Reading
H. Lee Swanson & Jennifer E. Kong
What is Specific Learning Disorder with impairment in reading?
The purpose of this chapter is to review work that provides an empirical foundation for the view that reading disorders (RD) reflect a fundamental deficit in phono - logical short-term memory (STM) and the executive component of Working Memory (WM). Several authors find that children and adults with RD experience considerable difficulty on STM and WM tasks (e.g., Brandenburg, Klesczewski, Fischbach, Schuchardt, Büttner, & Hasselhorn, 2015; De Weerdt, Desoete, &
Roeyers, 2013; Gathercole, Alloway, Willis, & Adams; 2006; Swanson & Ashbaker, 2000; Wang & Gathercole, 2013). These memory deficits, depending on task demands, manifest themselves as a domain-specific constraint (i.e., the inefficient accessing of phonological representations) or a domain general constraint (i.e., capacity limitations in controlled attentional processing). Before discussing the research linking STM and WM to RD, operational definitions are reviewed.
Diagnostic features
The DSM-5 (American Psychiatric Association, 2013) views a Specific Learning Disorder in reading as reflecting a neurodevelopmental disorder of biological origin. This disability is manifested in reading performance markedly below age level that is not attributed to intellectual, developmental, neurological disorders, and poor instruction. This broad category includes more specific deficits referred to as dyslexia.
The incidence of dyslexia in the public schools has been reported to vary between
5% to 17% (McCandliss & Noble, 2003); although more conservative estimated prevalence rates range from 5% to 7% of the general population.
As indicated in the DSM-5, learning disorders in reading are assumed to have a biological base. In terms of studying the development progression of neurological dysfunction related to dyslexia or RD, a favored methodology is functional magnetic resonance imaging (fMRI). Richlan, Kronbichler, and Wimmer (2011) performed a meta-analysis of fMRI studies that included children or adults as a function of controls and dyslexics. Their synthesis suggested that the left occipital temporal and the temporo–parietal regions showed a hypo-activation in adults with dyslexia whereas a hypo-activation was observed in the anterior portion of the left occipito–temporal cortex for dyslexic children. Richlan’s (2012) reanalysis of the data also found little support for the assumption that standard neural anatomical models of developmental dyslexia are localized to problems primarily related to phonological decoding deficits in the left temporo–parietal regions (see also Paulesu, Danelli, & Berlingeri, 2014). Rather, Richlan found evidence that points to dysfunction in dyslexics in the left hemisphere reflecting a larger reading network that included the “under activation” of the occipital–temporal, inferior frontal and inferior parietal regions.
Working Memory (WM) and related Executive Function (EF) deficits
Working Memory and short-term memory
Some of the correlates of aforementioned neurological inefficiencies (under activation) are manifested as difficulties related to Working Memory (WM) and short-term memory (STM). Working Memory is defined as a processing resource of limited capacity, involved in the preservation of information while simultan - eously processing the same or other information (e.g., Baddeley & Logie, 1999;
Engle, Tuholski, Laughlin, & Conway, 1999). Tasks that measure WM assess an individual’s ability to maintain task-relevant information in an active state and to regulate controlled processing. WM tasks are those that require some inference, transformation and/or monitoring of relevant and irrelevant information (Baddeley
& Logie, 1999; Engle et al., 1999). WM tasks typically engage participants in at least two activities after initial encoding: (1) a response to a question or questions about the material or related material to be retrieved; and (2) a response to recall item information that increases in set size. The first part of the task is a distractor of initial encoding of items whereas the second part tests storage.
In contrast, tasks that measure STM typically involve situations that do not vary their initial encoding. That is, participants are not instructed to infer, transform or vary processing requirements. Although WM or complex span tasks share the same processes (e.g., rehearsal, updating, controlled search) as STM or simple span tasks, simple tasks have a greater reliance on phonological processes than WM or complex span tasks (see Unsworth & Engle, 2007, pp. 1045–1046, for a review). In contrast,
complex span tasks are highly associated with attention in a number of WM models.
