The potential impact of literacy intervention on speech sound production in students with intellectual disability and communication difficulties

ABSTRACT

A small body of research and reports from educational and clinical practice suggest that teaching literacy skills may facilitate the development of speech sound production in students with intellectual disabilities (ID). However, intervention research is needed to test the potential connection. This study aimed to investigate whether twelve weeks of systematic, digital literacy intervention enhanced speech sound production in students with ID and communication difficulties. A sample of 121 students with ID were assigned to four different groups: phonics-based, comprehension-based, a combination with both phonics- and comprehension-based intervention and a comparison group with teaching-as-usual. Speech sound production was assessed before and after the intervention. The results on the data without the imputed variable suggested a significant positive effect of systematic, digital literacy interventions on speech sound production. However, results from sensitivity analyses with imputed missing data was more ambiguous, with the effect only approaching significance (ps = .05–.07) for one of the interventions. Nonetheless, we tentatively suggest that systematic, digital literacy intervention could support speech development in students with ID and communication difficulties. Future research should be done to confirm and further elucidate the functional mechanisms of this link, so that we may have a better understanding and can improve instruction and the pivotal abilities of speech and reading.

KEYWORDS:

• Intellectual disability
• communication disorder
• speech sound production
• literacy
• educational intervention

Introduction

Individuals with intellectual disabilities (ID) often face challenges in communication, which in turn affects the ability to participate in daily life activities and education. A limited ability to communicate, can cause misunderstandings and frustration, both for the individual with ID and their communication partner (Drager et al., 2010; Light et al., 2008). Communication is essential to learning and wellbeing, and it is important to support various forms of expression, including speech, for individuals with ID (Terband et al., 2018). Additionally, a significant number of individuals with ID and communication difficulties have problems developing functional reading skills (Ainsworth et al., 2016; Erickson & Geist, 2016; Ratz & Lenhard, 2013). There is some evidence suggesting that literacy interventions might positively impact speech sound production in this population (Buckley & Bird, 1993; Light et al., 2008). While prior studies are promising, they have several limitations. For instance, earlier studies have generally used small sample sizes and most lack a control group. The nature of the literacy intervention, for example, whether it focuses on word reading and code-based skills (e.g. phonics) and/or comprehension and meaning (The National Reading Panel, 2000), has also not been thoroughly explored in relation to speech outcomes. Finally, several of the available intervention studies have focused on the Down syndrome subpopulation rather than a wider ID population, meaning that questions remain regarding the generalisability of previous results.

Students with ID often require more time than typically developing children to acquire skills like speech and literacy (Patel et al., 2020). However, time is a limited resource, especially considering the numerous tasks that need to be addressed within a school day. This was pointed out by the National Agency for Education in the curriculum for schools for students with ID (Skolverket, 2022) in Sweden, which is the context for the current study. Thus, if a targeted approach to literacy instruction could improve speech and communication skills in individuals with ID, it would contribute to their participation in society. The aim of this study was to investigate if systematic, digital literacy interventions influenced speech sound production in school-aged students with ID and communication difficulties.

Connections between literacy and speech

There is a complex connection between literacy and speech (Hoover & Tunmer, 2018; Stackhouse & Wells, 1997; Tunmer & Hoover, 2019). Kolinsky et al. (2021) proposes that as literacy skills improve, the individual’s awareness of both written and spoken words strengthens, potentially leading to a positive effect on speech. There is research indicating that acquiring literacy skills can impact the quality of speech by enhancing the accuracy of underlying linguistic representations (Brady et al., 1994; Dodd & Gillon, 2001; Huettig & Pickering, 2019; Knight et al., 2015). More specifically, while phonological representations and speech production skills usually develop before literacy instruction and literacy acquisition take place, it has been proposed that learning to read and spell can refine phonological contrasts and affect the quality of speech (Rastle et al., 2011; Saletta, 2015). Several studies have found correlations between early literacy skills and quality of speech sound production in students with ID and communication difficulties (e.g. Barton-Hulsey et al., 2018; Burgoyne et al., 2021; Peeters et al., 2009; Samuelsson et al., 2023). While observational studies are important in establishing a potential connection between literacy learning and speech quality, determining causality in this relationship is difficult because of the correlational nature of the data.

