Relations between everyday executive functioning and language in youth with Down syndrome and youth with autism spectrum disorder

Abstract

Language and executive functioning are major impairments in many neurodevelopmental disorders, but little is known about the relations between these constructs, particularly using parent-report. Thus, the current research sought to examine relations between executive function and language in two groups -- Down syndrome (DS) and autism spectrum disorder (ASD). Forty-one youth with DS (M_age=11.2) and 91 youth with ASD (M_age= 7.7) were included. Results were as follows: in DS, executive function predicted pragmatic, but not structural language, after covarying for age, sex, and social functioning; in ASD, executive function predicted both. Findings highlight the interrelatedness of language and executive functioning and may have implications for intervention development.

Introduction

Language deficits are prominent among children with Down syndrome (DS) and autism spectrum disorder (ASD). However, the nature of language difficulties differs by diagnosis, as do the factors that predict individual differences in language abilities (Rice, Warren, Betz, & 2005). Understanding which abilities are key predictors of language outcomes for children with these two different neurodevelopmental disorders may inform the development of interventions that aim to target key mechanisms underlying language impairments and ultimately quality of life for these youth and their families. The current research focuses on a group of higher-level cognitive processes, referred to as executive function skills, and their relation to two aspects of language in children with diagnoses of DS or ASD – pragmatic (i.e., social language skills) and structural language (i.e., non-social language skills).

DS, the most common genetic cause of intellectual disability, is a disorder that occurs in 1 in 691 live births (Parker et al., 2010). It is caused by the presence of an extra (complete or partial) copy of chromosome 21 and is characterized by global learning difficulties, which worsen over childhood and adolescence (Carr, 2005; Hodapp & Zigler, 1990). While youth with DS are impaired relative to peers of the same chronological age on most aspects of cognitive and social development, some skills fall in line with overall cognitive challenges while others fall below mental age expectations. For example, some aspects of social functioning (Channell et al., 2015), adaptive functioning (Carr, 2012; Loveland & Kelley, & 1988), and visuospatial processing (Næss, Nygaard, Ostad, Dolva, & Lyster, 2016) are commensurate with mental age, while deficits in executive functioning (Lanfranchi, Jerman, Dal Pont, Alberti, & Vianello, 2010; Rowe, Lavender, & Turk, 2006), explicit memory (Martin, Klusek, Estigarribia, & Roberts, 2009), and several aspects of language (Martin et al., 2009) are typically in excess of cognitive limitations (Vicari, 2006). Language functioning, in particular, is described in the literature as a major area of deficit in DS (Sigman et al., 1999), and thus, DS is one of the two disorders investigated in the current study.

ASD is a neurodevelopmental disorder that is characterized by impairments in social and communication skills as well as repetitive and restricted behaviors and interests (American Psychiatric Association [APA], 2013). Diagnosed by behavioral phenotype, ASD occurs in approximately 1 in 59 children in the United States (Centers for Disease Control and Prevention [CDC], 2018). Along with the core deficits of ASD, impairments in adaptive and executive function, attention, sensory processing, and memory are also common (for a review, see Tsatsanis & Powell, 2014). Intellectual abilities are variable in ASD, with about one-third of children presenting with intellectual disability (Christensen, 2016) and some children presenting with average or above average intellectual abilities. Another variable aspect of the ASD phenotype is language functioning. While language deficits are not a criterion for diagnosis, they are prevalent among this population (Kjelgaard & Tager-Flusberg, 2001).

Language in DS and ASD

Multimodal in nature, language comprises many components that involve social (pragmatic) and non-social (structural) linguistic skills (Kennison, 2013). Pragmatics refers to the social aspects of language (Levinson, 1983) and includes the social, emotional, and communicative skills that help relay language meaning. In contrast, structural, or non-social, language refers to the “nuts and bolts” of language and includes abilities such as phonology, semantics, morphology, and syntax.

While youth with DS experience impairments in all aspects of language relative to chronological age expectations, structural language is often more impaired than pragmatic language (Abbeduto et al., 2006). Moreover, many aspects of structural language, such as phonological skills (Roberts, Price, & Malkin, 2007), expressive vocabulary (Chapman, Schwartz, & Bird, 1991; Chapman et al., 1998; McDuffie & Abbeduto, 2009; Miller, 1988), and syntax (Akoglu & Acarlar, 2014; Chapman et al., 1991; Martin et al., 2013) fall below mental age expectations while several aspects of pragmatic language, such as the ability to follow rules of conversation (Beeghly, Weiss-Perry, & Cicchetti, 1990; Roberts et al., 2007), are commensurate with mental age expectations. Given that the social use of language calls upon structural language skills, including syntactic abilities that are a significant weakness in DS (Chapman & Hesketh, 2001; Chapman, 1997; Fowler, 1990), the profile of relatively stronger pragmatic language abilities in DS is somewhat unexpected. Thus, further research is needed to identify what lower-level cognitive skills predict relatively stronger pragmatic abilities in DS.