These models consider the control of attention (e.g., Engle & Kane, 2004) as well as the scope and focus of attention (e.g., Cowan, Fristoe, Elliott, Brunner, & Saults, 2006). Clearly both WM and STM tasks involve sharing some common activities on the participant’s part. For example, both STM and WM tasks invoke controlled processes such as rehearsal. However, controlled processing on WM tasks emerges in the context of high demands on attention (e.g., maintaining a memory trace in the face of interference) and the drawing of resources from the executive system (see Engle et al., 1999, pp. 311–312, for discussion). In contrast, controlled pro - cessing on STM tasks attempts to maintain memory traces above some critical threshold (Cowan, 2005). This maintenance does not directly draw resources from the central executive system.
Theoretical model
A theoretical model that provides a descriptive account of STM and WM problems in participants with RD is the multicomponent model of Baddeley and Logie (1999).
In this model, WM comprised a central executive system that interacts with a set of two subsidiary storage systems: The speech-based phonological loop and the visual–spatial sketchpad. The phonological loop is responsible for the temporary storage of verbal information; items are held within a phonological store of limited duration, and the items are maintained within the store through the process of subvocal articulation. The phonological loop is associated with STM because it involves two major components discussed in the STM literature: a speech-based phonological input store and a rehearsal process. The visual–spatial sketchpad is responsible for the storage of visual–spatial information over brief periods of time and plays a key role in the generation and manipulation of mental images. The central executive is involved in the control and regulation of the WM system.
According to Baddeley (Baddeley, 2012), the central executive coordinates the two subordinate systems, focuses and switches attention, and activates representations within long-term memory (LTM). This model has been revised to include an episodic buffer (Baddeley, 2012), but support for the tripartite model has been found across various age groups of average achieving children (e.g., Alloway, Gathercole, Willis, & Adams, 2004; Gathercole, Pickering, Ambridge, & Wearing, 2004) as well as those with RD (Swanson, Kudo, & Guzman-Orth, 2016).
There are correlates in the neuropsychological literature that complement the tripartite structure, suggesting that some functional independence exists among the systems (e.g., Jonides, 2000). Functional magnetic resonance imaging (fMRI) studies suggest separate neural circuitry for the storage and rehearsal components of both the phonological and the visual–spatial system, with phonological system activity mainly located in the left hemisphere and visual–spatial system activity located primarily in the right hemisphere (e.g., Smith & Jonides, 1997). Executive control processes, on the other hand, are associated primarily with the prefrontal cortex (e.g., Nee, Brown, Askren, Berman, Demiralp, Krawitx, & Jonides, 2013).
In Baddeley’s model (2012), the phonological loop is specialized for the retention of verbal information over short periods of time. It is composed of both a phonological store, which holds information in phonological form, and a rehearsal process, which serves to maintain representations in the phonological store (see Baddeley, Gathercole, & Papagano, 1998, for an extensive review). Thus, the ability to retain and access phonological representations has been associated with verbal STM—but more specifically the phonological loop. Several recent studies suggest that deficits in the phonological loop deficit may lie at the root of word learning problems in children with RD (e.g., Melby-Lervåg, Lyster, & Hulme, 2012). Within the context of Baddeley’s model, a substantial number of studies support the notion that children with RD experience deficits in phonological processing (e.g., see Siegel
& Mazabel, 2013, for review), such as forming or accessing phono logical represen - tations of information. This difficulty in forming and accessing phonological representations impairs their ability to retrieve verbal information from STM.
Additional studies that have also implicated deficits in executive processing for individuals with RD, particularly as it applies to controlled attention (Swanson, 1993; Swanson & Ashbaker, 2000; Wang & Gathercole, 2013). Controlled attention is defined as the capacity to maintain and hold relevant information in “the face of interference or distraction” (Engle et al., 1999, p. 104). The deficits in controlled attention as applied to RD is inferred from three experimental outcomes: (a) poor performance on complex divided attention tasks; (b) poor monitoring, such as an inability to suppress (inhibit) irrelevant information; and (c) depressed performance across verbal and visual–spatial tasks that require concurrent storage and processing (see Booth, Boyle & Kelly, 2010; Swanson, 2006, 2011, for review of these studies).
