Language/cognitive predictors of literacy skills in 12-year-old children on the autism spectrum

Abstract

Aim

Some school age children with autism spectrum disorder (ASD) struggle with literacy development, yet, the individual differences are major and not well understood. Moreover, literacy attainment is multi-faceted and literacy skills and difficulties manifest in more than one way. The aim of this study was to describe this variability and to identify language/cognitive predictors of different literacy skills.

Methods

We assessed different literacy skills (word reading accuracy, reading fluency, reading comprehension, and spelling), along with a set of language/cognitive predictor skills (listening comprehension, rapid automatized naming, phonological awareness and nonverbal cognitive ability), in 12-year-old children with ASD without intellectual disability recruited from a longitudinal study in Sweden.

Results

There was great heterogeneity (from floor to ceiling levels) in literacy skills, with a statistically increased prevalence of poor reading comprehension and reading fluency compared with population norms. In regression analyses, it was shown that concurrent language/cognitive predictor skills (e.g. phonological awareness) were differentially associated with literacy subskills. Moreover, a longitudinal analysis showed that preschool language problems were associated with poor word reading accuracy and spelling in middle school.

Conclusions

The results confirm previous findings as well as provide new knowledge regarding profiles of literacy (difficulties) in children with ASD; interestingly, the identified predictors of literacy skills in ASD resembled those identified as important in general reading (and dyslexia) research, which might indicate that similar kinds of support and training would be beneficial.

Keywords:

• Autism
• reading
• listening comprehension
• learning
• spelling
• phonological awareness

Introduction

Autism Spectrum Disorder (ASD) is a behaviorally defined clinical condition characterized by the combination of marked difficulties with social-communicative functioning and markedly reduced flexibility in behaviors and interests (DSM-5) [1]. It is now well recognized that people who are diagnosed with ASD very rarely “only” have the specific constellation of symptoms that defines ASD. Hence, it has been proposed that children with ASD need to be assessed, treated and followed-up not only with a focus on the autistic symptoms per se but broadly and multi-professionally in order for the specific needs of each individual child to be met [2]. The current paper addresses literacy skills in school age children with an early diagnosis of ASD without intellectual disability (ID). A focus on literacy skills is justified given its importance in order for the individual to function independently in educational, leisure, and future work life settings.

General research into the reading capacities of children and adults has shown that literacy skills can be divided into separable but correlated subskills, with partly unique cognitive/language correlates [3,4]. Word reading refers to the ability to recognize written words, and can be assessed in terms of accuracy and fluency. Reading comprehension refers to the extraction of meaning from written text and is often assessed by having children read texts and answer comprehension questions independently. Finally, writing skills are included under the literacy umbrella, with proficiency in spelling constituting one central capacity [5]. Impairments in any of these skills are important to identify since once identified they can be ameliorated or compensated for through specific techniques [6,7]. Also, word reading accuracy, reading fluency, reading comprehension, and spelling should all be specified in making a formal diagnosis of Specific Learning Disorder according to the DSM-5¹, further underscoring their relevance in practical (clinical and educational) settings.

Children with ASD symptoms have been included in several recent studies of literacy performance. Most studies have noted a very considerable heterogeneity in reading capacity among children with ASD [8–11] in much the same way as language and cognitive abilities more generally have been shown to vary in this population [12]. Here, we aim to describe the variability in literacy skills in a population cohort of Swedish 12-year-old children with an early diagnosis of ASD without ID by describing both average performance and cases with impaired literacy. Also, we particularly aim to shed light on the nature of the variability in word reading accuracy, reading fluency, reading comprehension, and spelling by identifying concurrent language/cognitive predictors for each subskill. Guided by findings in general research on literacy [3–5,7] and employing a regression analytical approach, we wanted to determine if we could predict literacy subskills from the following language/cognitive predictor skills: listening comprehension, rapid automatized naming (RAN), phonological awareness, and nonverbal cognitive ability. This has been done separately in prior studies that have focused on one or two literacy subskills [8,9,11,13,14]. The present study contributes to the current literature by including several literacy outcomes in one age-homogenous sample and using similar assessment formats. These features make it possible to more systematically identify whether there are certain language/cognitive capacities that uniquely predict different literacy subskills. From a practical point of view, it is important to understand why some children with autism excel in literacy while others struggle greatly, in order to help everyone develop their full potential. From a theoretical point of view, it is continuously important to know to what extent general models of literacy ability hold for those on the autism spectrum. The ASD cohort in this study has been followed by the research group since the children were aged 2–4 years [12,15]. In Kjellmer et al. [16], we reported on the language ability of the children with ASD but without ID when they were aged 4–6 years and found that almost 60% had moderate to severe structural language problems, defined as impaired performance on two or more of a total of six oral language measures. At the current follow up for literacy assessment, the children had reached middle school age. Although our main focus in this new study was to examine the variability in literacy skills at age 12 years as well as to examine which specific language/cognitive skills that concurrently predict aspects of literacy in this population, we also examined how language problems identified at preschool age (according to [16]) relate to later literacy attainment.