Language is highly variable within ASD. While some children on the spectrum remain nonverbal throughout their lives (25% of those with ASD; Tager-Flusberg, Paul, & Lord, 2005), others develop structural language abilities that fall within normal limits (Kjelgaard & Tager-Flusberg, 2001; Lord & Paul, 1997; Tager-Flusberg, 2003). Despite heterogeneity in absolute levels of structural language abilities, there appears to be evidence for a ‘profile’ of relative strengths and weaknesses in structural language during the school age period for children with ASD. As reviewed by Boucher (2012), relative strengths (or lower levels of impairment) are observed in articulation abilities and syntax skills; in contrast, relative weaknesses are observed in language comprehension, semantic abilities, and aspects of morphological functioning. Lastly, despite variability in structural language skills in ASD, pragmatic language impairments are a hallmark of the disorder and thus are routinely observed in those on the spectrum (Tager-Flusberg, 2001).

Taken together, the pattern of structural versus pragmatic language abilities differs by neurodevelopmental disorder. Pragmatic language serves as a relative strength in DS and a relative weakness in ASD. The opposite can be said of structural language, a relative strength in ASD and a relative weakness in DS. Given the contrasting language abilities in these two groups, they may serve as model neurodevelopmental disorders for studying the shared and distinct relations between different cognitive abilities and language functioning. This was the goal of the current research.

In particular, we sought to examine relations between executive function and language. Executive function skills or higher level cognitive processes, such as working memory, inhibition, and cognitive flexibility (Miyake et al., 2000), are thought to be crucial for the completion of complex tasks and have been tied to language functioning in studies of children with specific language impairment as well as youth with typical development and those who are deaf (Botting et al., 2017; Kapa, Plante, & Doubleday, 2017). Moreover, intervention research in both typical (Barnett et al., 2008) and atypical populations (Kenworthy et al., 2014) has suggested that social communication skills may improve following interventions that target executive function, suggesting an intimate relationship between executive and language functioning.

Thus, we have framed this study around the following question: does variation in executive function predict meaningful variation in structural and pragmatic language in youth with DS or ASD. Although executive function and language data in the current study were obtained concurrently (and thus, we cannot speak to the directionality of the relationships we observe), we sought to focus on executive function as our predictor variable, as we have hypothesized that variability in different executive function skills may serve as facilitators or bottlenecks to language development. If this is the case, executive function skills may be appropriate alternative targets for intervention to improve language functioning in youth with neurodevelopmental disorders such as DS or ASD.

To date, research on the ties between executive function and language have been limited to investigations of single disorders or have relied upon laboratory-based executive function and language tasks (in most instances). These studies will be reviewed in the sections that follow. Then we will highlight gaps in the existing literature and describe the current research.

Relations between Executive Function and Language in DS and ASD

Few studies have examined the relation between executive function and language in DS; however, the existing literature suggests that further investigation is warranted. For example, Amadó, Serrat, & Vallès-Majoral (2016) found that working memory, cognitive flexibility, and inhibition, all aspects of executive function, were positively correlated with receptive vocabulary, one aspect of structural language. Another study identified syntax, both expressive and receptive, as a linguistic ability related to verbal working memory (Akoglu & Acarlar, 2014), one component of executive function. Finally, in the only study to date to use parent-report, Lee et al. (2017) found relations between overall executive function skills on the Behavior Rating Inventory of Executive Functioning (BRIEF) and general language abilities as measured by the Children’s Communication Checklist −2. They also found ties between BRIEF performance and lab-based measures of pragmatic language.

Similar to DS, the relations between executive function and language among youth with ASD have scarcely been investigated. One aspect of executive function skills, working memory (i.e., the ability to hold information to be processed or manipulated; Hitch & Baddeley, 1976), has been found to be associated with individual differences in the development of language in ASD. Specifically, research has documented a relationship between working memory and discourse comprehension (Schuh, Eigsti, & Mirman, 2016) as well as syntactic comprehension (Akoglu & Acarlar, 2014; Fortunato-Tavares et al., 2015). Language as a predictor of executive function has also been examined. Akbar, Loomis, and Paul (2013) found that structural and pragmatic language abilities (as assessed with direct testing) were related to several everyday executive function skills as measured by parent and teacher ratings on the BRIEF.

Gaps in the Literature and The Current Study

While there is a small body of literature to suggest that examining relations between executive function and language skills in youth with neurodevelopmental disorders such as DS and ASD may be fruitful, there are limitations in the existing research literature that the proposed study seeks to address. First, no research to date has examined relations between parent-reported executive function skills and everyday structural and pragmatic language abilities. Second, prior research has not controlled for other factors that may be contributing to the relation between executive function and language skills, such as overall intellectual abilities, sex, and social abilities. All of these factors relate either to language functioning and/or executive function skills (van der Schuit, Segers, van Balkom, & Verhoeven, 2011; Howe et al., 2015; Losh, Martin, Klusek, Hogan-Brown, & Sideris, 2012) . Thus, the current research sought to examine relations between executive function and both structural and pragmatic language functioning as measured by parent report. In particular, we examined executive function-language relations with and without controlling for the effects of nonverbal IQ, sex, and social abilities. It is hypothesized that executive function will account for significant variance in structural and pragmatic language abilities in youth with DS and those with ASD above and beyond the effects of sex, social abilities, and nonverbal IQ.