This problem in executive processing is particularly manifested in individuals with normal intelligence but with serious reading comprehension difficulties. Individuals with reading comprehension impairments can recognize words accurately, but have problems understanding the meaning of what they have read in terms of accuracy and speed. Although there are no population-based studies of this disorder, individual studies (e.g., Nation, Adams, Bowyer-Crane, & Snowling, 1999) suggest that approximately 10% of samples of children with reading problems have reading comprehension difficulties (Snowling & Hulme, 2012).
Synthesis of experimental literature
In order to understand the magnitude of memory problems in individuals with RD, a review of the quantitative literature is necessary. Swanson, Zheng, and Jerman (2009) synthesized published research that compared children with and without RD on measures of the phonological loop (STM) and the executive system of WM (tasks that included simultaneous processing and storage). Effect sizes were computed utilizing Cohen’s (1988) d index.Calculating the d indexfor any study involves dividing the difference between the two group means by either their average standard deviation or the standard deviation of the control group. To make
ds more interpretable, statisticians have usually adopted Cohen’s (1988) system for classifying ds in terms of their size (i.e., .00–.19 is described as TRIVIAL; .20–.49, SMALL; .50–.79, MODERATE; .80 or higher, LARGE). Cohen’s d(1988) is computed as d= Mean of children with RD – Mean of children without RD/
average of standard deviation for both groups.
In general, the Swanson et al. (2009) synthesis computed 578 effect sizes (ESs) on STM and WM measures across a broad range of age, reading, and IQ scores, yielding a mean ES across studies of –.89 (SD = 1.03) in favor of children without RD. 257 ESs were in the moderate range for STM measures (M = –.61, 95%
confidence range of –.65 to –.58) and 320 ESs were in the moderate range for WM measures (M = –.67, 95% confidence range of –.68 to –.64). Table 6.1 illustrates that ESs varied as a function of the type of measure. The results indicated that children with RD were distinctively disadvantaged compared to average readers on (a) STM measures requiring the recall of phonemes and digit sequences and (b) WM measures requiring the simultaneous processing and storage of digits within sentence sequences and final words from unrelated sentences. A mixed regression analysis that took into consideration the variance within and across studies yielded no significant moderating effects related to age, IQ, or reading level on memory effect sizes.
TABLE 6.1 Effect sizes for STM and WM measures as a function of children with and without reading disabilities
Category M SD K Weighted SE 95% CI for
effect size effect size Lower Upper Short-term memory
1. Phonological –0.83a 1.15 22 –0.39 0.05 –0.50 –0.29
2. Words –0.50 0.66 76 –0.55 0.03 –0.61 –0.48
3. Digits –1.49 2.2 55 –0.63 0.03 –0.69 –0.56
4. Letters –1.06 0.52 13 –1.10 0.07 –1.24 –0.95
Working Memory—D & C format
5. Counting span –0.88 0.55 32 –0.78 0.03 –0.84 –0.73
6. Listen/sentence –1.51 1.21 57 –0.84 0.03 –0.89 –0.79
7. Visual-matrix –0.69 0.63 72 –0.80 0.03 –0.86 –0.74
8. Complex visual. –0.52 0.17 20 –0.48 0.05 –0.57 –0.39
9. Concep. span –0.81 0.44 31 –0.37 0.04 –0.44 –0.30
10. Digit/sentence –1.47 2.25 24 –0.58 0.05 –0.68 –0.48
a = Negative ES in favor of children without RD; CI = confidence interval; Concep. = conceptual span; D & C = Daneman and Carpenter task format; K = number of dependent measures; Visual = Visual-matrix
Table adapted from Swanson, Zheng, & Jerman, 2009.
