The Use of Pictorial or Graphic Representation in Reading Comprehension Interventions for Students with Autism Spectrum Disorders: A Meta-Analysis

Reading is a complex process that involves decoding text, recognizing words, understanding their meanings, parsing sentences, making inferences, and actively monitoring comprehension (Perfetti & Stafura, 2014). While reading, readers form situation models through dynamic interactions between linguistic representations and background knowledge (Kintsch, 1988). This process allows readers to identify coherence in texts, draw inferences, and integrate information and relevant prior knowledge to construct a comprehensive understanding of what they read (Kintsch, 1988; McNamara & Magliano, 2009). Without this deeper integration, readers may focus only on surface-level details without inferencing, which may result in fragmented comprehension and difficulty connecting ideas across the text (Cain & Oakhill, 1999; Cain et al., 2001; Kintsch & Rawson, 2005; Miller & Keenan, 2009).

Many students with ASD have a detail-oriented cognitive processing disposition, which often affects proficiency in recognizing patterns and specific details (e.g., Frith & Happé, 1994; Happé & Booth, 2008; Nuske & Bavin, 2011; Rumpf et al., 2012; Schlooz & Hulstijn, 2014; van der Hallen et al., 2015). Although this detail-oriented processing style is a cognitive strength in some contexts, it can create barriers to the integration of global meaning when reading texts. Studies suggest that between 30 and 69% of individuals with ASD experience difficulties with comprehension (Henderson et al., 2014; McIntyre et al., 2017; Solari et al., 2019), particularly in higher-order skills such as making inferences, understanding figurative language, and recognizing implicit text structures (Brown et al., 2013; Tárraga-Mínguez et al., 2021). Despite the variability in the magnitude of these estimates, by any measure, it is notable that reading comprehension challenges are not rare among readers with ASD.

The unique challenges faced by individuals with ASD in reading comprehension have been explored through the lens of social cognitive theory, the central coherence framework (Blum, 2019; Frith, 1989; Happé & Booth, 2008; Happé & Frith, 2006). Frith (1989) proposed that these challenges arise from an imbalance in how information is integrated at different levels of processing. This framework contrasts the typical inclination to synthesize diverse information to form a comprehensive understanding, known as central coherence, with the ASD tendency to focus on local information (Frith, 1989). As noted by Frith and Happé (1994), individuals with ASD excel at retaining local details while struggling to form the global coherence required to understand overarching themes and broader inferences.

Given these challenges, incorporating PGR may be a powerful strategy to enhance reading comprehension for students, including those with ASD. In this study, a PGR encompasses visual supports (e.g., graphic organizers) and visual representations of narratives (e.g., comic strips, animated illustrations). Visual support includes a range of tools, including graphic organizers (e.g., semantic mapping, flowchart), pictorial representations, and other visual aids designed to scaffold understanding of new skills, behavioral expectations, or activities (Hume et al., 2014). Among these, graphic organizers stand out as effective tools for helping students visually display information, as demonstrated in formats like mind maps, charts, tables, and Venn diagrams (Urton et al., 2024). Research has demonstrated their effectiveness in improving comprehension across various student populations, including those with ASD (e.g., Bethune & Wood, 2013) and learning disabilities (e.g., Dexter et al., 2011; Kim et al., 2004). However, not all graphic organizers are equally effective for all tasks. For example, simplistic structures like the ‘beginning, middle, end’ framework may not provide enough cognitive challenge for students working on inferencing skills. Colliot and Jamet (2018) emphasized that the design of visual supports must balance cognitive demands to avoid hindering learning.

Representing information in a visual format can assist in creating mental representations, or situation models, by helping students organize and retain information from texts (Duke & Pearson, 2009). Previous studies have shown that mental models serve as the foundation for comprehension across various forms of media (i.e., Cohn, 2018; McNamara & Magliano, 2009; Kendeou et al., (2020; Magliano et al., 2016). In addition, visual representations not only aid in understanding literal content but also provide scaffolding for navigating non-literal constructs essential for critical analysis and knowledge transfer (Altun, 2018; Kendeou et al., 2020; Magliano et al., 2013). By fostering connections between discrete pieces of information, visual representations can encourage deeper comprehension of texts (Dexter et al., 2011).