Research investigating the effect of literacy interventions on speech production and communication in students with communication difficulties, with or without ID, is limited but important since such studies inform casual hypotheses. A study by Moriarty and Gillon (2006) that included children with apraxia of speech found that phonological awareness and letter-sound intervention improved not only word reading skills but also speech sound production analysed with percentage phoneme correct. Light et al. (2008) conducted a case study in which an 8-year-old girl with multiple disabilities who used augmentative and alternative communication (AAC) improved her literacy and communication skills through the Accessible Literacy Learning (ALL) curriculum. The improvement in literacy skills aligned with language and communication skills. Buckley and Bird (1993) also qualitatively described the effect literacy intervention seemed to have on children with Down Syndrome on their language and speech production (i.e. more words, longer sentences, correct grammar). Kennedy and Flynn (2003) and van Bysterveldt et al. (2010) have documented additional research demonstrating positive effects on speech resulting from literacy interventions among individuals with Down syndrome. Knight et al. (2015) did an experimental study and compared the pronunciation of words elicited through imitation, picture naming, or reading aloud in eight children (aged 11–14 years). The results showed that reading aloud produced significantly more accurate and intelligible speech, as calculated by the percentage of correctly produced consonants. Finally, a Swedish study (Tjus et al., 2001) found that a comprehension-based digital literacy intervention increased the number of verbal expressions (i.e. how much participants talked) in the 20 participating children (eleven with ID and autism, nine with ID and mixed sensory and motor problems). Taken together, these intervention studies suggest that practicing literacy skills may lead to improved communication, as well as language and speech production, according to various kinds of data (e.g. clinical judgement and test scores).

The current study utilised a controlled design to explore speech sound production among students with ID and communication difficulties following a period of systematic, digital literacy intervention. This study was conducted as part of a larger project (whose main outcomes are to be published elsewhere), aiming to compare the effect on literacy of different kinds of systematic, digital literacy interventions using phonics-based or comprehension-based literacy instructions, or a combination of both. The interventions were digital applications designed to integrate text, pictures/animations, and text-to-speech synthesis, and to work systematically with the students’ reading progress. Given the limited research and theory on the effect of specific types of literacy interventions on speech sound production, the aim of this study was to explore the effects of the three literacy interventions (phonics-based instructions, comprehension-based instructions, or the combination of both type of instructions) on speech, by comparing their impact on phoneme production to a teaching-as-usual condition.

Method

This study used a non-randomised controlled trial design, comparing data before and after systematic, digital literacy interventions. The study was approved by the Swedish Ethical Review Authority (case number 2019-03845 and 2020-06215). Prior to participation, written information supported with pictures was provided to the students and their caregivers for written consent.

Recruitment and allocation of schools to literacy instruction condition

To recruit participants, information about the literacy project was advertised through conferences, social media, and email lists via participating organisations. Teachers, special educators, or principals from schools for students with ID then expressed interest and indicated the potential number of participating students. Teachers were instructed to select students who had communication difficulties, were in need of AAC, and not able to decode independently or recognise a maximum of 20 whole words. Both applications used in this literacy intervention are designed for individuals in need of AAC.

For practical reasons, allocation was done on a school-basis, meaning all enrolled participating students from one school were in the same group. Schools could not choose which condition they would be allocated to. The literacy interventions focused on both phonics- and comprehension-based literacy instruction. Four groups were formed: the first received teaching-as-usual (Comparison); the second received phonics-based instruction (Phonics); the third received comprehension-based instruction (Comprehension); and the fourth received a combination of both phonics- and comprehension-based instruction (Combination). Participating teachers in all four groups filled out a digital logbook every week. The intervention groups reported details about their work regarding the applications, and the Comparison group reported their teaching content regarding literacy. When reviewing the logbooks from the Comparison group (teaching-as-usual), limited instruction was apparent. There were limited aspects such as letter-name knowledge, letter-sound knowledge, phonics, and even less, reading and comprehending texts. According to our data, the literacy instruction often seemed to lack a systematic and explicit format. Importantly, the three intervention groups did not receive any restrictions regarding other literacy instructions; thus, the applications could serve as a complement to other teaching. The recruitment had to be conducted in two waves due to the large number of students being recruited, and the four different groups started at different times between January 2020 and January 2021.