Methods

Data were compiled from an ongoing research study of youth with DS at Drexel University and two archival datasets -- one from a study of DS conducted at the National Institute of Mental Health (NIMH) Intramural Research Program and one from studies of youth with ASD that were archived in the National Database of Autism Research (NDAR).

Participants

Our sample consisted of 41 children with DS, 30 of whom were part of studies at the Drexel University and 11 of whom were part of studies at NIMH. Participants from Drexel University were recruited as part of a larger study on reading and language in individuals with intellectual disability. Participants at the NIMH were recruited to be in a study of brain development in DS (Lee et al., 2016). The entire ASD sample of 91 participants was gathered from NDAR. Participants from the NDAR dataset were derived from one of three studies: 1. Autism Genetics study at University of California, Los Angeles; 2. Intrinsic Brain Architecture Sleep study at New York University; and 3. Phenotype and Genetic Factors in Autism study at Massachusetts General Hospital.

Demographic information about the groups can be found in Table 1. As can be seen, the two groups differed on sex distribution (χ²(1) = 21.8, p<.001), with the ASD group having more males as expected. They also differed on age (t(64) = 6.84, p<.001 [df adjusted; equal variance not assumed]), with the ASD group being younger. Furthermore, they differed on full scale IQ (t(126) = −17.8, p<.001) and NVIQ (t(128) = −16.5, p<.001), with the DS group having lower scores in all cases. Because the groups are examined separately, these differences did not present a problem for our analyses and interpretation of results.

Table 1.

Demographic characteristics of DS and ASD groups and performance on language and executive functioning questionnaires

		DS (n=41)	ASD (n=91)	DS v. ASD
Sex	(% males)	39%	80%	χ2(1) = 21.8, p<.001
Age	M(SD)	11.2 (2.9)	7.7 (2.3)	t(64.1) = 6.8, p<.001
	Range	6-16	5-14
FSIQa,b	M(SD)	52.5(11.4)	102.0(15.5)	t(126) = −17.8. p<.001
	Range	24-75	61-145
NVIQc,d	M(SD)	55.1(13.2)	100.0(14.8)	t(128) = −16.5, p<.001
	Range	24-84	56-141
Adaptive Function Social Compositee,f	M(SD)	80.0(12.5)	80.6(14.1)	t(125) = −.25, p>.8
	Range	57-100	53-123
CCC-2 GCC	M(SD)	74.0(14.2)	78.7(17.7)	t(130) = −1.5, p>.1
	Range	49-105	42-134
CCC-2 Structural Language Compositeg	M(SD)	4.9(2.5)	7.1(3.0)	t(130) = −4.0, p<.001
	Range	1-10	1-13.8
CCC-2 Pragmatic Language Compositeg	M(SD)	6.8(2.4)	5.9(2.8)	t(130) = 1.7, p>.9
	Range	1.75-11.25	1.5-14.75
BRIEF/BRIEF-2 GECh	M(SD)	60.4(8.8)	64.3(12.4)	t(105.7) = −2.2, p<.05
	Range	41-77	36-91
BRIEF BRI	M(SD)	58.5(10.5)	64.2(12.8)	t(119) = −2.2, p<.04
	Range	40-82	37-93
BRIEF MCI/CRI	M(SD)	60.1(7.6)	62.4(12.1)	t(116.7) = −1.3, p>.1
	Range	44-77	35-91

Abbreviations: : DS = Down syndrome; ASD = autism spectrum disorder; FSIQ = Full Scale IQ; NVIQ = Nonverbal IQ; CCC-2 GCC = Children’s Communication Checklist – 2^nd ed. General Communication Composite; BRIEF GEC = Behavior Rating Inventory of Executive Functioning Global Executive Composite; BRI = Behavior Regulation Index; MCI = Metacognition Index; M = mean; SD = standard deviation; M = mean; SD = standard deviation; FSIQ, NVIQ, and GCC are standard score means (normative M = 100, SD = 15). p<.05 are indicated in bold font

Note:

^an=39 for DS

^bn=89 for ASD

^cn=40 for DS

^dn=90

^en=37 for DS

^fn=90

^gComposite calculated based on mean of scaled scores (normative mean = 10; SD = 3) for scales A-D for Structural Language and E-H for Pragmatic Language

^hT-score (normative mean = 50; SD=10; higher scores denote greater difficulties).