Blum (2019) challenges the deficit assumption in narrative comprehension among students with ASD, proposing that comprehension difficulties may be more attributable to modality. Blum (2019) compared the impact of comic-plus-text narratives with traditional text-only formats on inferential reasoning and found that although students with ASD exhibited difficulties with inferential reasoning in text-only conditions, their performance improved significantly when engaging with comic-plus-text, especially for those with prior experience with comics. The study suggests that visual narratives can serve as an alternative literacy that can scaffold comprehension by making implicit narrative elements more explicit (Blum, 2019; McVicker, 2007; Rozema, 2015). Multimedia scaffolds such as comics are not merely supplementary tools but essential components for encouraging mental representations and building coherence between narrative elements (McVicker, 2007).

Several studies have suggested frameworks that support inferential thinking across different media, including written, graphically or auditorily represented information (e.g., Blum et al., 2020; Loschky et al., 2015; Kendeou et al., (2020); Ness-Maddox, 2022). Inference plays a significant role in bridging implicit gaps by retrieving and integrating relevant information (Kintsch, 1998); McNamara & Magliano, 2009; Oakhill, 1984), and studies suggest that inferential comprehension relies on shared cognitive processes across multiple formats, including both textual and visual narratives (e.g., Kendeou et al., 2020; Kim, 2016; Magliano et al., 2013).

The Inferential Language Comprehension (ILC) framework (Kendeou et al., 2020) highlights that leveraging visual narratives—both static and dynamic—facilitates inference-making because they provide contextual scaffolding that aids comprehension. According to Kendeou et al. (2020), the ability to integrate information from visual cues helps students to bridge narrative gaps and form coherent mental models of the information presented. Furthermore, studies indicate that questioning techniques alongside visual supports further enhance inferencing skills, as they actively prompt learners to engage in the activation and integration process necessary for comprehension (e.g., Kendeou et al., 2020; Magliano et al., 2013).

Ness-Maddox (2022) investigated the extent to which readers generate different types of inferences depending on the modality of the material, comparing text-based and non-linguistic graphic narratives. In this study, college students were given either text or non-linguistic graphics and asked to engage in think-aloud activities while answering recall questions. The findings revealed strong correlations between text and non-linguistic graphic narratives for certain inference types, particularly anaphoric, bridging, elaborative, and internal state inferences. However, the frequency of specific inferences varied by modality. Participants engaging with text narratives generated more goal statements, predictions, and affective responses, whereas non-linguistic graphics prompted more bridging inferences, anaphoric inferences, and emotion inferences. Notably, those in the graphic narrative condition generated more emotion inferences, whereas those in the text condition provided more detailed story recall (Ness-Maddox, 2022).

Collectively, these studies highlight a key insight: inferential thinking extends beyond any single modality and is influenced by the way information is presented (Blum et al., 2020; Kendeou, 2015; Kendeou et al., 2020; Ness-Maddox, 2022). Given these findings, an important question emerges—how frequently have reading comprehension interventions effectively applied PGR to support students in K-12 settings in forming situation models? Addressing this question is critical for developing instructional strategies that leverage the full potential of visual scaffolds to enhance reading comprehension.

Guo et al. (2020) conducted a meta-analysis on the effectiveness of graphic displays in supporting students’ reading comprehension and ultimately included 36 articles. Their broader inclusion criteria encompassed various types of visual aids such as pictures and pictorial diagrams (see p. 9 in Guo et al., 2020). Of the identified studies, only sixteen studies exclusively included flow diagrams, which is relevant to the focus of the present study. In addition, 58% of the studies included adult participants. Guo et al. (2020) indicated that less than half of the studies reported participants’ reading skills (n = 18), which may or may not include students with reading difficulties or disabilities. Despite the limited number of reading comprehension interventions with PGR, especially in K-12 settings, there have been a few studies that included PGR to improve students’ reading comprehension.

