Sex Differences and Parent–Teacher Discrepancies in Reports of Autism Traits: Evidence for Camouflaging in a School Setting

Abstract

The purpose of this study was to examine sex differences and informant discrepancies in parent- and teacher-reports of autism traits. Data were drawn from the Simons Simplex Collection to create a sex-matched sample of autistic youth (N = 388; 4–17 years). Included participants had both parent and teacher reports of autistic traits from the Social Responsiveness Scale (SRS). Within each sex group, parent and teacher raw SRS scores were compared. Scores within each informant group (parent/teacher) was assessed for sex differences. Predictors of parent–teacher discrepancies were examined. Despite no sex differences in parent-reported autistic traits, teachers reported males as having more autistic traits compared to females. Parents of females reported significantly more autistic traits than teachers across multiple domains. Being older and female were significant predictors of increased parent–teacher discrepancy for multiple domains. These results suggest discrepancies between the observed autistic traits for females at home and school and builds on the growing body of literature highlighting potential camouflaging across development in autistic youth: parent–teacher discrepancies may reflect ways that autistic females are overlooked by teachers due to conscious changes in behavior or gender-based expectations of female characteristics. Discussion of discrepancies on an individual basis may therefore alleviate potential long-term consequences of camouflaging.

Parents and teachers serve as important informants of an autistic child’s behavior and needs, as well as primary advocates for a child receiving an autism assessment or services. Reports across informants are relied upon during the diagnostic process (Campbell et al., 2020), and parents and teachers are key contributors to a child’s Individualized Education Plan in school settings. Given the critical role of both parents and teachers, disagreement between the two can adversely impact diagnostic decisions, access to supports, and accurate tracking of development for these children (Garbacz et al., 2016).

Existing studies suggest low to moderate levels of agreement between parents and teachers of autistic children on ratings of autism traits (Lopata et al., 2016; Stratis & Lecavalier, 2017; Szatmari et al., 1994). Examinations of inter-rater reliability between parents and teachers have produced mixed results. Thompson and Winsler (2018) reported a moderate and significant correlation between parent and teacher ratings on a preschool developmental screening measure for social skills, but not for behavioral concerns. Studies with school-aged children have found significant, but low in magnitude correlations (Lopata et al., 2016), or significant correlations only for children classified as more severely impacted by autism (Azad et al., 2016). Similarly, Levinson et al., (2021) found that discrepancies between ratings of behavioral concerns decreased with increasing clinical ratings of autism severity. The directionality of this discrepancy appears to vary. The majority of studies suggest that parents rate their children as having more difficulties in social or behavioral development than teachers (Levinson et al., 2021; Lopata et al., 2016; Palmer et al., 2023; Sturrock et al., 2020), yet others report greater teacher endorsement of difficulties than parents (Szatmari et al., 1994; Thompson & Winsler, 2018).

Although this body of research provides foundational evidence of a trend in parent–teacher discrepancies for autistic children, it faces an issue that is consistent across the majority of autism research: a substantial lack of female participants. Recent reports suggest that 3.8 school-aged male children are diagnosed with autism for every 1 school-aged female child (Maenner, 2023). While the school age ratio has decreased over time (Baio et al., 2018), it remains more imbalanced than the near 1:1 ratio seen in adults (Rutherford et al., 2016) and some elevated-likelihood sibling samples (Burrows et al., 2022). This suggests that while there have been improvements in identifying autistic females earlier in life, females are still diagnosed less and later than males (Harrop et al., 2024). The majority of published research in autism reflects these sex ratios, and the lack of female representation has made it difficult to generalize prior study findings to autistic females.

Only one study, to our knowledge, has directly compared parent and teacher reports and evaluated sex differences between informants of autistic children’s behavior. Sturrock and colleagues (2020) compared parent and teacher scores on the Strengths and Difficulties Questionnaire (SDQ; Goodman, 1997) for school-aged autistic males and females. The SDQ is not specific to autism, but rather measures child mental health and behavior through scores on emotional regulation, conduct, hyperactivity, peer relations, and prosocial behavior. In this study, parent reports of concerns for their children on the SDQ were greater than those reported by teachers. Further, this discrepancy was significantly greater for females than for males. A similar study using the SDQ found that although there were no sex differences for parent-reported scores, teachers reported significantly more concerns for males than females (Mandy et al., 2012).

