ArticlePDF Available

Abstract and Figures

The purpose of this study was to determine the relation between reading and working memory (WM) in the context of three major theories: the domain-specificity theory (debate) of WM, the intrinsic cognitive load theory, and the dual process theory. A meta-analysis of 197 studies with 2026 effect sizes found a significant moderate correlation between reading and WM, r = .29, 95% CI [.27, .31]. Moderation analyses indicated that after controlling for publication type, bilingual status, domains of WM, and grade level, the relation between WM and reading was not affected by types of reading. The effects of WM domains were associated with grade level: before 4th grade, different domains of WM were related to reading to a similar degree, whereas verbal WM showed the strongest relations with reading at or beyond 4th grade. Further, the effect of WM on reading comprehension was partialled out when decoding and vocabulary were controlled for. Taken together, the findings are generally compatible with aspects of the domain-specificity theory of WM and the dual process theory, but, importantly, add a developmental component that is not currently reflected in models of the relation between reading and WM. The findings suggest that the domain-general central executive of WM is implicated in early reading acquisition, and verbal WM is more strongly implicated in later reading performance as readers gain more experience with reading. The implications of these findings for reading instruction and WM training are also discussed.
Content may be subject to copyright.
A Meta-Analysis on the Relation Between Reading and Working Memory
Peng Peng
University of Nebraska, Lincoln
Marcia Barnes
University of Texas, Austin
CuiCui Wang
Beijing Normal University
Wei Wang
University of Central Florida
Shan Li
Beijing Normal University
H. Lee Swanson
University of California, Riverside
William Dardick
George Washington University
Sha Tao
Beijing Normal University
The purpose of this study was to determine the relation between reading and working memory (WM) in the
context of 3 major theories: the domain-specificity theory (debate) of WM, the intrinsic cognitive load theory,
and the dual process theory. A meta-analysis of 197 studies with 2026 effect sizes found a significant moderate
correlation between reading and WM, r.29, 95% CI [.27, .31]. Moderation analyses indicated that after
controlling for publication type, bilingual status, domains of WM, and grade level, the relation between WM
and reading was not affected by types of reading. The effects of WM domains were associated with grade
level: before 4th grade, different domains of WM were related to reading to a similar degree, whereas verbal
WM showed the strongest relations with reading at or beyond 4th grade. Further, the effect of WM on reading
comprehension was partialed out when decoding and vocabulary were controlled for. Taken together, the
findings are generally compatible with aspects of the domain-specificity theory of WM and the dual process
theory, but, importantly, add a developmental component that is not currently reflected in models of the
relation between reading and WM. The findings suggest that the domain-general central executive of WM is
implicated in early reading acquisition, and verbal WM is more strongly implicated in later reading perfor-
mance as readers gain more experience with reading. The implications of these findings for reading instruction
and WM training are also discussed.
Public Significance Statement
This study showed that working memory has moderate relations with reading, and these relations are as
strong for more foundational reading skills as they are for comprehension. More importantly, these
relations may vary as a function of development: working memory primarily exerts an impact on reading
early on, with reading also shaping the further development of verbal working memory in particular.
Keywords: reading, working memory, dual process theory, grade level, cognitive load
Supplemental materials: http://dx.doi.org/10.1037/bul0000124.supp
This article was published Online First October 30, 2017.
Peng Peng, Department of Special Education and Communication Dis-
orders, College of Education and Human Sciences, University of Nebraska,
Lincoln; Marcia Barnes, Department of Special Education, University of
Texas, Austin; CuiCui Wang, State Key Laboratory of Cognitive Neuro-
science and Learning and IDG/McGovern Institute for Brain Research,
Beijing Normal University; Wei Wang, Department of Psychology, Uni-
versity of Central Florida; Shan Li, State Key Laboratory of Cognitive
Neuroscience and Learning and IDG/McGovern Institute for Brain Re-
search, Beijing Normal University; H. Lee Swanson, Graduate School of
Education, University of California, Riverside; William Dardick, Depart-
ment of Educational Leadership, George Washington University; Sha Tao,
State Key Laboratory of Cognitive Neuroscience and Learning and IDG/
McGovern Institute for Brain Research, Beijing Normal University.
We thank Lee Branum-Martin at Georgia State University and Emily
Tanner-Smith at University of Oregon for their suggestions on the statistics
analysis in this paper.
Correspondence concerning this article should be addressed to Peng Peng,
Department of Special Education and Communication Disorders, College of
Education and Human Sciences, University of Nebraska, Lincoln, 301 Barkley
Memorial Center, P.O. Box 830738, Lincoln, NE 68583-0738, or to Sha Tao,
State Key Laboratory of Cognitive Neuroscience and Learning and IDG/
McGovern Institute for Brain Research, Beijing Normal University, China.