Participants

Recruitment involved 137 students. Sixteen of these students were not included in the analysis: nine students chose to or were unable to participate in testing (3 with mild ID, 4 with moderate ID, 1 with severe ID and 1 unknown level of ID). Additionally, five did not participate in the intervention due to Covid-related conditions and did not attend school during this time. For two students, video recordings were missing. A final group of 121 students (69 boys and 52 girls) participated in the test administration and intervention for this study. From the first wave of recruitment, 58 participants were assigned to the Comparison group (n = 28) and to the Phonics group (n = 30). The participants from the second wave of recruitment were assigned to the Comprehension group (n = 29), and the Combination group (n = 34).

The ages of the students ranged from 6 years, 9 months to 21 years, 5 months, with an average of 13 years and 9 months. Diagnostic information (based on ICD-10, 2007), was obtained from the caregivers (Table 1). The ID levels were mild 32%, moderate 55%, severe 10%, and 3% unknown. Many students had comorbid diagnoses, such as autism, ADHD, or sensory impairment. A minority (n = 28) had a rare or unknown diagnosis, which included a combination of chromosome and genetic disorders. In the group of students with cerebral palsy, four were classified as spastic, five as dyskinetic and two was unspecified. Twelve students were reported to have hearing impairment, ranging from mild to severe. Forty-three students were reported to have visual impairment, 12 of whom could not have their vision fully corrected by glasses. Students who needed hearing or visual aids used them during the study.

Table 1. Categorisation of participants based on caregiver-reported age, gender, diagnoses and means of communication. Diagnostic categories of autism, Down syndrome, cerebral palsy, and rare diagnosis are in addition to intellectual disability.

		Comparison	Phonics	Comprehension	Combination
Mean age (year:month)		13:11	15:1	12:3	13:6
Gender (boys/girls)		15/13	17/13	16/13	21/13
ID	Total	28	30	29	34
	Mild	5	8	15	11
	Moderate	17	16	13	21
	Severe	6	4	0	2
	Unknown level	0	2	1	0
Diagnoses
	ID -no second diagnosis	4	4	5	8
	Autism	12	5	14	10
	Down syndrome	8	10	3	6
	Cerebral palsy	3	3	2	3
	Rare diagnosis	9	8	5	8
Means of communication
	Nonverbal communication^a	25	17	23	23
	Manual signing	18	20	17	20
	Pictorial support^b	22	19	19	24
	Speech generating device	5	5	4	4
% correct speech sound production at T1		50%	64%	73%	72%

^aNonverbal communication includes facial expressions, gestures, eye contact, vocalisation and body movement. ^b Pictorial support includes pictorial schedules, single pictures, picture communication boards, and picture communication books.

The groups were not fully balanced regarding ID level (Table 1), and there was a statistical significant difference in the distribution across levels between the Comparison group and the Comprehension group (χ² (1, N = 56) = 11.5, p = 0.003) more students with mild ID and none with severe ID in the Comprehension group. This unbalance could not be corrected during data collection because recruitment was conducted in two waves. Although there are statistical differences in frequency regarding ID level between these two groups, a comparison remains valuable. Indeed, a previous study on the same group of participants indicated that neither IQ scores nor age accounted for more than 3% of the variance in speech sound production (Samuelsson et al., 2023), suggesting that these variables probably do not explain the observed differences in speech sound production between groups in this study.

The students’ expressive speech ability varied from minimal to significant impairment. All students had access to some form of AAC in their daily school environment for expressive support, comprehension support, or both. The use of AAC varied, including manual signs, pictorial support, and technology-based tools such as speech-generating devices. Most students used multiple AAC modalities for different situations. For more information on diagnoses and means of communication, see Table 1.