Despite the paucity of research for sex-specific parent–teacher discrepancies within the field of autism, the patterns found in these two studies have been relatively well-documented for discrepancies between parents and clinicians. Several studies have found sex differences in parent-reported autism traits, despite no observed clinical differences between males and females (Dworzynski et al., 2012; Ratto et al., 2018; Ross et al., 2022). While this disparity may reflect implicit bias in the evaluation of autism characteristics in females, a phenomenon reported by autistic females and their families (Cridland et al., 2014; Lockwood Estrin et al., 2021), it may also reflect differences in how clinician-rated observational measures (e.g., ADOS-2, Lord & Rutter, 2012) and parent-report measures (e.g., Social Responsiveness Scales, Constantino & Gruber, 2005, 2012) capture autism traits in females (Ratto et al., 2018). In one study evaluating clinician perspectives towards sex differences, 36% of surveyed clinicians reported relying more on clinical impressions than on formal instruments when assessing females (Jamison et al., 2017). This parallel body of work not only highlights a need to further examine sex differences in informant reports, but it also underscores the importance of considering the instruments used for informant report. The current study evaluates informant reports of autism characteristics using the same tool, thus limiting the possibility that informant discrepancies could stem from differences in measurement.

Given the pattern of disagreement between parents and teachers among samples of primarily male autistic children, and the evidence of female-specific disparities between parents and clinicians, it is important to determine whether discrepancies between parents and teachers are exacerbated for autistic females on ratings of autism traits. Teachers have been less likely to identify a female vignette as autistic even when she is described with the same traits as a male vignette (Whitlock et al., 2020), and playground behavior amongst groups of girls has been shown to camouflage the social struggles of autistic females at school (Dean et al., 2017). Qualitative reports from late-diagnosed women and mothers of adolescent girls suggest that teachers tend to overlook evidence of autism if a young girl is well behaved or performs well academically (Bargiela et al., 2016; Tierney et al., 2016). However, no study has examined teacher perceptions towards females on a tool used for autism screening and diagnosis, a factor that may bring to light the translational impact of such biases evident in previous work.

The central research question for this current study is, “Are there differences between how parents and teachers report the presence of autism characteristics in autistic males and females?”. To answer this, we leverage a large dataset of school-age children (4–17 years) to assess sex differences between parents and teacher reports of autism traits and related characteristics. Additionally, we compare these informant scores for each sex group and evaluate informant (parent–teacher) discrepancies. Finally, we test for possible predictors of informant discrepancy within this sample. Based on previous evidence, our overall hypothesis was that differences between parents and teachers would be greater for females than males.

Methods

Data Source

The current study utilizes data from the Simons Simplex Collection (SSC; Fischbach & Lord, 2010). The SSC comprises data from proband (autistic individuals, N = 2,58), parents, and unaffected siblings who were recruited as part of a multi-site genetic consortium. Recruitment for SSC was conducted at 12 different universities in the United States between 2008 and 2010 (Fischbach & Lord, 2010). Standardized phenotypic and genotypic data were collected across sites; for information about site reliability, measurement, and supervision, see Fischbach and Lord (2010). Exclusion criteria included not meeting threshold for a clinical diagnosis of autism, having any first-degree, autistic relatives, having medically significant perinatal events, having low mental age (< 18 months nonverbal age), or genetic evidence of Fragile X or Down Syndrome (Fischbach & Lord, 2010). The use of the SSC dataset for the current study was approved by the Institutional Review Board (#22–2481) at University of North Carolina, Chapel Hill.

Participants

Participants included youth ages 4–17 years (M = 8.88 years, SD = 3.53; 50% male) with a DSM-IV diagnosis of Autistic disorder, Aspergers, or PDD-NOS. To be included in the current study, participants were required to have complete data from both parent and teacher forms of the Social Responsiveness Scale (SRS; Constantino & Gruber, 2005). Of the 375 females in the SSC dataset, 194 met this criterion. Sex-balanced groups were achieved by matching each female to a corresponding male, with complete parent and teacher SRS data, based on age and full-scale intelligence quotient (IQ). The resulting male and female groups did not significant differ on chronological age, t(377) = − 1.44, p = 0.015, d = 0.15, IQ, t(377) = − 0.01, p = 0.98, d = 0.001, or clinician-rated autism characteristics, t(385) = 1.02, p = 0.31, d = 0.1. Demographic information and individual characteristics for the full sample and group differences are available in Table 1.