E-mail: kevpp2004@hotmail.com or taosha@bnu.edu.cn
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Psychological Bulletin © 2017 American Psychological Association
2018, Vol. 144, No. 1, 48–76 0033-2909/18/$12.00 http://dx.doi.org/10.1037/bul0000124
48
Working memory (WM) refers to the capacity to store informa-
tion for short periods of time while engaging in cognitively de-
manding activities (Baddeley, 1986). In contrast to short-term
memory (STM), which is a passive information storage system,
WM is the system responsible for storing and processing informa-
tion simultaneously (e.g., see Miyake & Shah, 1999, for review).
Theoretically, WM plays an important role in reading performance
because many reading tasks involve simultaneous information
processing and storage. For example, to comprehend a text, indi-
viduals first visually process the words; then match the words to
the phonological, orthographic, and semantic representations in
long-term memory; and finally combine these representations with
the context to construct an understanding of the passage. WM is
purported to be involved in this process by keeping relevant
information in STM, retrieving information from long-term mem-
ory, and integrating all sources of information to form an accurate
representation of the situation described by the text (van den
Broek, Mouw, & Kraal, 2016). Despite the theoretically strong
relation between reading and WM, substantial differences have
been reported across studies in the size of this relation. Some
studies have found that the contribution of WM to reading is
negligible, with R
2
around 0 (e.g., Koltum, 2003;O’Shaughnessy
& Swanson, 2000), whereas other studies have found that WM
almost fully explained reading performance, with R
2
values rang-
ing from .60 to .81 (e.g., Daneman, 1991;Weissinger, 2013;
McIntyre, 2015).
It is important to gain better insight into the degree to which
WM is related to reading performance and the factors that influ-
ence this relation. To our knowledge, only two prior reviews have
investigated these issues (i.e., Daneman & Merikle, 1996;Savage,
Lavers, & Pillay, 2007). Specifically, Savage et al. (2007) con-
ducted a comprehensive literature review on the relation between
reading and WM. They concluded that (a) in comparison to foun-
dational reading skills such as decoding, WM is more important
for advanced reading skills such as reading comprehension and (b)
whether the relation between reading and WM is influenced by
domains of WM (i.e., verbal vs. visuospatial) remains unclear. The
Savage et al. (2007) study is a comprehensive literature review. A
systematic meta-analytic investigation of the studies cited in that
review as well as studies published afterward would be important
for replicating the findings as well as for addressing questions that
review could not answer.
In contrast to Savage et al. (2007);Daneman and Merikle (1996)
used the meta-analytic method to systematically investigate the
relation between WM and language comprehension. Based on the
inclusion of 77 studies, Daneman and Merikle (1996) found that
compared with STM, WM showed stronger correlations (r
.30 –.52) with comprehension indexed by vocabulary, listening
comprehension, and reading comprehension. Furthermore, verbal
WM showed stronger correlations with comprehension, compared
to verbal/numerical WM or numerical WM. Although Daneman
and Merikle (1996) adopted the meta-analytic method, they did not
investigate specific relations between WM and different compre-
hension skills (i.e., vocabulary, listening comprehension, and read-
ing comprehension). Nor did they analyze relations between WM
and foundational reading skills such as phonological coding and
decoding. Relations of visuospatial WM and comprehension were
not investigated, likely reflecting the lack of primary data to
analyze at the time.
There are two additional important issues that have not been
sufficiently investigated in the previous review and meta-analysis.
These are whether WM makes unique contributions to reading
comprehension when relevant foundational reading skills are con-
trolled for, and whether the relations of WM and various reading
components change across development. The present study aims to
replicate previous findings with an updated corpus of studies as
well as address the questions mentioned above that previous stud-
ies did not or were unable to answer. Specifically, we use the
meta-analytic method to systematically investigate the relation
between reading and WM with a focus on several moderators that
might explain variations in this relation. These moderators include
domains of WM (i.e., verbal WM, numerical WM, visuospatial
WM), types of reading (i.e., phonological coding, decoding, vo-
cabulary, comprehension), and grade level (i.e., before 4th grade
vs. at/beyond 4th grade, reflecting the transition from learning-to-
read to reading-to-learn; Christopher et al., 2012). Moreover, we
examine the unique contributions of WM to reading comprehen-
sion after controlling for decoding and vocabulary.
From a theoretical perspective, the present meta-analysis con-
siders the relation between reading and WM in the context of
several cognitive theories. By examining whether reading is re-
lated to WM in different domains (i.e., verbal WM, numerical
WM, and visuospatial WM), the findings may contribute to the
theoretical debate on whether WM’s relation to reading is domain-
general (Baddeley, 1986;Engle, 2002;Engle & Kane, 2003)or
domain-specific (Daneman & Carpenter, 1980;Ericsson &
Kintsch, 1995;Friedman & Miyake, 2004). The investigation of
how WM is involved in different reading skills from a develop-
mental perspective addresses two general cognitive theories: In-
trinsic cognitive load theory and dual process theory; that is,
whether the relation between reading and WM is mostly affected
by the complexity of reading tasks, as suggested by the intrinsic
cognitive load theory (Chandler & Sweller, 1991;Sweller, 1994),
or whether reading experience influences the degrees of WM
involvement in reading, as implied by the dual process theory
(Evans & Stanovich, 2013).