Assessment

Speech sound production

The assessment of speech sound production was carried out using a sub-test from the Assessment of Phonology (Frylmark, 2015), a tool designed to assess phoneme production. For this study, the subtest of an action picture illustrating a café visit was selected to elicit speech sound production. It contained 28 target words representing 138 phonemes, covering all Swedish phonemes except/m/(although all phonemes were not in every position of the word). All participants had reached the age of six, which is the typical age at which Swedish children are expected to be able to produce all Swedish phonemes (Blumenthal & Lundeborg Hammarström, 2014). The students were first asked to describe the action picture verbally, with the aim of obtaining all 28 target words in their spontaneous speech. If this was not possible, the administrator pointed to specific objects or actions, eliciting a response by the participant using a prompt such as ‘this boy is eating a … ’. If the student did not respond, the administrator pronounced the word and asked the student to repeat it. All 121 students were assessed at T1, including those with limited speech, and all 138 phonemes were individually scored as correct or incorrect for all participants who underwent the assessment. At T2, 105 students were tested. The absence of 16 students at T2 was due to COVID-19 pandemic-related regulations that prevented access to some of the schools, thus we were unable to conduct the necessary assessments. No data have been discarded.

The test sessions were video-recorded with a Sony HDR-CX240 with a built-in microphone to enable later transcription and analysis. The camera was placed on a table approximately 1 metre from the student.

Speech analysis and reliability

The transcription of all target words on phonemic level was conducted by either the first or fourth author. Each phoneme (138) in the 28 target words was scored as either correct or incorrect in terms of place and manner of articulation. Omissions and substitutions were scored as incorrect. If the child added morphological endings to the target word, such as pluralising ‘bun’ to ‘buns’, the plural ending was not analysed. However, the Swedish allophones for/r/were scored as correct. One point was awarded for each correctly produced phoneme. We have not made any attempt to distinguish between speech errors considered ‘delayed’ versus ‘atypical’ in the manner that has been done in some research on speech errors in English speaking children (c.f. Dodd et al., 1989; Shriberg et al., 2019; Stackhouse & Wells, 1997). The participants’ overall performance on speech sound production before the intervention, indicated by the percentage of correctly produced phonemes is presented in Table 1.

An interrater reliability analysis of speech sound production was conducted using a subset of 20 recordings, with ten students selected at random from each transcriber, which happened to represent all five diagnostic subgroups (ID = 2, autism = 6, Down syndrome = 7, cerebral palsy = 2, rare/unknown = 3). The agreement between the two raters on all 138 phonemes was calculated using Cohen’s kappa, with showed substantial agreement (κ = .78) (Landis & Koch, 1977).

Literacy instruction

The literacy instructions were based on two applications with different instructional approaches, which were aimed at improving the literacy skills of students with communication difficulties, including those using AAC. The applications were designed to be multimodal, integrating pictures/animations, text, and text-to-speech synthesis. Both were developed with a systematic approach to support students through their learning progression. The applications were designed to encourage teacher-student interaction and included features to increase motivation in order to facilitate the learning process. The teachers were also provided with a set of augmentative pictures and boards to support visual structure and communication. Animega-interactive sentences (is) had a comprehension-based approach (Comprehension group), and Accessible Literacy Learning (ALL) had a phonics-based approach (Phonics group), both described below. In the Combination group, teachers were instructed to alternate evenly between the two instructional strategies (using both ALL and the Animega-is applications) throughout the entire intervention period.

A comprehension-based application: Animega-interactive sentences

The comprehension-based instruction was represented by the application Animega-is from Topic DOS AB (Heimann & Lundälv, 2020), which offered two learning modes. The ‘create mode’ allowed the learner to construct events using text buttons, which were then accompanied by animations or video clips that corresponded to the created event. In the ‘test mode’, the learner’s proficiency was evaluated by having the learner view the animation, choose words and create a sentence that best represents what was viewed. The app featured several levels of complexity and built-in comprehension tests and encouraged exploration with the support of a teacher. The language material and accompanying animations provided motivation and opportunities for the learner to express their imagination, humour, and thoughts, and engage in conversations. The app was based on the use of recasts (Clarke et al., 2017) and aimed to achieve error-free co-construction of meaning through text, animations, and supportive interaction. The earlier software-based versions (Alpha Interactive Language Series, DeltaMessages, and Omega-interactive sentence) have all been used successfully in prior intervention research focusing on children with autism, hearing disability, dyslexia and intellectual disabilities (Fälth et al., 2013; Gustafson et al., 2011; Heimann et al., 1995; Tjus et al., 1998).