Table 1.

Characteristics of the Included Sample of Participants

	Full Sample (N = 388)	Males (n = 194)	Females (n = 194)
Age (Years)
Mean (SD)	8.88 (3.53)	8.63 (3.25)	9.14 (3.25)
Range	4 – 17.58	4 – 17.67	4 – 17.58
IQ
Mean (SD)	74.8 (27.14)	74.78 (27.17)	74.82 (27.18)
Range	19 – 142	19 – 142	19 – 141
ADOS CSS
Mean (SD)	7.65 (1.69)	7.74 (1.74)	7.57 (1.66)
Range	4 – 10	4 – 10	4 – 10
Race (n)
African American	17	7	10
Asian	13	5	8
More than one race	30	17	13
Native American	1	1	0
Not Specified	3	1	2
Other	13	8	5
White	311	155	156
Ethnicity (n)
Hispanic	41	20	21
Non-Hispanic	246	174	172
Not Specified	1		1
Household Income (n)
Less than $20,000	30	16	14
$21,000 – $35,000	38	18	20
$36,000 – $50,000	48	24	24
$51,000 – $65,000	45	27	18
$66,000 – $80,000	67	32	35
$81,000 – $100,000	58	27	31
$101,000 – $130,000	11	7	4
$131,000 – $160,000	23	10	13
Over $161,000	54	28	26
Not reported	14	5	9
Maternal Education (n)
Associates	33	17	16
Bachelor’s	143	72	71
GED	3	0	3
Graduate School	103	47	56
High School	24	10	14
Some College	78	47	31
Some High School	3	1	2
Not reported	1	0	1

Measures

Clinician-Reported Autism Characteristics

Autism characteristics were assessed by clinicians using the Autism Diagnostic Observation Schedule-2nd edition (ADOS-2; Lord & Rutter, 2012). The ADOS-2 is a semi-structured behavioral assessment of autistic traits including communication, social reciprocity, and restricted/repetitive behaviors. The ADOS has five modules, based on chronological age and expressive language level (toddler and modules 1–4). To compare across modules, the calibrated severity score (CSS; range = 0–10) was used to measure the presence and intensity of autism characteristics as rated by clinicians.

Cognitive Ability

Depending on the age and verbal abilities of the participant, full-scale intelligence quotients (FSIQ) were assessed using the Wechsler Abbreviated Scale of Intelligence (Wechsler, 2011), Wechsler Intelligence Scale for Children—4th Edition (Wechsler, 2003), Differential Ability Scales-II (Elliott, 2007), or Mullen Scales of Early Learning (Mullen, 1995). Standard scores (M = 100, SD = 15) accounting for age and sex were used for each measure and thus are comparable across participants.

Informant-Reported Autism Characteristics

The Social Responsiveness Scale (SRS; Constantino & Gruber, 2005) is a parent- and teacher-report of autism-related characteristics across 65 items rated on a 4-point Likert Scale. The SRS is the original precursor to the currently used SRS-2 (Constantino & Gruber, 2012) and was developed to capture autism characteristics in individuals ranging from 4- to 18-years-old. The original SRS is identical to the SRS-2 School Age Form, used with the same age range (Bruni, 2014). Items comprise five subdomains, including social communication, social motivation, social awareness, social cognition, and repetitive/restricted behaviors (henceforth “RRBs”; Called “autistic mannerisms” in the SRS). The SRS also provides an overall total score as a continuous measure of severity ranging from 0 to 195. The SRS boasts strong internal consistency (Constantino et al., 2007), test–retest reliability (Constantino et al., 2003), and interrater reliability specific to parents and teachers (Pine et al., 2006). Parent- and teacher-reported raw scores were used for all six domains (five sub domains and the total score). Raw scores were used instead of sex-normed standardized scores to not prematurely adjust for sex (Kaat et al., 2021) and to avoid basing our interpretations on outdated norms.