From a practical perspective, findings from the present meta-
analysis may have implications for reading instruction and WM
intervention. Specifically, many cognitive skills have been related
to reading, but all of them may not be equally important. Meta-
analysis is one way to obtain relatively accurate estimates of the
size of the relation between two variables, which may be useful for
thinking about whether and under what conditions WM can be
considered a “pressure point” for reading (Perfetti & Stafura,
2014). Moreover, a number of studies in recent years have inves-
tigated whether training WM has effects on WM outcomes, as well
as transfer effects on academic skills such as reading (e.g., Dahlin,
2011;Morrison & Chein, 2011;Peng & Fuchs, 2017). Although
some of these studies found effects of WM training on WM
outcomes, most failed to find far-transfer effects to academic skills
(e.g., Jacob & Parkinson, 2015;Melby-Lervåg & Hulme, 2013;
Shipstead, Redick, & Engle, 2012). Several questions have been
posed regarding the potential of WM training to increase academic
skills, including (a) whether training in different domains of WM
(e.g., verbal, numerical, or visuospatial) is appropriate for the type
of academic skill being targeted (Peng & Fuchs, 2017); (b)
whether the academic target of WM training (i.e., more complex
aspects of reading such as reading comprehension or foundational
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
49
READING AND WORKING MEMORY: A META-ANALYSIS
aspects of reading such as decoding) makes a difference (e.g.,
Dahlin, 2011;Peng & Fuchs, 2017); (c) whether certain child-level
characteristics (e.g., age and/or academic skill-level) or features of
the training itself (e.g., intensity and duration of training) moderate
training effects (Shipstead et al., 2012;Wang, Zhou, & Shah,
2014); and (d) whether cognitive training is delivered on its own or
combined with academic skills interventions might be important
for facilitating transfer to academic skills (e.g., Barnes et al.,
2016). Meta-analysis provides one type of empirical correlational
evidence that may help inform some of these current questions in
WM and reading intervention research. In the following sections,
we describe various theories of WM and associated hypotheses
regarding the relation between reading and WM.
Domain-Specificity Theory (Debate) of
Working Memory
Although many researchers acknowledge that WM tasks mea-
sure the ability to simultaneously process and store information
(Miyake & Shah, 1999), there are several ongoing debates about
the nature of WM tasks in different domains (i.e., verbal WM,
numerical WM, visuospatial WM). One is whether WM tasks that
use materials from different domains measure a domain-general
WM ability or domain-specific WM abilities. The domain-general
WM model proposed by Baddeley (1986) contends that WM
consists of two “slave systems” that are responsible for short-term
maintenance of domain-specific (i.e., verbal, numerical, visuospa-
tial) information and a central executive that coordinates the on-
going processing and storage of information in the slave systems.
The central executive directs attention to relevant information,
suppressing irrelevant information and inappropriate actions, and it
coordinates cognitive processes when more than one task must be
accomplished simultaneously. It also differentiates WM from STM
and, for this reason more than any other, the central executive is
considered by many to be the core component of WM or to
represent the construct of WM (e.g., Engle, 2002;Engle & Kane,
2003). Based on domain-general views of WM, the relation be-
tween reading and WM should not be influenced by the domain of
WM that is being measured.
However, other researchers claim that the operation of WM
depends on domain knowledge, and thus is strongly affected by
domain specificity (Ericsson & Kintsch, 1995). According to Er-
icsson and Kintsch (1995), long-term memory can supplement or
facilitate WM. When individuals are knowledgeable in a particular
domain, they can process (encode and retrieve) information in that
domain more efficiently than information in domains they are less
knowledgeable about (Ericsson & Kintsch, 1995). In accordance
with this view, WM integrates domain-specific skills, knowledge,
and procedures to meet the particular demands of learning tasks
within a particular domain. Thus, the relation between reading and
WM may be influenced by the domain of WM such that compared
to visuospatial WM, verbal WM would be expected to show a
stronger relation with reading.
There is evidence to support both domain-general and domain-
specific hypotheses on the relation between reading and WM.
Some studies have not found that the domain of the WM task
(numerical, verbal, visuospatial, and tasks involving at least two
domains) is differentially related to reading (e.g., Swanson &
Alexander, 1997;Swanson & Berninger, 1995;Karasinski & Ellis
Weismer, 2010), whereas other studies have found verbal WM to
be more strongly related to reading than visuospatial WM (e.g.,
Oakhill, Yuill, & Garnham, 2011;Zheng, 2009). In the Daneman
and Merikle (1996) meta-analysis, verbal WM was more strongly
related to reading than was numerical WM.