A phonics-based application: Accessible literacy learning

The phonics-based instruction was represented by the application ALL from Tobii Dynavox (n.d.). The app consisted of modules for basic literacy skills that combined direct phonics-based instruction with meaningful comprehension-based literacy activities. In the present study, the teachers were instructed to work only with the modules using phonics-based instruction, including sound blending, phoneme segmentation, letter-sound correspondences, and word decoding. The app had three levels of complexity and followed the students’ progress through the stages. Exercises included pictures to facilitate participation without speech. Seven studies that conducted interventions with the ALL Curriculum were reviewed by Yorke et al. (2020). The results indicated a reasonably large gain (39%) and a large effect (estimated with Tau-U = .78). All these studies had between one to eight participants ranging in age from 3;6 to 18;7. In one of the included studies, a case study by Light et al. (2008), the participant also showed an increase in communication skills, in addition to literacy development.

Procedure

Prior to the start of the intervention, the teachers participated in either a half day workshop (groups working with phonics-based instruction or comprehension-based instruction) or two half-day (the group working with both instructions) workshops about the specific literacy training application(s). The intervention was carried out over 12 weeks with a recommendation of 90 minutes a week distributed over three sessions (but with a strong recommendation to adjust to the motivation, needs and abilities of the individual students).

The students underwent individual assessment of speech sound production before and after the literacy intervention. This assessment was carried out at their schools, in a quiet setting, and students had the option of having a teacher or assistant present during the assessment. The tests were administered by authors 1, 2 and 4. The first two authors are experienced speech and language pathologists, and the fourth author has experience from behavioural and psychological assessment as a researcher.

A pictorial schedule was used to visualise the test order to the students. Throughout the assessment session, verbal instructions were complemented with the use of picture communication and manual signs as necessary. The assessment lasted approximately 45–75 minutes with short breaks included. During these breaks, students could choose between various activities, such as eating a snack, drinking water or taking a short movement break, with pictures presented to aid their choice.

Statistical analysis

A logistic mixed-effects model was applied to test the effect of time (counted as Days from the first test occasion, standardised and centred at T1 for day 0), group (dummy coded using three binary variables, with the Comparison group defined as 0), and the interaction between time and each of the three dummy coded group variables, on whether a phoneme was produced correct. Analyses were performed with R Statistical Software (version 4.2.2; R Core Team, 2022), running R Studio (version 2023.09.0; RStudio Team, 2023). The glmer function from lme4 (Bates et al., 2015) was used to model the binary outcome, using a logit function. For the random effects structure, we included both random intercepts and random slopes for participants, keeping it maximal (Barr et al., 2013). Maximum likelihood estimation with BOBYQA optimisation was applied, and the alpha level was set to .05 for all effects. Tables showing the results from the glmer models were generated using the sjPlot package (Lüdecke, 2023). To deal with missing data in the outcome variable, multiple imputation using the mice package (van Buuren & Groothuis-Oudshoorn, 2011) was used. Predictors were set to any existing value on the outcome variable, the time variable, dummy variables for intervention groups and two behavioural measures of early literacy skills (i.e. letter-sound identification and phonological awareness) available in the larger project of which this study is part (see Samuelsson et al., 2023). The use of auxiliary variables, associated with the variable for which you are imputing data typically improves the precision of the imputation model (van Ginkel et al., 2020).

For model diagnostics, the DHARMa package (Hartig, 2022) was used. The dispersion and outlier tests were not significant, suggesting that the model variance matched the variance in the data and that no individual case was too influential on the model. The Kolmogorov-Smirnoff test suggested that residuals were non-normal. However, visual inspection of the Q-Q plot indicated only a slight deviation from normality. With the large number of observations (N = 32844) the Kolmogorov-Smirnoff test is likely to detect very small deviations that have little practical consequences. Thus, the general model fit was considered acceptable.