Measures

Language Functioning

The Children’s Communication Checklist- Second Edition (CCC-2; Bishop, 2003) was used to evaluate structural and pragmatic language skills. It is a parent-report questionnaire assessing everyday language skills among children ages 4 to 16 years, 11 months. Test-retest reliability ranges between .86 and .96, while internal consistency ranges between .69 and .85. The measure consists of 70 items within ten scales: A. Speech, B. Syntax, C. Semantics, D. Coherence, E. Initiation, F. Scripted Language, G. Context, H. Nonverbal Communication, I. Social Relations, and J. Interests. Speech, Syntax, Semantics, and Coherence (A-D) are scales that assess aspects of structural language. Initiation, Scripted Language, Context, and Nonverbal Communication (E-H) measure aspects of pragmatic language. Finally, the last two scales of the CCC-2, Social Relations and Interests (I-J), measure nonverbal behaviors associated with ASD. For the purposes of the current research study, the first eight scales were analyzed to characterize structural and pragmatic language in all of our participants with DS and ASD. We did not include scales I and J. Composites for structural and pragmatic language were created by averaging scaled scores (mean=10; SD=3) across scales A-D and E-H, respectively, consistent with methods in Lee et al. (2012). The General Communication Composite (GCC) is a standard score that summarizes overall language skills.

The CCC-2 is intended for children who speak in sentences (>2 word utterances) and do not have permanent hearing loss. For the DS sample (for which we engaged in original data collection), participants whose parents endorsed that they used sentences on the CCC-2 and/or indicated that they used phrases of 3 or more words together when speaking during a parent screening interview were included. In addition, one participant with DS with inconsistent (infrequent) sentence/ 3-word phrase use was included. Analyses were run with and without this participant and results were largely the same. There were six participants for whom parents indicated that they had some hearing loss; however, none of these hearing difficulties impacted neuropsychological testing. Given how common hearing problems are in children with DS (~25%; Nightengale, Yoon, Wolter-Warmerdam, Daniels, & Hickey, 2017) and the inability to determine if the hearing loss was ‘permanent,’ we elected to include these participants. However, analyses were run with and without these participants and results remained largely the same (if not were strengthened for structural language analyses with these participants excluded). Because of the consistency in findings and the small sample size of the DS group, we included all participants with DS in primary study analyses. For the ASD sample, which originated from the NDAR data repository, such detailed information was not available; thus, we included all participants with scored CCC-2 data available in the NDAR database that met other study criteria.

Executive Functioning

Because data were compiled from multiple studies, two versions of the Behavior Rating Inventory of Executive Function (parent report) were included in the current study -- the Behavior Rating Inventory of Executive Function (BRIEF; Gioia, Isquith, Guy, & Kenworthy, 2000) and the recently updated BRIEF-2 (Gioia, Isquith, Guy, & Kenworthy, 2015), both of which are normed for individuals ages 5 to 18. Parents of participants with DS (n=11) seen at NIMH and participants with ASD (n=91) that were a part of the NDAR sample completed the BRIEF. The BRIEF-2 was completed by parents of participants with DS (n=30) seen at Drexel University. The BRIEF and BRIEF-2 have high internal consistency coefficients (mid to upper .90s) and test-retest reliability (mid to upper .80s).

The Global Executive Composite on the BRIEF was the primary independent variable in our analyses predicting structural and pragmatic language. This measure is highly correlated across the instruments (r=.97). In addition, 74% of the questions are the same across the BRIEF and BRIEF-2. For the most part, the BRIEF-2 represents a shortening of the BRIEF (i.e., 18 items were not included in the BRIEF-2 that were a part of the BRIEF), with the exception of the Inhibit, Shift, and Organization of Materials scales, in which 7 total new items were added to the BRIEF-2 that were not present on the BRIEF. Despite these differences, the correlation between the GEC on the two instruments is exceptionally high and actually higher than the test-retest reliability of the instrument. Thus, to increase our sample size in the DS group, we utilized the GEC score from the BRIEF and BRIEF-2 in our primary analyses.

For exploratory analyses in which we were interested in understanding more about how cognitive and behavioral executive dysfunction related to language scores, we elected to use only one version of the instrument in the DS group – the BRIEF-2. This eliminated the 11 children with the original BRIEF for supplementary analyses. We did this because the factor structure of the BRIEF changed from the BRIEF to BRIEF-2. Specifically, the BRIEF-2 has three factors while the original BRIEF had only two. Thus, we felt that for these analyses, it was more conservative to eliminate the participants with DS who had the original BRIEF. As such, for exploratory analyses, the BRIEF-2 Behavior Regulation Index and Cognitive Regulation Index served as the primary variables of interest in the DS subgroup included in these analyses (n=30) while the BRIEF Behavior Regulation Index and Metacognition Index served as the two primary indices of interest in the ASD group (n=91).

Social Abilities

Again, compilation of data across multiple sites necessitated the inclusion of multiple measures of social abilities, a covariate in our analyses. Parents of individuals in the DS and ASD samples completed either one of two measures of social functioning – the Vineland Adaptive Behavior Scale-Second Edition Parent/Caregiver Rating Form (Drexel University DS sample and NDAR ASD sample) or the Adaptive Behavior Assessment System – Second Edition Social composite (NIMH DS sample) to evaluate social abilities. Research comparing the two instruments reports a moderate correlation between the social domains (.60; Sparrow, Cicchetti, & Balla, 2008), which are named the Socialization domain and Social Composite domain for the Vineland Adaptive Behavior Scale and the Adaptive Behavior Assessment System, respectively. See Supplemental Material for additional details.