Development may contribute to the understanding of the
domain-specificity theory (debate) of WM. In early grades,
domain-specific knowledge, specifically verbal knowledge, is not
as well developed and represented in long-term memory as in later
grades. Thus, younger readers may rely less on retrieval of verbal
knowledge from long-term memory to help accomplish reading
tasks, and rely more heavily on WM, especially the domain-
general component of WM (e.g., the central executive) to perform
reading tasks. In contrast, in later grades, individuals have acquired
a strong knowledge base of verbal knowledge in long-term mem-
ory. The retrieval of verbal knowledge in long-term memory and
its integration with language-based information during reading
may specifically requires verbal WM. Thus, in the present study,
we further investigated the variability in the relations between
reading and different domains of WM, and whether the moderating
effect of WM domain is affected by grade level.
Relation Between Reading and Working Memory
In addition to the domain-specific/domain-general questions
about the nature of the relation of WM to reading above, the
relation between reading and WM has often been investigated and
interpreted within different WM models. The componential WM
model (Baddeley, 1986), as mentioned earlier, suggests that the
phonological loop and the central executive are especially impor-
tant to reading. The phonological loop plays an important role in
word learning, helping children establish the grapheme-phoneme
connection (Baddeley, 1979;McDougall, Hulme, Ellis, & Monk,
1994). The central executive is more critical for reading compre-
hension, coordinating and integrating information read from texts
(e.g., comprehension monitoring; making inferences; e.g., Cain,
Oakhill, & Bryant, 2004). The resource-sharing WM capacity
model claims that WM is a limited cognitive resource involved in
simultaneous information processing and storage (e.g., Daneman
& Carpenter, 1980;Engle, Cantor, & Carullo, 1992;Just & Car-
penter, 1992). According to this model, WM should show a strong
relation with reading comprehension because reading comprehen-
sion also requires simultaneous information processing and stor-
age. In attempting to understand what is being read, there is a
trade-off between processing and storage such that information
processing efficiency determines the storage capacity of WM that
is available for reading comprehension (Perfetti, 2007). For exam-
ple, inefficient word recognition lessens the amount of additional
information that can be maintained in WM to aid comprehension
during reading (Daneman & Carpenter, 1980). The long-term WM
model, as mentioned earlier, suggests that long-term memory can
supplement or facilitate WM (Ericsson & Kintsch, 1995). The
long-term WM model also emphasizes the important role of WM
in complex skills such as reading comprehension (Ericsson &
Kintsch, 1995). That is, relevant knowledge from long-term mem-
ory can enhance the efficiency of WM during reading comprehen-
sion by allowing individuals to effectively retrieve relevant back-
ground knowledge from long-term memory, keeping knowledge
accessible by means of retrieval cues in STM, and readily inte-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
50 PENG ET AL.
grating this knowledge with information extracted from the text
during reading to achieve the goal of comprehension. This hypoth-
esis is supported by empirical evidence showing that long-term
memory representations such as background knowledge partially
mediated the relation between WM and reading comprehension
(e.g., Hambrick & Engle, 2002;Nation, Adams, Bowyer-Crane, &
Snowling, 1999).
Although there are many commonalities among the WM models
presented above, these models emphasize somewhat different
mechanisms by which reading and WM are related. Both compo-
nential and capacity WM models emphasize that WM is a limited
domain-general construct and limitations in WM create bottle-
necks for reading (Baddeley, 1986;Engle et al., 1992). In contrast,
the long-term WM model emphasizes that knowledge in long-term
memory determines how WM is recruited during reading, suggest-
ing that long-term memory plays a more important role than WM
in reading (Ericsson & Kintsch, 1995). Also, almost all these WM
models focus on the importance of WM for complex reading skills
such as reading comprehension (Baddeley, Logie, Nimmo-Smith,
& Brereton, 1985;Cain et al., 2004;Daneman & Carpenter, 1980;
Engle et al., 1992;Ericsson & Kintsch, 1995), with less attention
to whether and how WM is related to other reading skills such as
phonological coding, decoding, and vocabulary (Christopher et al.,
2012). Moreover, although these WM models focus on how WM
influences reading, they are not developmental in nature; that is,
they do not address the possibility that the relation between WM
and reading may vary with development.
In the present study, we investigated the relation between read-
ing and WM using different theoretical perspectives to guide the
choice of variables and moderators that are included in analyses. In
addition to theories of domain-specificity discussed above that
prompted us to examine different domains of WM in relation to
reading, we also drew on the intrinsic cognitive load theory and the
dual process theory to derive hypotheses about relations between
WM and different types of reading and whether these relations
vary as a function of development.