Analytic Approach

SRS Differences by Sex and Informant Group

To determine whether parent- and teacher-reported SRS scores differed for males and females in this sample, we conducted a series of Welch’s independent t-tests. Effect sizes for mean comparisons were calculated using Cohen’s d. First, we tested for sex differences within each informant group (i.e., mean teacher-reported scores for males compared to teacher-reported scores for females). Mean scores for males and females were compared for the raw SRS total and for each raw domain score, but not for the SRS T-score due to its sex-specific scoring approach.

Next, we tested for informant differences within each sex group (i.e., mean parent-reported scores for females compared to teacher-reported scores for females). Mean scores were compared between parents and teachers using T-tests for all SRS scores including the raw total and the raw domain scores.

Predictors of Informant Differences

We calculated an informant discrepancy value for each included SRS score by subtracting the teacher-reported score from each respective parent-reported score (e.g., parent-reported Social Motivation score—teacher-reported Social Motivation score). Therefore, a positive informant discrepancy value indicates that a parent endorsed greater autism traits/severity than the teacher for a given child, whereas a negative value indicates the inverse. Bland–Altman (1986) plots were used to visually depict the mean and range of these informant discrepancy values by SRS score and sex group, as well as to determine whether informant discrepancy was associated with SRS score severity. This was done by regressing informant discrepancy values on the average of the parent and teacher scores (Bland & Altman, 1986), and is an approach used by previous studies characterizing parent–teacher discrepancies (Lopata et al., 2016; McDonald et al., 2016). To explore possible predictors of any differences in parent- and teacher-reported autism characteristics, we used additional linear regression models with informant discrepancy as the dependent variable. Within each model, informant discrepancy was regressed on the following child characteristics: sex, race, age, FSIQ score, and ADOS CSS score.

Results

Sex Differences within Informant Groups

No significant sex differences were found for any parent-reported SRS score (p’s > 0.05). However, consistent sex differences were found across teacher-reported SRS scores, with teachers reporting significantly higher scores for males than females in all domains (Fig. 1): Raw total, t(378) = − 4.23, p < 0.001, d = 0.43, Social Awareness, t(380) = − 3.24, p = 0.001, d = 0.33, Social Cognition, t(377) = − 3.64, p < 0.001, d = 0.37, Social Communication, t(379) = − 3.97, p < 0.001, d = 0.4, Social Motivation, t(381) = − 2.84, p = 0.004, d = 0.29, and RRBs, t(384) = − 4.34, p < 0.001, d = 0.44.

Fig. 1.

Mean raw total and raw domain scores for males and females by parent and teacher report. No sex differences were found for any parent-reported scores, but teachers reported males in the sample as having significantly higher scores (indicative of more/more intense autism characteristics) for all scores

Informant Differences Within Sex Groups

There were no significant differences between informants for males on the raw total or any domain scores (Table 2). All mean parent-reported scores for females were higher than teacher-reported scores. The female group had significant differences between informants for the raw SRS total score, along with the domain scores of social awareness, social cognition, and RRBs, with differences in the social communication domain trending towards significance (Table 2).

Table 2.

Mean (SD) SRS scores and informant differences within each sex group

	Females			Males
SRS Score	Parents	Teachers	Informant Difference	Parents	Teachers	Informant Difference
Raw SRS Total	100 (26.4)	91.49 (33.31)	t(366) = 2.79 p = .006 d = 0.28	102.71 (29.91)	104.91 (29)	t(386) = −0.75 p = 0.45 d = 0.08
Social Awareness	12.71 (3.72)	11.73 (4.74)	t(373) = 2.35 p = 0.02 d = 0.24	13.35 (3.96)	13.12 (3.98)	t(386) = 0.55 p = 0.58 d = 0.06
Social Cognition	18.99 (5.4)	17.22 (6.57)	t(371) = 2.90 p = 0.004 d = 0.29	19.38 (5.88)	19.48 (6.54)	t(385) = −0.17 p = 0.87 d = 0.02
Social Communication	34.11 (9.44)	32.05 (12.23)	t(363) = 1.86 p = 0.06 d = 0.19	35.1 (11.03)	36.69 (10.72)	t(386) = −1.44 p = 0.15 d = 0.15
Social Motivation	15.43 (5.61)	14.49 (6.46)	t(378) = 1.52 p = 0.13 d = 0.15	15.14 (6.04)	16.26 (5.77)	t(385) = −1.86 p = 0.06 d = 0.19
RRBs	18.75 (7.14)	16 (7.86)	t(382) = 3.61 p < .001 d = 0.37	19.73 (6.48)	19.35 (7.32)	t(380) = 0.54 p = 0.59 d = 0.06