Intellectual Ability

While not the focus of the current research, an estimate of NVIQ was desired to (1) characterize the nonverbal cognitive abilities of the samples and (2) to be used as a covariate in analyses. We refrained from covarying for overall IQ, which encompasses verbal and nonverbal cognitive abilities, given our focus on language abilities as the dependent variable of interest. Again, because data derived for this sample was archival, multiple standardized measures of IQ were utilized in this study. These included Kaufman Brief Intelligence Test-Second edition, the Differential Abilities Scales – Second Edition, Wechsler Abbreviated Scale of Intelligence-Second edition or the Wechsler Preschool and Primary Scale of Intelligence-4th edition. See Supplemental Material for additional details.

Statistical Analyses

Prior to conducting primary statistical analyses, descriptive statistics were obtained for the two groups and evaluated for the presence of outliers and deviations from normality. To evaluate the contribution of executive function abilities to structural and pragmatic language, separate parallel regression analyses were completed for the DS and ASD groups. Specifically, four sets of hierarchical multiple regression analyses were completed for each disorder. For each regression equation, structural or pragmatic language scores served as the dependent variable, while executive function served as the independent variable. In order to examine executive function’s contribution to language skills above and beyond the effects of sex, NVIQ, and social abilities, hierarchal regressions were conducted in which these three variables served as covariates that were entered into the model in Step 1 while executive function was entered in Step 2.

Because the ASD group had a significantly higher IQ than the DS group and research suggests that language and nonverbal intellectual abilities are correlated in ASD (Thurm, Lord, Lee, & Newschaffer, 2007), we also sought to examine relations between executive function and language in youth with ASD with differing intellectual ability levels. Thus, we split the ASD group based on whether their nonverbal IQ was above or below the group’s median – 102. Regression analyses were then conducted on both halves of the ASD group – high and low IQ – separately.

Lastly, in order to explore executive function-language relations further, two multiple regression analyses per diagnostic group (one for structural language and one for pragmatic language) were completed in which the Behavior Regulation Index and the Metacognitive/Cognitive Regulation Index were examined for their relative contribution to structural and pragmatic language in both diagnostic groups.

Results

Prior to addressing primary study questions, the performance of the DS and ASD groups on the two measures of interest was examined. Means and SDs for the CCC-2 composite scores and the BRIEF Global Executive Composite are provided in Table 1. The groups did not differ on general language abilities (t(130) = −1.49, p>.05) or pragmatic language abilities (t(130) = 1.66, p>.05). However, they did differ on structural language (t(130) = −4.04, p<.001), consistent with prior research documenting that this is an area of significant challenge for youth with DS (Abbeduto et al., 2006). On the BRIEF Global Executive Composite, both groups scored just below the clinical cutoff of T= 65; their scores on this measure did not differ. Lastly, for descriptive purposes, we evaluated correlations among all variables within each group. These correlations are presented in Table 2.

Table 2.

Correlations among study measures for participants with ASD (above the diagonal) and DS (below the diagonal)

	NVIQ	Social Composite	BRIEF GEC	CCC-2 Structural Language	CCC-2 Pragmatic Language
NVIQ	╲	.08	−.05	.31**	.20
Social Composite	.40*	╲	−.52**	.41**	.61**
BRIEF GEC	−.24	−.51**	╲	−.38**	−.59**
CCC-2 Structural Language	.27	.39*	−.46**	╲	.74**
CCC-2 Pragmatic Language	.43**	.45**	−.66**	.69**	╲

Abbreviations: NVIQ = nonverbal IQ; BRIEF GEC = Behavior Rating Inventory of Executive Functioning Global Executive Composite; CCC-2 = Children’s Communication Checklist – 2^nd ed.

^*indicates p<.05

^**indicates p<.01

Question 1a: Is executive function a concurrent predictor/correlate of language in DS?

In the absence of covariates, GEC contributed unique variance to structural language in DS (p < .01). When covariates were entered in the model, the model was significant (p < .01). However, investigation of individual predictor variables revealed that (1) sex, NVIQ, and social abilities (entered in one step) were significantly predictive of structural language (p<.01), but (2) executive function added no additional variance to the prediction of structural language (p>.05) after accounting for step 1.

In evaluating pragmatic language, GEC alone in the model contributed significant unique variance (p<.001). When covariates were entered, the model remained significant (p<.01). In this case, both steps in the model were significant: (1) sex, NVIQ, social abilities (p<.05), and (2) executive function (p<.01) added unique variance. Table 3 displays R, R² change, and F-statistics for these models.

Table 3.