Intrinsic Cognitive Load Theory
Reading is a complex multicomponent construct that is often
conceptualized in strands including (but not limited to) phonolog-
ical coding, decoding, and vocabulary, and comprehension (Na-
tional Reading Panel, 2000). One framework to understand the
relation between reading and WM is the intrinsic cognitive load
theory. According to this theory, there is an inherent level of
difficulty associated with a specific task, which may not be easily
altered by external factors such as instruction and learning expe-
rience (Chandler & Sweller, 1991;Sweller, 1994). That is, tasks
with multiple steps and sequential thinking are assumed to be more
difficult than tasks involving a single step or where the task can be
accomplished through direct retrieval from long-term memory.
Thus, based on the intrinsic cognitive load theory, complex read-
ing tasks such as reading comprehension that have multiple, se-
quential storage and processing features are hypothesized to draw
more WM resources than foundational reading tasks such as pho-
nological coding, decoding, or vocabulary, performance on which
may rely more on retrieval from long-term memory, at least for
more experienced readers for whom most words and their mean-
ings may be recognized or retrieved from lexical memory.
Dual Process Theory
Another framework for conceptualizing the relation between
reading and WM is the dual process theory of higher cognition
(Evans & Stanovich, 2013). This view provides an account of how
task performance is accomplished by two different processing
systems: autonomous processes and controlled processes (Evans &
Stanovich, 2013). Autonomous processing refers to the processing
of domain-specific familiar information that is efficient, seemingly
effortless and takes few cognitive resources, whereas controlled
processing refers to the processing of novel information that needs
conscious and sequential thinking, and requires many high-level
cognitive resources such as reasoning and WM (Bargh, 1994;
Evans & Stanovich, 2013). Based on the dual process theory, the
involvement of WM in a reading task is not just determined by task
complexity, but, more importantly, by how efficiently the reading
task can be performed, which is closely associated with reading
experience. For beginning readers (e.g., readers before 4th grade),
limited reading experience and less well developed verbal knowl-
edge means that their reading skills are less likely to be automated.
Thus, for beginning readers, tasks such as phonological coding and
decoding may require significant WM resources. For more expe-
rienced readers (e.g., readers at/beyond 4th grade), WM may be
less involved in these more foundational aspects of reading given
that word recognition and the retrieval of relevant verbal informa-
tion from long-term memory is likely to be more efficient. How-
ever, more experienced readers would still be expected to draw
heavily on WM resources on complex reading tasks such as
reading comprehension, which become more complex at higher
grades with the inclusion of new academic vocabulary and longer
texts. Thus, grade level, as well as the interaction between types of
reading and grade level, may exert effects on the relation between
reading and WM.
Typically, individual studies only investigate the relation be-
tween WM and one or two reading skills and focus on a single
population (e.g., beginning readers or experienced readers). Such
studies are insufficient on their own to test some of the hypotheses
about the relation of reading and WM that can be generated by the
intrinsic cognitive load or the dual process approaches. In the
current meta-analysis, hypotheses that are generated by these two
theoretical approaches can be addressed by synthesizing studies
across four major reading skills (i.e., phonological coding, decod-
ing, vocabulary, and comprehension) and among populations with
different levels of reading experience. In the following sections,
we discuss empirical findings and hypotheses about the relation of
WM and these different reading skills using the frameworks pro-
vided by the intrinsic cognitive load theory or the dual process
theory.
Phonological Coding
Phonological coding refers to a broad set of phonological pro-
cessing skills such as identifying and manipulating units of oral
language parts (e.g., words, syllables, onsets and rimes, phonemes)
and retrieval efficiency of phonological codes (e.g., rapid letter,
number, or object naming; Torgesen, Wagner, & Rashotte, 1999).
Many studies of reading in English have shown that phonological
coding plays a critical role in early reading development such that
individuals who are better at detecting, manipulating, and retriev-
ing sounds in words learn to decode words more easily (e.g.,
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
51
READING AND WORKING MEMORY: A META-ANALYSIS
Bradley & Bryant, 1978;Stanovich, 1992). Theoretically, WM is
implicated in being able to decompose spoken words into their
constitute syllables, onset-rimes, or phonemes; putting syllables,
onset-rimes, and phonemes together to form words; and effectively
retrieving sounds for sequential displays of symbols and objects.
This is because WM is required for the simultaneous processing
and storage of these phonological representations (e.g., Oakhill &
Kyle, 2000).
Phonological coding tasks show small to moderate relations
with WM (e.g., Hester & Hodson, 2004;Swanson & Alexander,
1997). The size of these relations could be influenced by several
factors. One factor is the complexity of different phonological
coding tasks, as suggested by the intrinsic cognitive load theory.
For example, some researchers suggest that tasks that require the
manipulation of small sound units—phonemes—are more difficult
and thus require more WM resources than tasks with larger sound
units (e.g., syllables; e.g., Cormier & Dea, 1997;Tunmer &
Hoover, 1992). Another potential factor affecting the relation
between phonological coding and WM is grade level. As suggested
by the dual process theory, beginning readers may draw, to a
greater extent than more experienced readers, on WM resources to
perform phonological coding tasks. Experienced readers are
thought to directly retrieve phonological representations from
long-term memory to perform phonological coding tasks, thereby
relying less on WM.