Methods

The current study group constituted a subgroup from a larger cohort of children with ASD included in a population-based longitudinal research project. The larger cohort was a community-representative group of 208 children (176 boys and 32 girls), birth-years 2002–2006, aged 2.5–4.5 years at first assessment, Time 1 (T1). Two years later, 198 of these 208 children were included in a comprehensive clinical follow-up, Time 2 (T2) [12,15]. The children and their parents had met with a multi-professional clinically experienced research group at T1 and T2. At the 2-year follow-up, at ages 4.5–6.5 years, 107 children met criteria for autistic disorder, 70 for other ASDs (PDD NOS or Asperger syndrome), and 21 did not meet full criteria for any ASD, but had autistic symptoms plus various other developmental deviations in areas of speech and language, attention-activity regulation, and behavioural problems. In the total T2 group, intellectual levels varied with 95 children having ID (and two children were not assessed). The remaining children either had an average intellectual level, i.e. an IQ of 85 or above, or a borderline intellectual function, in the IQ area of 70–84. Thus, the included children had variable intellectual levels, some had cognitive delays, but at the time of school entry none of the children had a diagnosed intellectual disability. At the time for the current follow up, none of the participating children attended the special school form available for children with ID in Sweden (compulsory school for children with intellectual disability) according parental reports. At T2, the speech and language pathologists in the research team performed a comprehensive language assessment of the children without ID. Of those 101 children, parents of 7 declined participation and 11 children were excluded since Swedish was not their first language, which means that 83 children participated in that part of the T2 assessment, as reported in Kjellmer et al. [16].

Over a 3-year period, 70 children (out of the 83) who subsequently had reached the age of 12 years (born between February 2002 and February 2005) were invited to a follow-up assessment of language/cognitive abilities and literacy. Of these 70 children, parents of 43 (40 boys, 3 girls) accepted to participate with their child, but two of the children never contributed with any literacy data during the assessment. Parents of 27 children did not want their child to participate or could not be reached. Hence, the present study contains literacy and language/cognitive skills data from 41 children with ASD without ID.

The participants were assessed with a battery of tests covering different literacy and language/cognitive skills (Table 1). The tests were administered in a fixed order. The full examination of each child took approximately three hours including a break during which refreshments were offered to the participants.

Table 1. Test materials used in the study.

Literacy skills	Test name and description
Word reading (orthographic) accuracy	Sight decoding subtest from ITPA-3 [17]. The test entails reading a list of printed words. Scores are presented in age adjusted percentiles.
Reading fluency	Text reading fluency subtest from LOGOS [18]. The test entails reading connected text as swiftly as possible. Scores are presented in age-adjusted percentiles.
Reading comprehension	Reading comprehension subtest from LOGOS [18]. The test entails reading connected text and answering comprehension questions. Scores are presented in age-adjusted percentiles.
Spelling	Spelling subtests from DLS [19] for grades 4–6 and from LS [20] for grade 7. The tests entail spelling dictation. Scores are presented in age-adjusted stanines (M = 5, SD = 2).
Language/cognitive predictor skills
Listening comprehension	Listening comprehension subtest from LOGOS [18]. The test entails listening to connected text and answering comprehension questions. Scores are presented in age-adjusted percentiles.
Rapid automatized naming (RAN)	Rapid automatized naming (RAN) subtest from LOGOS [18]. The test entails naming serially presented familiar objects as swiftly as possible. Scores are presented in age-adjusted percentiles.
Phonological awareness	Sound deletion subtest from ITPA-3 [17]. The test entails deleting syllables and phonemes from words. Scores are presented in age-adjusted percentiles.
Nonverbal intelligence	Raven’s Progressive Matrices [21]. The test entails completing a visual pattern based on logical relations. Scores are presented in age-adjusted percentiles.

Descriptive statistics were used to describe average performance. We defined literacy problems using a cut off of ≤11th percentile or a stanine score of ≤2. To examine whether the prevalence of literacy problems in our group was increased compared to what would be expected, we calculated 95% confidence intervals (CI; http://vassarstats.net/prop1.html) around the obtained proportion of children with a result below the set cut-off for each test. If the lower limit of the CI was above the defined cut-off for literacy problems, the rate of children with problems was considered statistically increased relative to population norms. Second, in order to identify if, and which, language/cognitive skills that predict variability in literacy skills, we conducted four sets of linear regression analyses (using the Enter method) for each of the outcome measures, i.e. word reading accuracy, reading fluency, reading comprehension, and spelling. Finally, non-parametric Mann–Whitney U tests were used to compare children with moderate to severe and mild or no preschool oral language problems, respectively, on the literacy scores from the current follow up.