Missing data

Overall, the missing rate for observations was 2.6% for the dependent variable at T1 and 15.4% at T2 (see Table 2). The pattern of missing data was not equal across the four groups. Specifically, the Comparison and Phonics groups had more missing data than the two other groups, especially at T2 (see Table 2). In this case, the reason for missingness was known in all cases: regulations related to the COVID-19 pandemic prevented access to some of the schools, and we were unable to perform the necessary assessments. Still, missingness was associated with the group variable in our data. Thus, we treated missing data under a Missing at Random assumption, for which estimates can be unbiased (Enders, 2010; van Ginkel et al., 2020). However, to conduct a sensitivity analysis of our results, we applied multiple imputation to create five new data sets and then replicated the analysis from the original data for each of the separate data sets with imputed values. We compared consistencies between results across each of the five different models on imputed data sets and the results in the original model and assumed that a high level of consistency would reflect robustness of the results in the original analysis.

Table 2. Missing data pattern for the first (T1) and the second (T2) test occasion, across the four groups (Phonics, Comprehension, Combination, and Comparison). N = total phoneme items.

	Phonics		Comprehension		Combination		Comparison		Overall
	T1 (N = 4140)	T2 (N = 4140)	T1 (N = 4002)	T2 (N = 4002)	T1 (N = 4692)	T2 (N = 4692)	T1 (N = 3864)	T2 (N = 3864)	T1 (N = 16698)	T2 (N = 16698)
Response
0	1489 (36.0%)	1087 (26.3%)	1052 (26.3%)	649 (16.2%)	1319 (28.1%)	1037 (22.1%)	1786 (46.2%)	1408 (36.4%)	5646 (33.8%)	4181 (25.0%)
1	2597 (62.7%)	2186 (52.8%)	2880 (72.0%)	2900 (72.5%)	3367 (71.8%)	3379 (72.0%)	1772 (45.9%)	1483 (38.4%)	10616 (63.6%)	9948 (59.6%)
Missing	54 (1.3%)	867 (20.9%)	70 (1.7%)	453 (11.3%)	6 (0.1%)	276 (5.9%)	306 (7.9%)	973 (25.2%)	436 (2.6%)	2569 (15.4%)

Results

Testing the effect of literacy intervention on speech sound production

In the model on the data without the imputed variable, we tested for the main effects of time (Days) and of the three dummy-coded contrasts between each of the intervention groups (Phonics, Comprehension, and Combination) and the Comparison group. We further tested for the interaction between time and the three dummy-coded contrasts to see if any of the intervention groups had a stronger effect of time than the Comparison group.

Firstly, the effect of Days was not statistically significant (see Table 3). This means that the speech sound production of participants in the Comparison group did not improve between the first and second testing occasion. The effects of Comprehension group (beta = 2.19, 95% CI [0.61, 3.76], p = 0.006) and Combination group (beta = 1.72, 95% CI [0.25, 3.19], p = 0.022) were positive and statistically significant, which means that speech sound production was at a higher level in these intervention groups compared to the Comparison group at T1 (= day 0). Further, all three interaction effects between the intervention group variables and Days were positive and statistically significant (see Table 3). This means that all intervention groups had a higher level of change than the Comparison group, which received teaching-as-usual, between the first (before intervention) and second (after intervention) testing occasion. Looking at the point estimates and confidence intervals, the positive effect seemed most convincing for Comprehension group.

Table 3. The results for the logistic mixed-effects model, testing the effects of time (Days), intervention groups (Phonics, Comprehension, and the Combination), and interaction terms between time and intervention groups.

Coefficient	Odds Ratios	95% CI	p
Intercept	0.55	0.18–1.65	0.284
Days	0.89	0.67–1.17	0.407
Phonics vs Comparison	3.80	0.79–18.18	0.095
Comprehension vs Comparison	8.90	1.85–42.86	0.006
Combination vs Comparison	5.58	1.28–24.32	0.022
Days*Phonics vs Comparison	1.52	1.00–2.31	0.048
Days*Comprehension vs Comparison	1.86	1.27–2.72	0.001
Days*Combination vs Comparison	1.51	1.05–2.18	0.027
Random Effects
Error (σ^2)	3.29
Random intercept (τ00)	9.49
Random slope (τ11)	0.33
ICC	0.74
N	121 participants
Observations	30391
Marginal R2/Conditional R2	0.067/0.760

Days = Time in days, standardised and centred at day 0/T1. Phonics vs Comparison = The main effect of the Phonics intervention. Comprehension vs Comparison = The main effect of the Comprehension intervention. Combination vs Comparison = The effect of the Combination intervention. ICC = Intraclass correlation coefficient.