Related cognitive difficulties (cognitive profile)
Obviously, STM and WM are not the only processes implicated in RD. A recent meta-analysis by Kudo, Lussier, and Swanson (2015) analyzed comparative studies (children with RD vs. children without RD) in English journals that were published between 1963 and 2010. The average standard scores across studies on classifi- cation measures (e.g., reading, math, spelling) are shown in Table 6.2. The search nar rowed down from 9,719 articles to 485 actual studies that were data based. The 485 potential studies were further evaluated as to whether they met the follow- ing criteria: (1) children with defined as RD (e.g., reading disabilities, dyslexia, Specific Learning Disorders in reading) were compared to children without RD (i.e., no indication of a learning or behavior deficit) in grades 1 to 12 (age 5 to 18);
(2) within the RD group, at least one RD subgroup has no reported comorbidity (e.g., math disabilities, ADHD); and (3) each study reported a mean score on a standardized (norm referenced) measure of reading and intelligence for each com - parison group (e.g., Wechsler tests or selected subtests). Thirty-five studies met these criteria, yielding 568 ESs with an overall mean weighted ES (ES) of .79 (SE = .01).
As expected, large mean ESs occurred on several achievement measures besides reading in areas such as math (M = 1.28), vocabulary (M = .80), spelling (M = 1.14), and writing (M = 1.01). More importantly, this study syn thesized the cognitive process differences between children with and without RD.
The study addressed two issues. The first focused on identifying those measures of cognition that yielded large differences between children with and without RD.
The synthesis found that moderate to high mean ESs (corrected for sample size)
TABLE 6.2 Psychological assessments for children with and without RD on normed tests Reading disabled Average achieving Effect size
(N = 595) (N = 605)
Variable Mean SD Mean SD Mean SD
Reading compreh. 82.61 6.57 107.82 11.00 1.87 1.11
Reading recognition 82.82 5.02 105.69 7.24 2.41 .63
IQ general 100.19 12.66 107.56 6.67 .45 .54
IQ verbal 95.83 11.63 109.04 6.79 .89 .59
Fluency/RAN 99.19 23.31 118.81 33.74 1.04 .73
Phonological processing 86.50 14.72 101.65 15.54 .94 .47
Word attack 82.54 3.26 102.26 5.01 1.82 .65
Math 92.26 9.32 103.09 9.58 .88 .56
Vocabulary 98.45 6.93 105.71 3.11 .52 .44
Spelling 82.16 4.67 105.42 6.28 1.89 .43
The reporting of performance on various measures is in standard scores (Mean= 100, SD = 15).
A standard score below 85 is one standard deviation below the mean. Compreh. = Comprehension;
RAN = Rapid naming of Letter or Numbers Table adapted from Kudo et al., 2015.
in favor of children without RD emerged on measures of cognition related to rapid naming speed (M = .89), phonological awareness (M = 1.00), verbal Working Memory (M = .68), and verbal short-term memory (M = .64).
The second issue focused on whether performance differences (effect sizes) between children with and without RD were moderated by age, IQ levels, ethnicity, or gender. Few eligible studies reported on the specific effects of ethnicity and gender and, therefore, the influence of these variables on performance outcomes could not be pursued further. However, an important finding was that chrono - logical age was not a significant moderator of the overall ES between children with and without RD. In contrast, IQ moderated the overall outcomes, even when competing measures (e.g., the cognitive and academic measures that had been found to be significant in other studies) were entered into the hierarchical linear modeling.
There is also a parallel to these findings on children with adults. Swanson and Hsieh (2009) found that 52 studies that met inclusion criteria yielded 776 ESs comparing adults with and without RD with an overall effect size (ES) of .72 (positive ESs in favor of adults without RD, SD = .54).
Three important findings emerged in this quantitative synthesis. First, the trend found for children with RD on cognitive measures ( Johnson, Humphrey, Mellard, Woods, & Swanson, 2010; Kudo et al., 2015), also occurred for adults with RD on cognitive measures of naming speed (M = .96), phonological awareness (M = .87) and verbal memory (M = .62). The results on the comparative measures (i.e., those not used as part of the classification criteria) also yielded high to moderate ESs in favor of adults without RD on measures of math (M = .75), vocabulary (M = .71), spelling (M = 1.57), and writing (M = .72).