Decoding
Decoding is the ability to translate written language into speech
with accuracy and/or fluency (e.g., Gough & Tunmer, 1986;
Melby-Lervåg & Lervåg, 2014). To decode a word or nonword,
one needs to convert the letters into sounds sequentially and then
blend them sequentially, at least during the initial acquisition of
decoding, which requires the simultaneous processing and stor-
age of letters/sounds. The relation between decoding and WM
may be affected by grade level. For example, beginning readers
who are still in the stage of learning the grapheme-phoneme-
correspondence (GPC) rules cannot apply the GPC rules effi-
ciently in the decoding process. For experienced readers, GPC
rules can be applied more efficiently in the reading of nonwords,
and familiar words are often directly retrieved from long-term
memory based on their orthography (Ehri, 1992).
Another factor that may influence the relation between decoding
and WM is the way in which decoding is measured. Usually,
decoding is measured by either real word and/or nonword reading
tasks. According to the triangle model of word recognition, to
decode a word, the phonology, orthography, and semantics of the
word are often activated, which facilitates word recognition (Se-
idenberg & McClelland, 1989). Because semantics is thought to
facilitate the association between orthography and phonology for
decoding (e.g., Seidenberg & McClelland, 1989;Vellutino, Scan-
lon, & Spearing, 1995), the lack of a semantic representation for
nonwords might make nonword reading more dependent on WM
than real word reading, which would be consistent with the dual
process theory. That being said, nonwords used in decoding tasks
are almost always orthographically regular, which means individ-
uals can read nonwords solely based on GPC rules (Coltheart &
Leahy, 1992), whereas real-word decoding tasks often consist of a
mixture of orthographically regular and irregular words. To read
irregular words (e.g., plumber), GPC rules cannot be consistently
or completely applied and using GPC rules to decode irregular real
words may actually interfere with their pronunciation (e.g., the
letter “b” sounds different in “lumber” and “plumber”). Thus, the
pronunciation of irregular words, particularly low frequency irreg-
ular words that are not easily retrievable from memory, may
require more effortful cognitive processing. Following this logic
and the dual process theory, reading real words, or at least some
types of real words, may require more WM resources than reading
nonwords.
Decoding is also indexed by the accuracy/fluency of sen-
tence/passage reading or word list reading. Compared with
word list reading or the reading of single words, the semantic
associations among words in sentence- or passage-level reading
are likely to facilitate word reading accuracy (Stanovich, 1980).
For example, based on the dual process theory, readers may be
more accurate and faster to read “it is a sunny day and people
are out hiking” than to read a jumbled list of the same words.
Thus, WM may show weaker relations with sentence/passage
reading accuracy/fluency than with word list reading accuracy/
fluency.
Vocabulary
There is also some evidence that WM is related to vocabulary
(e.g., Babayig
˘it, 2015;Tighe, Wagner, & Schatschneider, 2015).
To perform a vocabulary task (either point to a picture correspond-
ing to a word or explain what a word means), WM may be used to
simultaneously process verbal/visual information, activate relevant
background knowledge and concepts, and integrate those sources
of information. Because the retrieval of relevant background
knowledge is especially critical in vocabulary tasks (Echols, West,
Stanovich, & Zehr, 1996), language development, as indexed by
grade level, may play a role in the relation between WM and
vocabulary based on the dual process theory. That is, compared
with experienced readers who often automatically retrieve relevant
knowledge from long-term memory, beginning readers may need
to allocate more of their WM resources to searching in long-term
memory and integrating this more slowly retrieved knowledge
with less familiar phonological and orthographic representations to
perform vocabulary tasks.
The type of vocabulary measure may also affect the relation
between vocabulary and WM. There are two major types of
vocabulary tasks: receptive and expressive. Receptive vocabulary
tasks (e.g., Peabody Picture Vocabulary Test) typically involve the
understanding of spoken, written, or a signed word, tapping the
breadth of vocabulary knowledge (Ouellette, 2006). In contrast,
expressive vocabulary tasks (e.g., Wechsler Abbreviated Scale of
Intelligence: Vocabulary) typically require individuals to express
or produce words (e.g., explain what police is or name the word
depicted in an illustration), tapping the depth of vocabulary knowl-
edge (Ouellette, 2006). Because expressive vocabulary appears to
develop after receptive vocabulary over the course of early lan-
guage development, and some expressive vocabulary tasks may be
more semantically demanding and involve more complex language
output, it is reasonable to assume that, based on the intrinsic
cognitive load theory, performance on expressive vocabulary tasks
should show a stronger relation with WM than that of performance
on receptive vocabulary tasks.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
52 PENG ET AL.
Comprehension
Comprehension in the present study is defined as the ability to
comprehend a passage in either oral format (listening comprehen-
sion) or written format (reading comprehension). Comprehension
is often considered the most complex literacy task that draws on
the coordination of many different reading and cognitive skills,
and thus it is hypothesized, based on the intrinsic cognitive load
theory, to show the strongest relation with WM than the other
components of reading above. Indeed, some research indicates that
WM strongly correlates with reading comprehension even when
other foundational reading skills are controlled for (e.g., Cain et
al., 2004;Christopher et al., 2012). However, there are also studies
reporting that the relation between WM and comprehension is not
strong or significant (e.g., Walczyk & Taylor, 1996;Winke, 2005).