Sensitivity analyses with imputed values

As mentioned above, there were missing values in the data set as a result of restrictions during the COVID-19 pandemic. We therefore conducted sensitivity analyses with five sets of imputed data. The obtained results were somewhat more ambiguous than for the original data set. As was the case in the analysis on the original data, the main effect of Days was not significant. Also, the main effects of group, Comprehension and Combination were still significant before the intervention period. However, the interactions between intervention group variables and Days were all non-significant, although the interaction between Comprehension and Days approached the cut-off for statistical significance (p-values ranging from 0.05 to 0.07). Thus, the analyses on the data sets with imputed data did not reveal strong evidence of an effect of the literacy interventions on speech sound production. For detailed results from the analyses on the data sets with imputed data, see Supplement A.

Discussion

The present study investigated whether systematic, digital literacy interventions could enhance speech sound production, measured by the number of correctly pronounced phonemes, in students with ID and communication difficulties. We compared the effects of three different systematic, digital literacy instructions (phonics-based, comprehension-based, and a combination of the two) to teaching-as-usual. This study also included a large group of students with various aetiologies and different levels of ID. Thus, the present study was novel both in terms of the comparisons that we could do, and by including a larger sample than any previous study with a similar purpose did. Our results suggest that participating in systematic, digital literacy interventions may have a small positive effect on speech sound production compared to teaching-as-usual, although this conclusion was less clear when the analysis was based on data with imputed values. We propose that for children with ID and communication difficulties, literacy intervention in combination with other more specific interventions, could be a way to boost their speech sound development. Enhancing speech sound production could greatly benefit students with communication difficulties by facilitating their engagement in the classroom and educational interactions, as well as when interacting with others, and thus increase their participation McCormack et al. (2009); Samuelsson et al. (2024). A better understanding of the potential effects of literacy instruction beyond improved reading ability may help practitioners and researchers to identify effective interventions for improving both literacy and speech development in school-age children with ID, as well as contribute to improving theories describing the complex relationships between speech, language, and literacy development.

Although the main results suggest that all three interventions may show positive results on speech sound production, there is a need for future investigations to fully understand their effectiveness. Specifically, it is essential to examine the specific instructional approach and content used in these interventions. In another study, we reported that the comprehension-based application was perceived as more stimulating for conversation than the phonics-based application by the teachers in the Combination group (Samuelsson et al., 2024). This may have positively influenced speech production in favour of the comprehension-based intervention. However, the specific mechanisms involved in the potential effect of literacy instruction on speech need to be identified in future work.

It has been suggested that literacy intervention could yield positive effects on speech, since it supports enhanced phonological awareness that could in turn improve the precision of utterances (Brady et al., 1994; Dodd & Gillon, 2001; Huettig & Pickering, 2019; Knight et al., 2015). It is also possible that literacy intervention entails a focus on articulation and sounds in the written words that emphasise speech sound production (Saletta, 2015; Saletta et al., 2016). Another possibility is that engaging in conversations while working with various literacy instructions can contribute to overall language stimulation, potentially fostering language development (Clarke et al., 2017). Finally, by combining the visual representation of the letter, word or sentence with the speech output from the applications, we may strengthen the association with the spoken word, graphic symbol and internal phonological representation of the written word (Bishop et al., 2020; Blischak et al., 2003; Gevarter et al., 2016). Thus, phonological awareness, articulation awareness, semantic networks and overall language stimulation and other potential mediators should be tested in future research.