Danaei et al. (2020) reported mixed results on the impact of PGR in reading comprehension interventions. They compared two types of interventions, augmented reality (AR) story books or traditional print books, to enhance students’ reading comprehension, with a particular focus on how graphic representations influence learning outcomes. The study involved 34 neurotypical children aged 7 to 9 in Iran. After reading, they were asked to retell the story and answer comprehension questions. AR storybooks enhance children’s reading comprehension by integrating dynamic elements such as animations, sound effects, and narration. Results revealed a significant improvement in overall reading comprehension for children who engaged with the AR-enhanced storybook. These children performed better in retelling and answering comprehension questions, particularly in implicit understanding. However, no notable difference was found between the groups in retelling the theme and setting.

Another example of using PGR in comprehension interventions is the Technology-based Early Language Comprehension Intervention (TeLCI; McMaster et al., 2024), which supports the development of reading comprehension in young children by emphasizing inferencing skills through dynamic visual narratives (i.e., videos) rather than traditional text-based reading and decoding-based approaches. McMaster et al. (2024) found that Grade 1 and 2 neurotypical students who used TeLCI demonstrated growth in their ability to make inferences. However, when comparing the experimental group with the control group, the improvements in inferencing were not significantly different, which indicates that TeLCI’s effectiveness may be comparable to other standard reading comprehension interventions. Although the aforementioned studies (i.e., Danaei et al., 2020; McMaster et al., 2024) did not specifically target students with ASD, these findings highlight the potential for inferencing skills to transfer across media.

Tárraga-Mínguez et al. (2021) conducted a systematic review analyzing the effectiveness of reading comprehension interventions specifically designed for children with ASD. The review covered 25 studies published between 2000 and 2019, including 196 participants aged 5 to 18 years. Tárraga-Mínguez et al. (2021) evaluated various interventions that targeted specific reading comprehension sub-processes, such as understanding inferences, identifying main ideas, and recognizing text structures. Their findings underscored the effectiveness of structured interventions, particularly those incorporating direct instruction, collaborative learning, and the use of visual supports such as concept maps and graphic organizers. These tools significantly improved comprehension by providing a structured framework for organizing information, which is particularly beneficial for children with ASD. However, out of the 25 studies reviewed, only one study used digital concept maps of texts that were relevant with PGR (i.e., Browder et al., 2017).

This meta-analysis aims to better understand how PGR impacts reading comprehension specifically among students with ASD, distinguishing their effects across different modalities, age groups, intervention settings, and task types. It deliberately focuses on PGRs rather than the broader term visual support, to ensure specificity in the types of interventions analyzed. This study specifically examines interventions that use structured visual formats to facilitate comprehension by promoting situational models. The following section will discuss a more detailed definition of PGR. This study addresses three questions:

To extract numerical data from graphical representations in the original studies, the juicr package in R was employed (Lajeunesse, 2021). This package provides a graphical user interface that facilitates extraction and conversion of image-based data into numerical values. The second author played a key role in setting up the juicr and establishing guidelines for its use. The first author conducted the initial data extraction for all five studies, with the third author reviewing and verifying the extracted data for accuracy. Discrepancies between the two—39 disagreements out of 445 data points—were resolved through consensus discussions. An interrater agreement yielded a reliability rate of 91.24%.

To determine the appropriate meta-analytic model, both fixed-effects and random-effects models were estimated using the metafor package (Viechtbauer, 2010; see Table 2). Because ATDs require an effect size metric that accounts for immediate treatment effects across multiple sessions, Tau-U was selected as the primary metric due to its ability to synthesize data across different SCED designs. Tau-U accounts for changes in both level and trend change, adjusts for positive baseline trends when necessary, and does not suffer from the ceiling effects observed in other nonoverlap techniques (Parker et al., 2011). Tau-U effect sizes were weighted within each study, using the Tau-U Calculator from Single Case Research (Vannest et al., 2016).