Calculated informant discrepancy values (teacher score subtracted from the parent score) revealed wide variation in the directionality of disagreement (Fig. 2; Supplementary Figs. 1, 2, 3, 4 and 5), with females having significantly higher positive values (parent > teacher) than males for all SRS scores (Supplementary Table 1). Regressions of informant discrepancy values on the average of informant scores revealed that for females autism trait severity was a significant predictor of informant discrepancy for all domains, except RRBs (Supplementary Table 1). The negative regression coefficients suggest that for females, as informant perception of autism severity increases, informant discrepancy decreases (indicating less disagreement and/or more teacher-reported than parent-reported difficulties).

Fig. 2.

Bland–Altman plots of the informant discrepancy (difference score; parent minus teacher) score regressed on the average of parent/teacher raw total SRS score. The averaged score was a significant predictor of informant discrepancy for females (β = − 0.34, p < 0.001, Adj R² = 0.05), but not for males (β = − 0.006, p = 0.96, Adj R² = −0.005)

Predictors of Informant Discrepancy

Results of the linear regression analyses run for each SRS score discrepancy are detailed in Table 3. These models revealed multiple child characteristics predicted discrepancies between informants: Being female predicted greater disagreement (parent > teacher) on the SRS total score and on four of the five domains (social cognition, social communication, social motivation, and RRBs). Older child age also predicted greater informant discrepancy (parent > teacher) for the SRS total score and all domains. Being non-white was predictive of increased disagreement (teacher > parent) on the social awareness and social communication domains. These models suggest that the included child characteristics account for a small amount of variance (Adj. R² = 0.018–0.061) in informant discrepancy.

Table 3.

Linear Regression Analysis of Informant Differences (Parent – Teacher) on Child Characteristics

	SRS Score
	Raw SRS Total β (SE)	Social Awareness β (SE)	Social Cognition β (SE)	Social Communication β (SE)	Social Motivation β (SE)	RRBs β (SE)
Sex – Female	9.369*** (3.527)	0.678 (0.489)	1.622** (0.727)	3.221** (1.313)	1.815** (0.727)	2.067** (0.874)
Race – Non-white	−6.614 (4.465)	−1.080* (0.617)	−1.204 (0.920)	−3.212* (1.660)	−0.706 (0.920)	−0.438 (1.107)
Age	0.172*** (0.042)	0.011** (0.006)	0.033*** (0.009)	0.063*** (0.016)	0.028*** (0.009)	0.036*** (0.010)
IQ	0.055 (0.066)	0.014 (0.009)	0.017 (0.014)	0.004 (0.025)	0.004 (0.014)	0.016 (0.016)
ADOS CSS	−1.596 (1.063)	−0.052 (0.147)	−0.277 (0.219)	−0.486 (0.395)	−0.364 (0.219)	−0.424 (0.263)
Observations	388	388	387	387	388	388
Adjusted R 2	0.06	0.018	0.055	0.061	0.044	0.048
Residual Std. Error	34.604 (df = 382)	4.780 (df = 382)	7.125 (df = 381)	12.862 (df = 381)	7.132 (df = 382)	8.576 (df = 382)
F Statistic	6.485*** (df = 5; 382)	2.455** (df = 5; 382)	5.524*** (df = 5; 381)	6.006*** (df = 5; 381)	4.558*** (df = 5; 382)	4.934*** (df = 5; 382)

^*p < 0.1;

^**p < 0.05;

^***p < 0.01

Discussion

The purpose of the current study was to test for disparities between informants (parents and teachers) reporting autism characteristics in school-aged autistic males and females. We predicted that any discrepancies between characteristics would be exacerbated in autistic females. Our results support this hypothesis and have important implications for how teachers perceive autistic males and females. Our results also build on a growing body of literature describing how autistic females’ behavior may be camouflaged in a school setting.