Another focus for future research is to investigate moderators that may exist within the group of students with ID and communication difficulties exploring how these factors may influence the extent to which literacy intervention affects speech production. Clearly, factors such as oral-motor function (Mogren et al., 2020) and oral-sensory perception (Aswathy et al., 2016) play significant roles in speech sound production for students with and without ID. Additionally, environmental factors such as the quality of language input (Ramírez-Esparza et al., 2014) and the availability of speech and language therapy (Law et al., 2003) impact speech development. The student’s own motivation is also important when it comes to speaking and learning to read (Logan et al., 2011). Therefore, if we are to better understand for whom and why literacy intervention could be beneficial for speech production, it is crucial to consider several factors when designing interventions aimed at improving speech production in individuals with ID and communication difficulties. An integrated and multidisciplinary approach that considers the individual’s strengths and weaknesses is most likely the best way to improve speech production and overall communication abilities, and we believe that this applies in the current context as well.

The role of the professional in an educational setting, especially when it comes to special education (FUB, 2023), the ability to interact with and to motivate students (McTigue et al., 2020), is of course another important factor contributing to educational success. Overby et al. (2007) also noted that a child’s degree of intelligibility might impact a teacher’s anticipations concerning the child’s academic achievements. Consequently, the expectations resulting from this interaction could potentially play a role in determining the effectiveness of literacy intervention. This notion finds support in a recent systematic literature review of research on a phonics-based literacy instruction, GraphoGame, where McTigue et al. (2020) concluded that working independently with Graphogame did not lead to clear progress for many young readers. Instead, it was the teacher’s didactic interaction with the student when working with the application that made the difference. In the present study, one possible explanation for the seemingly stronger effect for the Comprehension group than the other two intervention groups, is that working with the comprehension-based application may initiate more such interactions than working with the phonics-based application. This could then also explain why the Combination group showed the second strongest effect, and the Phonics group the weakest. However, which approach that is more effective and why should be investigated in future studies.

When it comes to limitations, missing data and unbalanced groups regarding ID level and speech sound production were present. The missing data was partly due to complications during the COVID-19 pandemic, and the unbalanced groups were because recruitment was conducted in two waves. The Comprehension group had a higher frequency of students with mild ID than the Comparison group, and the probability of correctly produced phoneme was also higher for the Comprehension and Combination group compared to the Comparison group. Better cognitive and adaptive skills, as indicated by ID level, and better speech sound production at the start of the study, may have influenced the outcome of speech sound production in favour of the Comprehension group. With that being said, we do not believe that these effects are particularly problematic since previous study involving the same participants as this study have demonstrated that IQ and age explained only 3% of the variance in speech sound production (Samuelsson et al., 2023). Thus, it is unlikely that differences in these variables explain solely why the Comprehension group shoved a steeper development in speech sound production in present study. Also, we tested the robustness of our results by replicating our analysis on five data sets with imputed values. The significant effects of intervention did not survive in these sensitivity analyses, although the imputed models indicated a close-to statistically significant effect for the comprehension-based application (Animega-is). While, collectively, this unfortunately dampens the conclusions that can be drawn in our study, we nonetheless find our positive trend-level findings to be interesting and important. This is especially true given the lack of prior research on literacy interventions on speech outcomes in this population.

Further research is needed to better understand the potential effect of literacy instruction on speech development. It is hoped that this relatively large systematic, digital literacy intervention study will stimulate such future research. Such research should also include both mediating and moderating variables in order to investigate if and how improving literacy skills can affect speech sound production for certain students. A better understanding of the possible effects of literacy instruction beyond improved reading ability may help practitioners and researchers to identify effective interventions for improving literacy, speech, and broader communication development in school-age children with ID.

Acknowledgments

We express our sincere thanks to all the participating teachers, students, and families for welcoming us into the school, especially during the challenging times of the COVID-19 pandemic.

Disclosure statement

Mikael Heimann and Mats Lundälv are co-creators and copyright holders of Animega- interactive sentences, which is available for free on platforms where applications can be downloaded. The data collection and data analyses have been performed by the other co-authors and no conflict of interest has been present.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/02699206.2024.2374915.

Additional information

Funding

The research was supported by grants from the Marcus and Amalia Wallenberg Foundation [Grant no. 2018-0084], the Swedish Research Council [Grant no. 2018-04702], Sävstaholm Foundation [Grant no. 2023-006], and Linnea and Josef Carlsson Memorial Foundation.