. It will also be important for future research to examine how best to treat children in the Consistent Large PB groups, who appear to have the most severe externalizing symptoms and lowest levels of cognitive functioning, while also showing the least insight into impairment.

Table 3.

Group Means and Standard Deviations for Key Study Variables in the Behavioral Domain

	Behavioral Competence
	Consistent Small NB		PB Change		Consistent Large PB
	(n = 29)		(n = 103)		(n = 46)
Measure	M	SD	M	SD	M	SD
Externalizing Symptoms
HI Symptoms	1.37a	.44	1.79b	.48	2.11c	.42
IA Symptoms	1.78a	.44	1.98a,b	.51	2.04b	.39
ODD Symptoms	.97a	.45	1.33b	.61	1.35b	.50
Internalizing Symptoms
Dep. Symptoms	.44a	.36	.37a	.30	27b+	.24
Anx. Symptoms	1.44	.46	1.22	.52	1.27	.45
Cognitive Functioning
Executive Processes	100.89	12.06	103.19	12.13	98.23	12.73
Cognitive Fluency	86.31	14.64	90.90	13.22	90.00	15.58
Broad Attention	94.90a	11.18	103.57b	12.05	94.61a	10.58
Working Memory	96.45a,b	12.78	102.96a	14.54	94.18b	10.62

Note. Dep. = Depressive; Anx. = Anxiety; HI = hyperactivity/impulsivity; IA = inattention; Significant post hoc comparisons are identified by superscripts a, b, and c. Means that differ significantly within each domain of competence are identified by different superscripts, means that do not differ between groups are identified by the same superscript.

⁺= Marginally significant post hoc comparison.