Implications of Sex and Informant Differences

Despite no differences in how males and females were rated by parents, teachers rated males as having significantly more autistic traits than females for total and all SRS domain scores. This finding is particularly interesting given the current literature examining sex differences in parent and clinician report, where parents were reporting greater autistic traits in females than males despite a lack of sex differences as reported by clinicians (Dworzynski et al., 2012; Ratto et al., 2018; Ross et al., 2022). Participants in the current study sample did not differ on a measure of clinician-rated autism characteristics, which is in line with results from these previous studies. Although clinician ratings were not directly compared to parent or teacher reports in this study, sex differences found among teachers, but not among parents or clinicians, suggests a gap between teachers and clinicians that may warrant further evaluation in future studies.

In addition to teacher-reported sex differences seen across the sample, we also found significant differences between parent and teacher reports of autistic traits for the same child. However, these differences were only seen for female autistic children: on average, parents of female children rated their child as having significantly more or more intense traits than their teacher did for the SRS total score as well as in the domains of social awareness, social cognition, and RRBs. These significant informant discrepancies were not seen for males in the sample. Previous qualitative evidence from autistic females and their parents (Cook et al., 2018; Tierney et al., 2016), from playground observations (Dean et al., 2017), and from teachers themselves (Whitlock et al., 2020) suggests a potential sex or gender “bias” towards traditional “male” traits and towards children socialized as boys when it comes to reporting autism characteristics. Sex disparities in parent–teacher informant discrepancies in the current study may add empirical support to these claims.

Predictors of Informant Discrepancy

We conducted linear regressions to explore whether child characteristics predicted calculated informant discrepancy values for SRS scores. Age was a significant predictor of informant discrepancy for all SRS scores, such that parents and teachers disagreed more about autism traits as children got older. Being female was also a significant predictor of increased discrepancy for the SRS total score and for social cognition, social communication, social motivation, and RRB domains. Finally, being non-white was a significant predictor of decreased discrepancy in the social awareness and social communication domains. This aligns with the results of studies indicating that mothers of White children endorsed more autism traits than those of Latino children (Blacher et al., 2014), and parents and teachers of ethnically diverse children had significantly higher agreement for social behavior than for other traits (Thompson & Winsler, 2018). Through Bland–Altman (1986) regressions within each sex group, we explored whether the averaged parent and teacher scores predicted informant discrepancy. For females, but not males, a higher average parent–teacher SRS score (indicating more perceived autism severity) predicted less disagreement between informants. This was significant for all scores but the RRB domain for females. The positive association between perceived autism severity and agreement between informants for females aligns with prior research (Azad et al., 2016; Levinson et al., 2021). It is important to note that all models explained a very small amount of variance in informant discrepancy values, suggesting that other factors, such as parent or teacher beliefs or characteristics, may account for the majority of this disagreement.

Extension of Discrepancy-Measured Camouflaging to School Settings

Sex differences among teacher-reported scores as well as informant discrepancies for females in this sample may also be explained by the concept of camouflaging, which describes the ways in which an individual masks or compensates for their autistic traits within social settings (Cook et al., 2021). Although the term camouflaging was not yet in our lexicon at the time of data collection, the rise of camouflage-focused research in recent years has revealed that autistic individuals have been experiencing this phenomenon long before it was formally conceptualized and posited as a potential explanation for why young girls may be overlooked for a diagnosis (Attwood, 2007; Gillberg, 1991; Hull et al., 2019; Lai & Baron-Cohen, 2015; Lai et al., 2017). Camouflaging is particularly important to consider for school-based observations, where children may be monitoring their behavior or may “blend in” with their peers (Dean et al., 2017). Parents in one study attributed parent–teacher differences in reports of autism traits to variations in child behavior between home and school settings (Szatmari et al., 1994). After finding exaggerated parent–teacher discrepancies for females on a measure of child mental health and behavior, Sturrock et al., (2020) suggested that teacher underreporting of females’ difficulties may be due to these difficulties being camouflaged in school. The current study brings further evidence to this theory with discrepancies on a measure specific to autistic traits.