To assess potential publication bias, Kendall’s Rank Correlation test (Kendall & Gibbons, 1999) was used to examine asymmetry in effect size distribution. This test evaluates the proportion of data pairs that show improvement over time and offers an interpretable measure of trend consistency (Parker et al., 2011). In addition, a forest plot (Fig. 2) was generated to visualize the effect sizes and confidence intervals for each included study included in the meta-analysis. Each study’s point estimate (i.e., Tau-U) and 95% confidence interval are displayed, along with study labels for reference. The overall effect size is represented by a diamond shape, which indicates the aggregated effect across studies. The funnel plot (Fig. 3) was also generated to examine the potential presence of publication bias by plotting effect sizes against their standard errors.

With the five reviewed studies, both fixed- and random-effects models yielded nearly identical estimates, with an effect size of 0.85, a standard error of 0.19, and a highly significant z-value of 4.521 (p < 0.0001), as shown in Table 2. Their confidence intervals [0.48, 1.22] also completely overlap. Notably, the heterogeneity tests reveal I² of 0.00% and a non-significant Q statistic (p = 0.98). The random-effects model confirms this with a tau² estimate of 0. Also, the fixed-effects model demonstrates lower information criteria (AIC = 3.15, BIC = 2.76) compared to the random-effects model (AIC = 5.04, BIC = 3.82). Therefore, the fixed-effects model was selected.

The fixed effects mean of 0.85 [0.48, 1.22] supports the overall moderate-to-strong effectiveness of PGR. The studies analyzed each show positive Tau-U results, with four of the five showing statistically significant positive effects (range = 0.82–1.00). Browder et al. (2017) reported the highest intervention effect suggesting a large, positive impact. The lowest was reported in Schatz (2017) with a moderate effect size of 0.61 [−0.42, 1.64].

The five reviewed studies included different types of PGR, and some variability of PGR effects was observed across modalities. Browder et al. (2017) and Kouo and Visco (2021) only used technology-based PGR. The former reported an effect size of 1.00 [0.13, 1.87], and the latter reported 0.95 [0.02, 1.89] with a moderate average variability in effects. The other three studies used both tech-based and paper-based modalities in their interventions. Drill and Bellini (2022) reported a high effect size of 0.84 [0.07, 1.61], whereas Schatz (2017) with the same interventions reported 0.61 [−0.42, 1.64]. Sartini (2016) also used both modalities and reported an effect size of 0.82 [0.15, 1.49].

Variability in intervention effects was also observed when considering individual and contextual factors. Regarding age groups, three out of five studies (i.e., Browder et al., 2017; Kouo & Visco, 2021; Sartini, 2016) focused on elementary school students, whereas the remaining two (i.e., Drill & Bellini, 2022; Schatz, 2017) included middle school participants in residential settings. There were four different types of task types: literal comprehension questions, inferential comprehension questions, combined literal and inferential questions, and Maze.

Publication bias was assessed by examining funnel plot asymmetry using the rank correlation test with Kendall’s tau. The analysis yielded Kendall’s tau of 0.00 and a p-value of 1.00, which indicates no significant correlation between study effect sizes and their precision. This lack of correlation suggests that there is no systematic bias in the selection of studies; specifically, smaller studies are not disproportionately reporting extreme effects, which often signals publication bias. Although these results are reassuring, it is important to acknowledge that tests for funnel plot asymmetry may have limited power, particularly when the number of studies in the meta-analysis is small, as is the case in this study.

This meta-analysis examined the effectiveness of PGR in reading comprehension interventions for students with ASD. Findings indicate that although all studies demonstrated moderate to strong positive effects, there was notable variability in intervention outcomes. The fixed effects mean effect size of 0.85 [0.48, 1.22] suggests that, on average, PGR provides significant benefits for students with ASD. However, variations in effect sizes within and across studies emphasize the importance of factors such as intervention design and target skills.