Results of Hierarchical Multiple Regression for DS and ASD samples

	Independent variables		Dependent variables
		df^	Structural Language				Pragmatic Language
			R	R2 change	F change	p	R	R2 change	F change	p
EF entered in model without covariates:
DS n=41	EF	1,39	.46	.22	10.7	<.01	.67	.44	30.9	<.001
ASD n=91	EF	1,89	.38	.15	15.3	<.001	.59	.35	47.3	<.001
EF entered in model with covariates:
	STEPS:
DS n=36	1) Sex, NVIQ, Social abilities	3,32	.61	.37	6.2	<.01	.54	.29	4.34	<.02
	2) EF	1,31	.65	.06	3.2	<.1	.73	.24	15.9	<.01
	Total Model	R2 =.43, F[4,31] = 5.8, p<.01					R2 = .53; F[4,31] = 8.8, p<.01
ASD n=89	1) Sex, NVIQ, Social abilities	3,85	.49	.24	9.13	<.001	.64	.41	19.28	<.001
	2) EF	1,84	.53	.04	4.6	<.05	.72	.11	18.68	<.001
	Total Model	R2 = .28, F[4,84] = 8.27, p<.001					R2 = .51, F[4,84] = 22.14, p<.001

NVIQ = nonverbal IQ; EF = executive functioning as measured by the GEC; Bold font indicates significant findings; Note: sample size is reduced after covariates are added to reflect missing data on NVIQ and Social abilities

^{^}indicates the df refers to the change statistic

Question 1b: Is executive function a concurrent predictor/correlate of language in ASD?

GEC accounted for significant variance in structural language when entered as the sole independent variable (p<.001). The model was also significant with sex, NVIQ, social abilities (step 1) entered first, and executive function (step 2) entered second into the model (p<.001). Executive function, entered in Step 2 (p<.05), significantly predicted structural language scores above and beyond the effects of sex, NVIQ, and social abilities (entered in step 1, p<.001).

For pragmatic language, GEC contributed unique variance in the model when entered alone (p<.001). The model was also significant overall (p<.001) when covariates were included. Individually, (1) sex, NVIQ, social abilities (p<.001) and (2) executive function (p<.001) predicted variance in pragmatic language. Table 3 displays R, R² change, and F-statistics for these models.

Because of differences in the IQ of youth with DS and ASD and the wide nonverbal IQ range present in the ASD group (standard scores = 61-145) in the current research, we sought to examine executive function-language relations in subgroups of children with ASD – those with nonverbal IQ scores above and below the sample median score of 102. The same analyses described above for the whole group were also conducted on both halves of the ASD group. Among those with an IQ <102, GEC did not contribute unique variance to structural language when entered in the absence of covariates (p>.05). In addition, the overall model failed to reach statistical significance after sex, NVIQ, and social abilities were included in Step 1 and executive function was entered in Step 2 (p=.05). In contrast, executive function contributed unique variance to pragmatic language when entered in the model alone (p<.01). However, while the model for pragmatic language was significant (p <.05) with the covariates and executive function included in the model, neither (1) sex, NVIQ, social abilities (p>.05), nor (2) executive function (p >.05) predicted unique variance in pragmatic language skills.

Among those with an IQ ≥ 102, GEC significantly predicted structural language (p<.001). Next, in the evaluation of covariates and executive function, the structural language model containing all variables was significant (p<.001). (1) Sex, NVIQ, social abilities, (p <.01) and (2) executive function (p<.001) predicted unique variance in structural language. Similarly, GEC significantly predicted pragmatic language (p<.001). With covariates, the overall pragmatic language model continued to be significant (p<.001), with (1) sex, NVIQ, social abilities, (p < .001), and (2) executive function (p<.001) significantly predictive of pragmatic language. Results are reported in Table 4.

Table 4.

Results of Hierarchical Multiple Regression for ASD group split by median IQ

	Independent variables		Dependent variables
			Structural Language				Pragmatic Language
		df^	R	R2 change	F change	p	R	R2 change	F change	p
EF entered in model without covariates:
ASD IQ: 61-101 n=44	EF	1,42	.14	.02	.82	>.3	.40	.16	8.09	<.01
ASD IQ: 102-145 n=45	EF	1,43	.67	.44	34.4	<.001	.69	.48	39.0	<.001
EF entered in model with covariates:
	STEPS:
ASD IQ: 61-101	1) Sex, NVIQ, Social abilities	3,40	.14	.02	.28	>.8	.41	.17	2.68	<.1
	2) EF	1,39	.26	.05	1.96	>.17	.48	.06	3.28	<.1
	Total Model	R2 =.07, F[4,39] = .7, p>.5					R2 = .23; F[4,39] = 2.94, p<.05
ASD IQ: 102-145	1) Sex, NVIQ, Social abilities	3,40	.56	.31	6.01	<.01	.73	.53	15.02	<.001
	2) EF	1,39	.71	.19	15.02	<.001	.82	.14	15.83	<.001
	Total Model	R2 = .50, F[4,39] = 9.84, p<.001					R2 = .67, F[4,39] = 19.4, p<.001

NVIQ = nonverbal IQ; EF = executive functioning; Bold font indicates significant findings

^{^}indicates the df refers to the change statistic

Question 2a: Are there differential relations between two domains of executive function – Behavioral Regulation and Cognitive Regulation– and the two overarching domains of language in DS?

For this exploratory aim, two linear multiple regressions were conducted, with the Behavior Regulation and Cognitive Regulation Indices entered together as independent variables and either structural or pragmatic language composites as the dependent variables.