The study was approved by the Regional Ethics Committee in Stockholm (approval number 2006/61-31/2 and amendment approved 14^th of November 2014).

Results

All mean literacy and language/cognitive scores are presented in Table 2. As can be seen, the average scores are close to age-typical performance or only slightly below. The variability was however note-worthy. In particular, the prevalence of literacy problems was statistically increased in the case of reading fluency (n = 8/39; 26% [95% CI: 14–43%]) and reading comprehension (n = 8/39; 26% [95% CI: 14–43%]), but not significantly so for word reading accuracy (5/40; 13% [95% CI: 5–26%]) or spelling (8/41; 20% [95% CI: 10–34%]).

Table 2. Literacy and language/cognitive test results.

	N	Min	Max	Mean	SD
Literacy skills
Word reading accuracy, percentile	40	1	75	44.9	24.6
Reading fluency, percentile	39	1	100	45.3	35.6
Reading comprehension, percentile	39	1	100	50.1	33.9
Spelling, stanine	41	1	7	4.5	2.1
Language/cognitive predictor skills
Listening comprehension, percentile	40	0.9	100	44.1	30.0
Rapid automatized naming, percentile	41	1.7	97.7	53.6	31.4
Phonological awareness, percentile	40	1	75	45.7	25.9
Nonverbal intelligence, percentile	41	1	99	60.6	32.7

Next, regression analyses were conducted in order to determine if the language/cognitive scores predicted individual differences in word reading accuracy, reading fluency, reading comprehension and spelling. Statistical details are reported in Table 3. The results show that while we were able to explain a substantial and significant amount of the variance for all literacy measures (all R² > 45%), the unique contributions of the different language/cognitive skills were found to differ for specific literacy subskills. For both word reading accuracy and spelling, the phonological awareness (sound deletion) and listening comprehension scores made unique contributions. For reading fluency, RAN contributed to the model besides phonological awareness, whereas the role of listening comprehension fell just shy of significance. Finally, reading comprehension was predicted only by listening comprehension. Nonverbal intelligence scores (Raven’s Progressive Matrices Test) did not uniquely predict any of the literacy scores in the regression analyses.

Table 3. Outcome from regression analyses (Method: Enter) with the four literacy measures as dependent variables.

	Word reading accuracy			Reading fluency			Reading comprehension			Spelling
	R^2/F	β	t	R^2/F	β	T	R^2/F	β	t	R^2/F	Β	T
Listening comprehension		.437	2.98**		.231	1.85^		.549	3.77**		.351	2.57*
RAN		−.063	−.406		.386	2.94**		.042	.27		−.158	−1.09
Phonological awareness		.437	2.21*		.351	2.33*		−.063	−.36		.639	3.86**
Nonverbal intelligence		−.01	−.062		.009	.065		.254	1.54		−.096	−.62
	.45/6.85**			.61/12.80**			.46/7.13**			.52/9.33**

^p < .10, *p < .05, **p < .01.

Finally, we compared literacy scores of participants that were identified with moderate to severe language problems as preschoolers to those with mild or no language problems as defined in Kjellmer et al. [16]. Non-parametric Mann–Whitney U tests showed that children with early moderate to severe language problems scored significantly lower in word reading accuracy (U = 71.5, p = .016) and spelling (U = 53, p = .002). There were no significant differences in reading fluency (p = .051) or in reading comprehension (p = .151) although the trend was in the same direction on these measures (see mean scores in Table 4).

Table 4. Literacy scores in children with and without preschool moderate to severe language problems.

Literacy skills	Subgroup	N	Mean	SD
Word reading accuracy*	Moderate/severe early language problems	29	40.34	24.59
	Mild/no early language problems	10	62.70	10.21
Reading fluency	Moderate/severe early language problems	28	39.61	36.35
	Mild/no early language problems	10	65.41	25.72
Reading comprehension	Moderate/severe early language problems	28	44.93	35.99
	Mild/no early language problems	10	65.03	25.51
Spelling*	Moderate/severe early language problems	30	4.00	2.03
	Mild/no early language problems	10	6.20	1.23

*denotes to a statistically (<.05) significant group difference.