Also, the participants’ racial distribution is not entirely representative of the broader U.S. public school student population. While race was not reported for 13% of participants, the available data suggest an overrepresentation of White students and an underrepresentation of other racial and ethnic groups compared to national averages. Future research would benefit from intentionally including a more diverse student population to enhance the generalizability of findings regarding interventions for students with ASD.

Another key limitation of this meta-analysis is its intentional focus on interventions using structured visual formats (i.e., PGR), which potentially limited the generalizability of the findings to other types of PGR. As indicated by Guo et al. (2020), including broader categories of visual aids such as pictures, pictorial diagrams, flow diagrams, and other mixed types of graphic displays can provide more comprehensive insights into the effectiveness of PGR in supporting reading comprehension. Future studies should therefore explore a wider range of visual formats to better understand the overall impact of PGR on reading comprehension for students with ASD.

While this meta-analysis intended to examine the differences in effect size based on the instructional settings, comparisons across instructional settings (e.g., small group vs. individualized instruction) were not possible due to the lack of variability in study settings. Future research should investigate whether group-based PGR interventions produce similar or different outcomes compared to one-on-one interventions, particularly in classroom environments where peer interaction and collaborative learning may play a role in comprehension development. Also, studies should explore whether integrating multiple modalities (e.g., paper- and tech-based approaches) yields additive benefits. Understanding these factors can help practitioners optimize reading comprehension interventions and enhance educational outcomes for students with ASD.

The findings of this meta-analysis provide important insights for practitioners, instructional designers, and researchers working with students with ASD. The overall effectiveness of PGR suggests that visual support should be systematically integrated into reading comprehension interventions to enhance student engagement and understanding. Structured visual formats such as story maps can serve as effective scaffolds for improving reading comprehension in students with ASD, as shown in Schatz (2017) and Drill and Bellini (2022). These tools may help reduce cognitive load and provide clear structure, especially for students who benefit from concrete visual representations of abstract concepts (Schatz, 2017; Stringfield et al., 2011).

The variability in intervention effects underscores the need for individualized instructional approaches. Educators should consider student-specific factors such as cognitive processing styles, prior knowledge, and learning environment when designing reading comprehension interventions. For example, Blum (2019) found that students with prior experiences with comic strips produced higher levels of inferences when given graphic narratives. Implementing differentiated instruction and adapting visual supports based on students’ unique needs may result in more effective learning outcomes. Additionally, explicit instruction in how to use these supports effectively may enhance comprehension gains, particularly for students who struggle with independent application of visual strategies (Roux et al., 2014).

The subgroup analysis further highlights several important considerations for educational practice and research. Although tech-based interventions show promise, their effectiveness varies significantly, suggesting that implementation fidelity and instructional design play critical roles in determining success. In contrast, paper-based interventions appeared to yield more stable and consistent benefits, although further research is needed to confirm this pattern across larger samples. Elementary-aged students demonstrated stronger intervention effects compared to middle school students, which indicates that earlier exposure to PGR strategies may be more beneficial. These findings suggest that early intervention efforts should prioritize structured visual strategies to build foundational comprehension skills before students face more complex reading tasks.

In conclusion, this meta-analysis provides compelling evidence that PGR can significantly enhance reading comprehension for students with ASD. The overall moderate-to-strong effect size observed (Tau-U = 0.85) indicates that PGR interventions are beneficial, though the extent of effectiveness varies depending on modality, task type, and individual factors. Technology-based interventions showed strong results, but there was also notable variability, with paper-based methods demonstrating more consistent outcomes. The findings suggest that careful selection of appropriate visual supports tailored to the cognitive profiles and needs of students with ASD is critical. Ultimately, the integration of PGR into reading comprehension strategies holds promise for improving educational outcomes for students with ASD, particularly when designed to align with students’ unique learning profiles.