In the present study, we propose four factors that may explain
findings regarding the relations of WM and comprehension: com-
prehension format (listening or reading), whether the reading com-
prehension tasks are timed or not, type of text (expository or
narrative), and grade level. Based on the intrinsic cognitive load
theory, different types of comprehension task and whether the
reading comprehension assessment is time-limited may tax WM to
varying degrees. Compared to listening comprehension, reading
comprehension may draw more on WM because of extra informa-
tion processing such as decoding. However, because all informa-
tion is orally presented in listening comprehension tasks, storage
load requirements in WM during listening may be greater than
those for reading comprehension (i.e., because the pace of pro-
cessing during reading is under the reader’s control and the reader
can reread). In this view, listening comprehension may actually
require more WM resources (e.g., Cain & Bignell, 2014). Like-
wise, timed reading comprehension tasks may involve more WM
than untimed reading comprehension tasks because the time limit
poses a greater cognitive load. Expository text comprehension may
require more WM than narrative text comprehension, because, in
comparison with narrative texts, expository texts often use less
familiar text structures and are informationally dense with less
familiar academic or subject-specific vocabulary (e.g., Best, Floyd,
& McNamara, 2008;McNamara, Graesser, & Louwerse, 2012).
Given the nature of reading development, reading experience
must also be considered when posing hypotheses about the relation
of WM and comprehension. Therefore, grade level should be
considered when studying the effect of aspects or types of com-
prehension on the relation between WM and comprehension. Spe-
cifically, reading development is a cumulative process in which the
reader shifts, with reading experience, from learning-to-read to
reading-to-learn (Chall, 1983;Etmanskie, Partanen, & Siegel,
2016). In the learning-to-read stage (i.e., before about 4th grade)
reading comprehension tasks tax decoding heavily and involve
mostly short to medium length narrative texts (e.g., Keenan, Betje-
mann, & Olson, 2008). Thus, decoding explains the majority of
variance in reading comprehension at this stage (Hoover & Gough,
1990). In contrast, in the reading-to-learn stage (i.e., 4th grade and
beyond), individuals are more fluent in decoding, and reading
comprehension tasks usually include longer and more difficult
expository texts (e.g., Keenan et al., 2008). Particularly, reading
materials beginning at 4th grade become significantly more com-
plex as textbooks serve as the mechanism to gain access to cur-
riculum, content, and knowledge (e.g., storybooks and narrative
texts are replaced with math and social studies textbooks; Etman-
skie et al., 2016). Thus, in this stage, language comprehension
becomes a more important component of reading comprehension
than decoding (Hoover & Gough, 1990).
Considering this reading development sequence, there may be
an interaction between types of comprehension and grade level,
which can be considered in the frameworks of both the intrinsic
cognitive load theory and the dual process theory. Specifically, for
the comparison between reading comprehension and listening
comprehension, reading comprehension may show stronger rela-
tions to WM before 4th grade. This may be because younger
children rely on WM for both decoding and comprehension in
reading comprehension tasks. In contrast, after 4th grade, reading
comprehension and listening comprehension may show compara-
ble relations with WM. This may be because experienced readers
need only allocate WM for comprehension not decoding on read-
ing comprehension task. Likewise, it is likely that before 4th grade,
children are still in the stage of accumulating foundational decod-
ing and vocabulary skills and thus timed reading comprehension
may place more cognitive load than untimed reading comprehen-
sion. In contrast, the cognitive load may not be different between
timed reading comprehension and untimed reading comprehension
at/beyond 4th grade when individuals usually allot most their WM
to comprehension, not to decoding or vocabulary processing. Re-
garding the type of text (expository text comprehension vs. narra-
tive text comprehension), before 4th grade, individuals may find
all comprehension tasks equally challenging due to less well-
developed foundational reading skills such as decoding and vo-
cabulary knowledge. Thus, text type may not influence the relation
between reading comprehension and WM. At/beyond 4th grade,
when individuals have mastered decoding and have more vocab-
ulary knowledge, text type may play a more important role such
that WM show stronger relation with expository text comprehen-
sion versus narrative text comprehension.