The overall model for structural language was not significant (R² = .15, F[2,27] = 2.4, p>.05); neither the Behavior Regulation Index (β= −0.06, p>.05) nor Cognitive Regulation Index (β = −0.35, p>.05) independently contributed significant variance to structural language. The results of the regression model with pragmatic language as the dependent variable indicated that the two predictors collectively accounted for significant variance in pragmatic language (R² = .35, F[2,27] = 6.7, p<.01). There was a trend for the Behavior Regulation (β = −0.41, p =.06) but not the Cognitive Regulation Index (β = −.22, p > .3) to predict pragmatic language. Table 5 summarizes β coefficients, t-tests, and p-values for the exploratory analyses.

Table 5.

Differential relations between domains of EF and Structural & Pragmatic Language for DS and ASD samples

		Structural Language				Pragmatic Language
		ß	t	p		ß	t	p
Question 1
DS n=30	CRI	−.35	−1.5	>.1		−.22	−1.03	>.3
	BRI	−.06	−.23	>.8		−.41	−1.9	.06
	Total Model	R2 = .15, F[2,27] = 2.4, p > .09				R2 = .33, F[2,27] = 6.7, p=.004
Question 2
ASD n=91	MCI	−.43	−3.1	<.01		−.26	−2.1	<.05
	BRI	.03	.21	>.8		−.39	−3.2	<.01
	Total Model	R2 = .17, F[2,88] = 8.8, p<.001				R2 = .36, F[2,88] = 24.67, p<.001

Bold font indicates significant findings; CRI=Cognitive Regulation Index [BRIEF-2]; MCI=Metacognition Index [BRIEF]; BRI=Behavior Regulation Index [BRIEF & BRIEF-2].

Question 2b: Are there differential relations between the two domains of executive function – Behavioral Regulation Index and Metacognition Index – and the two overarching domains of language in ASD?

Another two multiple regressions, with either of the language composites as dependent variables, measured the contributions of the Behavior Regulation and Metacognition indices to language functioning in ASD. The regression model predicting structural language was significant in ASD (R² = .17, F[2,88] = 8.76, p<.001). Whereas the Metacognition Index was significantly associated with structural language in ASD (β = −0.43, p <.01), the Behavior Regulation Index was not (β = 0.03, p>.05). The results of the regression model predicting pragmatic language indicated that the two predictors collectively contributed significant variance (R² = .36, F[2,88] = 24.67, p<.001) and that, individually, both the Behavior Regulation Index (β = −0.39, p <.01) and the Metacognition Index (β = −0.26, p<.05) significantly contributed variance to the prediction of pragmatic language. See Table 5 for β coefficients, t-tests, and p-values for these analyses.

Discussion

The purpose of the current study was to examine the contribution of executive function to variation in structural and pragmatic language functioning in youth with DS and ASD, as this is a topic that has not been studied and has implications for identifying alternative treatment targets to augment communication abilities in these two groups. Analyses evaluated whether executive function predicted unique variance above and beyond sex, NVIQ, and social abilities in youth with DS and those with ASD. In the sections that follow, we briefly review study findings and how they fit with the existing literature for each disorder.

Relations among language and executive function in DS

The first study question examined executive function abilities in relation to structural and pragmatic language in youth with DS. Within this sample, executive function did not significantly contribute unique variance to structural language, but did uniquely predict variance in pragmatic language, after sex, NVIQ, and social abilities were accounted for. Studies have documented a relation between executive function and aspects of structural language in DS (Akoglu & Acarlar, 2014; Amadó et al., 2016). The results of the current study are consistent with these prior studies in the absence of covariates; however, relations between executive function and structural language did not survive covariates in the current research. With regard to pragmatic language, the current finding is consistent with what has been found in the literature (Lee et al., 2017). This finding makes intuitive sense, as the ability to communicate with others in social situations (i.e. to utilize pragmatic language) taps into aspects of executive function skills, such as following the rules of conversation (Martin & McDonald, 2003), shifting and maintaining topics (Humphries et al., 1994), and monitoring and regulating one’s behavior (McDonald, 1993). These results highlight the importance of executive function skills to everyday pragmatic language in DS.

Relations among language and executive function in ASD

In Question 1b, we examined executive function in relation to language within individuals with ASD. Executive function significantly predicted variance in both domains of language with and without sex, NVIQ, and social abilities accounted for. This relationship is supported in the literature by Akbar et al. (2013), who found an association between the Global Executive Composite of the BRIEF and direct measures of structural and pragmatic language in ASD. Our study goes further, though, by examining the contribution of executive function in excess of other predictor variables and doing so using parent report of both executive function and language skills. Despite the social deficits characteristic of ASD, executive function contributed significant unique variability in language above and beyond social functioning (as well as sex and NVIQ). These findings suggest that components of executive function may meaningfully contribute to pragmatic language difficulties (or vice versa) and shape the way pragmatic language develops in ASD.