Camouflaging is thought to emerge at a young age (Hull et al., 2021; Jorgenson et al., 2020), however it is difficult to measure in children. Although a measure of self-reported conscious camouflaging is not yet available for children, recent research has operationalized camouflaging in children as the discrepancy between parent-reported autistic traits (considered to be a child’s “true” presentation of autism) and those observed by clinicians (Lai et al., 2017; Ross et al., 2022). Whereas this discrepancy-based approach is unable to capture a child’s awareness of camouflaging, it does illustrate the efficacy of camouflaging, whether it is intentional or not (Cook et al., 2021; Hannon et al., 2023), and may serve as a useful tool particularly within a school context. Indeed, discrepancies found in our results (i.e., higher parent ratings than teacher ratings) can help quantify descriptions of camouflaging from community members. For example, autistic girls have reported copying the interests or mannerisms of others in order to fit in with other girls at school (Cook et al., 2018). Late-diagnosed women and parents of autistic girls have described having their struggles be overlooked or even dismissed by teachers because they were well-behaved or high-achieving (Bargiela et al., 2016; Tierney et al., 2016). Because of differences in how boys and girls socialize at school, an autistic girl’s difficulty engaging with peers may be less noticeable to teachers (Dean et al., 2014, 2017).

Informant differences on a measure of autism traits likely reflect traits and needs that are being camouflaged using conscious effort to succeed in an academic environment, or that may be camouflaged by the environment itself and the gendered expectations embedded in it. In the current study, parents reported their female children as exhibiting significantly greater overall autism traits as well as traits related to social awareness, social cognition, and repetitive behaviors. Many items in these domains are descriptions of behaviors that adolescents and adults in previous studies have described consciously monitoring or reducing when camouflaging around others. For example, autistic individuals have described monitoring their facial expressions (“Expressions on his or her face don’t match what he or she is saying”) and body language (“Has repetitive, odd behaviors such as hand flapping or rocking”), limiting what they talk about (“Thinks or talks about the same thing over and over”), and learning social “rules” such as laughing at jokes they may not understand (“Has a sense of humor, understands jokes”) or following the interests of others (“Focuses his or her attention to where others are looking or listening”) as methods of camouflaging (Livingston et al., 2019, 2020; Tierney et al., 2016).

Domains in which parents and teachers did not significantly differ in their reports for females were social motivation and social communication, although the latter was trending towards significance (p = 0.06). The constructs of social motivation and camouflaging are tightly linked, as many individuals cite a desire for social closeness as a motivation for engaging in camouflaging behaviors (Cook et al., 2018; Livingston et al., 2019). As such, it is unsurprising that social motivation, or lack thereof, would be evident to both parents and teachers. While some items in the social communication domain are linked to common methods of camouflaging, such as forcing oneself to make eye contact (“Avoids eye contact or has unusual eye contact”), the majority may be better indications of a child’s ability to camouflage. For example, the ability to imitate another person (“Is able to imitate others’ actions”) and monitor physical proximity to others (“Knows when he or she is too close to someone or is invading someone’s space”) are both necessary in order to engage in some common camouflaging behaviors. Using the SRS or SRS-2 to identity similarities and differences between parent and teacher observations of behavior may help identify individuals who consciously (or unconsciously) camouflage within the school environment, which in turn could help alleviate long-term consequences of camouflaging, such as heightened depression and anxiety, as well as difficulties accessing services or establishing meaningful relationships.

Other results from this study align with those found in previous studies that operationalize camouflaging: IQ (ranging from 19 to 142 in the current study sample) was not a significant predictor of parent–teacher discrepancy, which is consistent with previous studies (Hull et al., 2021; Lai et al., 2017). Age was a significant predictor of parent–teacher discrepancy, such that older age predicted greater discrepancy, aligning with, to our knowledge, the only study that has measured camouflaging across childhood (Ross et al., 2022). This information is critical given the highlighted need in this body of work to expand current camouflaging knowledge to younger children and those with different intellectual abilities (Petrolini et al., 2023).

Strengths, Limitations, & Future Directions

By leveraging a large, national dataset, we were able to match our sample by both chronological age and cognitive ability. Our sample’s wide range in IQ scores is a particular asset when interpreting these results in the context of current camouflaging literature, which faces a lack of representation of both children and those with below-average cognitive ability (Petrolini et al., 2023). Further, using the sample measure for both informant groups allows for a more direct comparison between reports and eliminates a potential effect of measurement on this discrepancy.