Second, the synthesis also found that age, as well as gender ratio, were unrelated to the magnitude of ESs when the influence of all other classification variables was partialed out in the analysis.
Finally, ESs varied as a function of severity in RD and intellectual level. The key findings on this issue were that larger ESs in favor of adults without RD when compared to adults with RD occurred when the RD sample had relatively high IQs and reading scores below the 25th percentile.
Neurological profile
Comprehensive meta-analyses of cognitive and fMRI research studies comparing individuals with dyslexia with their average reading counterparts found that four cognitive variables across a wide age range played a major role in differentiating the two groups: phonological awareness, rapid naming speed, verbal STM and verbal WM. In addition, recent syntheses of the fMRI literature suggest there is a functional interaction failure between distinct brain regions that subserve diverse cognitive operations needed for reading in dyslexics. The literature suggests there are distinct regions (e.g., the left temporo–parietal regions) that have greater functional connectivity in normal controls than dyslexic samples, providing support for the notion that dyslexia (or RD) is a biologically based disorder.
Impact of disorder in daily functioning
The daily and social implications of children with RD as they move into adulthood are not well researched. This is because the criteria for defining adults with RD vary significantly across research studies, postsecondary educational institutions, and employment. Although several studies have identified some of the relationships between social variables among adults with low literacy (e.g., MacArthur, Konold, Glutting, & Alamprese, 2010; Sabatini, Sawaki, Shore, & Scarborough, 2010), the construct validity of reading assessment (e.g., measures, profiles) for adults with significant RD, has not been established (e.g., Fletcher, 2010).
As indicated in Gregg’s (2011, 2013) review, few studies exist to compre hensively inform our understanding of the adult populations with RD. Gregg’s review however, does point out that adults with RD are at risk across a number of dimensions related to life skills. For example, there are greater risk for adults with RD compared to nondisabled peers as evidenced by their high drop-out rates (Newman, Wagner, Cameto, & Knokey, 2009; Rojewski, Lee & Gregg, 2014, 2015), lower postsecondary enrollment and attainment (Wagner et al., 2005), restricted labor force participation and lower earnings (Day & Newburger, 2002).
Current debates
Although several debates related to RD have been discussed in the literature (Swanson, Harris, & Graham, 2013), this chapter focuses on just two. These two debates focus on the failure to link cognitive profiles to intervention and whether IQ is necessary in defining RD.
In terms of the first issue (the relationship between the cognitive characteristics of RD and treatment outcomes), research in the field of RD reflects a dichotomy between those who study RD from a cognitive neuroscience perspective and those that focus primarily on intervention (referred to as Response to Intervention, RtI).
A number of studies have suggested that information related to the cognitive characteristics of the sample with RD have minimal influence on treatment procedures (e.g., Burns, Petersen-Brown, Haegele, Rodriguez, Schmitt, Cooper,
& Van Der Hayden, 2016; Stuebing, Fletcher, Branum-Martin, & Francis, 2012).
Likewise, some studies have suggested that RtI fails to reduce the gap between children at risk for RD and those not at risk (Balu, Zhu, Doolittle, Schiller, Jenkins,
& Gersten, 2015; Tran, Sanchez, Arrelano, & Swanson, 2011). Clearly, before any potential advances in the field are to occur there is a need to consider how to integrate the positive outcomes of the two approaches.
The second issue relates to the role of IQ in the diagnosis of RD. Some studies suggest that IQ is irrelevant to defining RD (e.g., Stanovich & Siegel, 1994; see Hoskyn & Swanson, 2000; for review). These conclusions are based on findings showing that poor readers with low IQ do not differ significantly from children with RD on cognitive measures (see Hoskyn & Swanson, 2000, for review). There is, however, a paradox in these findings. These studies fail to explain how