Discussion

The present paper firstly confirmed prior research [8,9,22] by showing a great heterogeneity (from floor to ceiling levels) in literacy skills in this sample of 12-year olds with ASD without ID. Importantly, though, there were significantly increased rates of poor reading comprehension and reading fluency in our study cohort compared with population norms. The cross-sectional results obtained in the study also support the notion that different language/cognitive skills are uniquely predictive of different literacy subskills. These findings are important as they detail the language/cognitive background factors associated with each of the important literacy skills analyzed in the study. Interestingly, when it comes to the relative importance of the language/cognitive skills in predicting literacy scores, the pattern of results is likely to be familiar to many general reading researchers (i.e. outside the context of autism) (c.f., 3,7]. Thus, literacy seemed to be supported by many of the same underlying subskills in children with ASD as in non-ASD children. As pointed out by Bailey and Arciuli [14], potentially important issues are at stake in this context, because if there are similar concurrent predictors of literacy and literacy problems in children with ASD, then there are greater reasons to assume that standard models of (remedial) literacy instruction might be helpful also for children with ASD. From clinical and educational perspectives, this also highlights the importance of assessing literacy skills in students with ASD so that pedagogic interventions can be provided for those in need of it – instead of assuming that any literacy problems are secondary or “just part of the autism.”

Besides support for general phonological models of spelling and word reading accuracy [4,7], we observed a clear and distinct association between RAN and reading fluency, also in keeping with a large general literature on children’s literacy development [23]. Besides phonological awareness, also semantic processing, indexed by listening comprehension capacities, made a unique contribution in predicting word reading accuracy and spelling, which align well with predictions from the lexical quality hypothesis [5]. Finally, for reading comprehension, we found it to be statistically predicted mainly by listening comprehension. Also this finding is in line with a widely accepted general model of reading comprehension, namely the simple view of reading [24], which predicts that reading comprehension capacities in individuals with adequate word reading skills (accuracy and fluency) will closely mirror listening comprehension.

Many important questions are raised as results of these findings. For one, future research is needed to explain why the predictive patterns are similar but with increased prevalence of literacy difficulties in those with ASD. This question, in our view, relates to the broader question of why supposedly “distinct” neurodevelopmental disorders – such as ASD, dyslexia, ADHD, language disorder, etc. – more generally, tend to overlap to a very considerable degree [2,3]. Secondly, since we found similar concurrent predictors of literacy in children with ASD as in the non-ASD population, classroom-based research is needed in order to examine the actual practical implications. Although this might mean that the general targets of literacy instruction can be similar for children with ASD, in our experience, as well as educational research, the organization of this support needs to be adapted in many cases. For instance, explicit and structured teaching that creates few instances of ambiguity for the student with ASD concerning what to do, where to do it, with whom and for how long is presumed to benefit both social and academic development in the school milieu [cf. 25]. These principles of explicit and structured instruction need to be integrated into the child’s literacy instruction as well [26].

In addition, we find that in children with ASD without ID, preschool moderate to severe language problems – considered more globally –constitute a risk factor for later problems with word reading accuracy and spelling, also much in line with previous research on children without ASD [27]. For reading fluency and reading comprehension, the trend was similar in direction but non-significant. The unfortunately small sample sizes (and low power) in the subgroup comparisons (i.e. in the comparisons between those with and without early language problems) call for caution when interpreting those null results, however.

In sum, the present study contributes to the current literature by including several literacy outcomes in one well-defined, age-homogenous sample and using similar assessment formats. This set up made it possible to systematically identify which language/cognitive capacities that predict different literacy subskills in school-age children with ASD. In addition to theoretical implications, these results are of practical relevance for families and professionals supporting the reading development of children with ASD.

Acknowledgement

We would like to thank the following persons for assistance during data collection: Annika Kirkham, Christina Thulin, Ulrike Andersson Stockhaus, Maria Henriksson, Tintin Hamrin and Josefine Kyhle.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Jakob Åsberg Johnels

Jakob Åsberg Johnels, PhD, is an associate professor at the University of Gothenburg, including at the Gillberg Neuropsychiatry Centre and the Section for Speech and language pathology. He also performs academic skills assessments at the Child Neuropsychiatric Clinic (CNC) at Queen Silvia Children's Hospital, Sahlgrenska University Hospital.

Elisabeth Fernell

Elisabeth Fernell is a professor of child and adolescent psychiatry and a paediatrician specialised in child neurology and habilitation. She is a chief physician at the CNC at Queen Silvia Children's Hospital, Sahlgrenska University Hospital.

Liselotte Kjellmer

Liselotte Kjellmer is an assistant professor in special education at Stockholm University and a school-based speech-language pathologist in the municipality of Alingsås. She is also a national research collaborator of the Gillberg Centre at the University of Gothenburg.

Christopher Gillberg

Christopher Gillberg is Professor of Child and Adolescent Psychiatry at the University of Gothenburg, Sweden, since the mid-1980s. He is also a Chief Physician at the Sahlgrenska University Hospital and one of the world's most experienced, clinically active, child- and adolescent psychiatrists.

Fritjof Norrelgen

Fritjof Norrelgen, PhD, is a speech language pathologist at Karolinska University Hospital. He has extensive experience of clinical work with and research on children and adolescents with neurodevelopmental disorders.

Notes

1 In the DSM 5, also written expression of text and math and calculation skills are listed; these skills are not considered in the present paper.