The limitations of the current study highlight the work that should be done by researchers in this area moving forward. Data from this study was drawn from evaluations conducted between 2008 and 2010, with children meeting diagnostic criteria for autism based on DSM-IV. It is important to note that changes between DSM-IV and DSM5, particularly the removal of a language delay requirement for diagnosis and the inclusion of sensory behaviors, may impact the identification and profiles of autistic children, particularly females. However, a recent study found no impact of DSM version on the diagnostic rates for males or females (Harrop et al., 2024) and research using the Childhood Autism Rating Scales indicates stability of diagnosis between DSM-IV and DSM5 (Dawkins et al., 2016). While a strength of the SSC dataset is the focus on simplex families (only one diagnosed individual), thus removing potential bias from prior knowledge of being autistic or having an additional autistic family member, it is unclear whether the parent–teacher discrepancies found in the current study would be different in multiplex families. Further, it is important to consider the advancements in research, public knowledge, and professional training that have occurred since the time of data collection, and how these would affect parent and teacher reports. Although results from this study demonstrate the value of using the SRS or SRS-2 for evaluating informant discrepancies and identifying potential camouflaging, the discrepancies themselves might be diminished as new information regarding autism in females is disseminated. As such, replicating this study would be valuable to evaluate time-based trends in parent–teacher agreement, while also considering factors such as teacher knowledge of autism.

Our regression models revealed that child characteristics (sex, race, age, IQ, clinician-rated autism severity) collectively accounted for a very small amount of variance in informant discrepancy. Future studies should seek to evaluate the role that parent and teacher characteristics, such as knowledge of autism, involvement in an IEP plan, or level of education may impact the reporting of autistic traits. In the current study, we found that decreased parent–teacher discrepancy was associated with reports of increased autism characteristics: Without having access to information regarding whether teachers had autism-specific training, were in a disability-specific classroom setting, or spent more individual time with students, we were not able to determine the degree to which such factors could explain this relationship. Existing research suggests that discrepancies (particularly for females) may be due to some combination of “bias” towards stereotypically “male” presentations (Lockwood Estrin et al., 2021; Whitlock et al., 2020) as well as to the presence of camouflaging in a school setting (Bargiela et al., 2016; Sturrock et al., 2020; Tierney et al., 2016): Including measurements of an informants’ level of training in autism and familiarity with the child could disentangle these predictors of informant discrepancy.

Operationalizing camouflaging through informant discrepancies is a valuable way to observe this phenomenon in data from a time when tools that overtly ask about camouflaging did not exist, such as the Camouflaging Autistic Traits Questionnaire: Parent report (Hannon et al., 2023) or the Modified Questionnaire for Autism Spectrum Condition. However, future research should aim to determine if metrics from such camouflaging-specific tools have a relationship with levels of parent–teacher agreement. Although discrepancy-based measurement of camouflaging has been described as a useful approach for younger children for whom self-report is not an option (Lai et al., 2021), using parent report as a proxy measure of one’s “true” autistic phenotype places limits on determining its construct validity (Williams, 2022). As future research produces tools that can directly measure child-reported camouflaging, integrating such measurements into models assessing informant agreement will be critical.

Due to the relatively small number of non-White participants, we did not examine potential intersecting effects of race and biological sex on reported autism traits, which are established predictors of diagnostic discrepancies (Goldblum et al., 2023; Martin et al., 2023; Wallis et al., 2023). Gender is also an important construct to consider going forward. In this study, we focused on assigned sex at birth based on the availability of data. However, future research should consider how gender, both child identity and parent and teacher gendered views, may intersect with child sex to predict discrepancies in parent and teacher ratings. This is especially important given the intersection of autism and gender diversity (Heylens et al., 2018) and the impact of gender, such as child names, on teacher ratings (Whitlock et al., 2020).

Conclusion

This study examined sex differences and parent–teacher discrepancies on a measure of autism traits. Our observed patterns and predictors of such differences provide new, multifaceted knowledge of how reports of autism traits may be influenced by a child’s age, race, biological sex, and by the person reporting on the traits. Further, parent–teacher discrepancies on this measure may represent a child’s level of camouflaging across school and home contexts, and identifying such discrepancies early on may rectify resulting issues that impact a child’s mental health, academic success, and peer interactions.

Supplementary Material

The online version contains supplementary material available at https://doi.org/10.1007/s10803-024-06498-w.