Relations between the two domains of executive function and the two domains of language in DS and ASD

The exploratory analyses of the current study aimed to investigate differential relations between the two domains of language and the two domains of executive function in DS and ASD, with group analyses conducted in parallel. In particular, we examined the differential roles of both the Behavior Regulation Index and the Cognitive Regulation/Metacognition Index as they correlate with structural and pragmatic language in DS and ASD. Contrary to expectations, we did not find relations between either executive function domain and structural language in DS. Limited power was certainly a contributing factor to this null finding. In the case of pragmatic language, the Behavior Regulation Index but not the Cognitive Regulation index was marginally predictive (p=.06) in DS. The Behavior Regulation Index measures the ability to inhibit one’s responses, shift between cognitive sets, exert emotional regulation, and monitor one’s own behaviors. Collectively, these skills contribute to the ability to maintain and regulate one’s own behaviors, two important tools required for social relationships (Eisenberg, Fabes, Guthrie, & Reiser, 2000) and thus for social communication.

Analyses investigating relations between executive function and language in ASD revealed a significant relationship between the Metacognition Index -- but not Behavior Regulation Index -- and structural language. The Metacognition Index captures the aspects of executive function that involve initiating, planning and organizing, and sustaining working memory. Because structural language encompasses skills that require holding information online for comprehension (syntax and semantic skills) and planning/organizing speech output to compose a cohesive narrative (cohesion), it is not surprising that it calls upon metacognitive executive function abilities. On the other hand, examination of pragmatic language revealed no differential relations among the two domains of language. Both the Metacognition Index and Behavior Regulation Index significantly contributed variance to pragmatic language in ASD. Thus, while specific Metacognition Index-structural language relations in ASD exist, both domains of executive function are broadly implicated in pragmatic language. These findings suggest that pragmatic language abilities may rely on greater executive function resources than structural language in ASD and thus could explain some of the variance between abilities across domains.

Clinical Implications

The current study’s findings have implications for tailoring clinical interventions for language skills among these populations, as they suggest that the neurocognitive correlates of language vary (to some degree) by language domain, diagnostic group, and perhaps nonverbal intellectual ability levels. Understanding such relationships can inform intervention development for these groups. Findings from the current study suggest that executive function may play a more substantial role in structural language among youth with ASD than among youth with DS. This may be attributed, in part, to the lower intellectual ability level of our DS group, as structural language may call less upon executive function abilities among individuals with lower IQ. This interpretation is consistent with the fact that we did not find a significant relationship between executive function and structural language among those with an IQ < 102 in the ASD group, while a significant relationship was found for those with an IQ ≥ 102. Further research is needed with youth with heterogeneous intellectual ability levels and diagnoses in order to explore how intellectual ability levels, structural language, and executive function are related within and across diagnostic groups.

As executive function abilities were related to pragmatic language in both groups, the current study’s findings suggest that interventions targeting executive function abilities may have cascading effects on pragmatic language outcomes in youth with these (and perhaps other) neurodevelopmental disorders. An example of such an intervention is Unstuck and on Target (Cannon, Kenworthy, Alexander, Werner, & Anthony, 2011), which was designed to improve executive function in verbally fluent children with ASD. As stated in the Introduction, this intervention did have cascading positive effects on social communication skills in children with ASD. Future research on this and similar executive function interventions would benefit from evaluating their efficacy in improving executive function and augmenting social communication skills in groups with more limited structural language abilities, such as children with DS or youth with ASD with comorbid intellectual disability.

Limitations & Future Directions

The current research study has several limitations. First, the use of three different samples to obtain a sample size of 132 total participants prevented us from obtaining data on IQ, social, and executive function using the same measures within and across groups. We minimized cross-sampling between the two BRIEF versions in the exploratory analyses by constraining our DS sample to those with only BRIEF-2 data. We could not, however, extend these restrictions to the primary analyses for the DS group, as we were statistically underpowered to answer the primary study question with the smaller subsample of participants with the BRIEF-2 alone. However, given the exceptionally high correlation between the GEC on the BRIEF and BRIEF-2, we believe that including youth with DS with the two BRIEF versions was justified for primary analyses.

Along the same lines, because our data were compiled from three different studies, we were unable to obtain a sample of individuals with ASD and DS with similar intellectual ability levels and ages. As such, analyses were conducted in parallel and provided information about how executive function and language skills relate to one another within rather than across these two diagnostic groups. Future research would benefit from utilizing a sample of youth with DS and ASD with similar intellectual ability levels in order to further evaluate whether executive function-structural language relations in these groups.

Another limitation on this study was the limited statistical power observed for some analyses involving the DS group. This can be remedied in future research with larger sample sizes, perhaps by combining samples across multiple sites. Additionally, the use of a cross-sectional design to understand the relationships between language and executive function is a limitation of this study, as we cannot speak to the directionality of these findings. Thus, a next step for future research will involve adopting a longitudinal study design and examining the reciprocal relations between executive function and language functioning over time. Finally, intervention research that aims to target one domain but has cascading effects on other domains of functioning may also inform our understanding of the directional relationship among these variables and possibly help prioritize interventions that not only improve the targeted cognitive domain but related domains of functioning as well.