The Unique Contribution of Working Memory to
Reading Comprehension
Testing direct relations between different reading skills and
WM adds to our understanding of the relative importance (i.e.,
size) and nature of these relations. It is also important to
investigate how WM is related to advanced reading skills when
we control for relevant foundational reading skills. Addressing
this issue is important for understanding whether different
reading skills shared common/unique variance in WM (e.g.,
Savage et al., 2005). In the current study, we focus on the
unique contributions of WM to reading comprehension. We
chose reading comprehension, because it is the “goal” of read-
ing, the aspect of reading that is assessed in high-stakes reading
tests (National Center for Education Statistics, 2010), and the
most complex reading skill based on the intrinsic cognitive load
theory.
According to the Simple View of Reading (SVR, Hoover &
Gough, 1990), decoding and oral language are two major compo-
nents contributing to reading comprehension. Compared with de-
coding, oral language is a very broad construct including (but not
limited to) vocabulary, background knowledge, verbal reasoning,
and language structures (Scarborough, 2001). Because vocabulary
is often used as a general measure of oral language skills in reading
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
53
READING AND WORKING MEMORY: A META-ANALYSIS
comprehension research (e.g., Cain et al., 2004;Clarke, Snowling,
Truelove, & Hulme, 2010;Ouellette, 2006) and is a major com-
ponent of classroom comprehension instruction (e.g., National
Reading Panel, 2000), we use vocabulary as a proxy for oral
language ability in the analysis testing the unique relation of WM
to comprehension. In these analyses, we examine the unique
contributions of WM to reading comprehension controlling for
both decoding and vocabulary.
Indeed, empirical studies have shown that a large amount
(45% 85%) of the variance in reading comprehension can be
explained by decoding and vocabulary (e.g., Adlof, Catts, &
Little, 2006;Dreyer & Katz, 1992;Hoover & Gough, 1990;
Perfetti, 2007). However, decoding and vocabulary cannot ex-
plain all variance in reading comprehension. It is suggested
that, after controlling for decoding and vocabulary, the residual
variance in reading comprehension may still be explained by
other factors such as WM. The rationale is that in addition to
decoding and vocabulary, reading comprehension also involves
many other linguistic processes (e.g., comprehension monitor-
ing and inferences) that require WM (Perfetti, 2007). Some
studies have shown that after controlling for decoding and
vocabulary, WM still explains unique variance in reading com-
prehension (e.g., Cain et al., 2004), whereas other studies have
found that WM did not make unique contributions to reading
comprehension beyond decoding and vocabulary (Oakhill &
Cain, 2012). Thus, whether WM makes unique contributions to
reading comprehension beyond decoding and vocabulary is still
unclear.
Research Questions
To sum up, this meta-analysis seeks to address four major
questions. First, is there a significant correlation between reading
and WM, and if so, what is the size or strength of this relation?
Second, is the relation between reading and WM affected by
domains of WM, types or components of reading, and grade level?
Third, is there an interaction between domains of WM and grade
level, or between types or components of reading and grade level?
Fourth, is there a significant relation between reading and WM
comprehension after partialing out the effect of decoding and
vocabulary?
Specifically, regarding phonological coding, we investigated
whether WM is differentially related to phoneme manipulations,
syllable manipulations, and rapid naming. We also investigated
whether different aspects of decoding are differentially related to
WM. We focused on the comparison between word recognition
and nonword reading and between sentence/passage reading and
word list reading. We also investigated whether these comparisons
are influenced by grade level. Regarding vocabulary, we investi-
gated whether expressive vocabulary and receptive vocabulary are
related to WM to varying degrees. With respect to comprehension,
we examined whether there is a difference in the relation of WM
to listening comprehension versus reading comprehension, to
timed versus untimed reading comprehension, and to expository
text comprehension versus narrative text comprehension. Interac-
tions between grade level and these various aspects of comprehen-
sion were also tested.
Predictions Based on Domain-Specificity Theory
(Debate) of Working Memory
Based on the domain-general model of WM, the relation be-
tween reading and WM will not be influenced by domains of WM;
however, a domain-specific view of WM would lead us to predict
that verbal WM will be more strongly related to reading than will
visuospatial WM. Moreover, if we consider development, it could
be that in younger children, reading may relate to different do-
mains of WM to a similar degree, but may be more related to
verbal WM among older individuals.
Predictions Based on the Intrinsic Cognitive
Load Theory
Regardless of grade level, WM will be related to different types
of reading to different degrees, such that WM is more closely
related to comprehension than phonological coding, decoding, and
vocabulary. Within these broad aspects of reading, WM may show
a stronger relation with phoneme manipulation than that of syllable
manipulation or rapid naming. Expressive vocabulary may show a
stronger relation with WM versus receptive vocabulary. Reading
comprehension may show a stronger relation with WM compared
to listening comprehension. Performance on timed reading com-
prehension tasks may be more strongly related to WM than per-
formance on untimed reading comprehension tasks. Expository
text comprehension may be more strongly related to WM than is
narrative text comprehension.
Predictions Based on the Dual Process Theory
The relations of WM to various components of reading may
interact with grade level (a marker of reading experience). WM
may be involved to a similar degree in both foundational and
higher level reading skills for younger children. For older individ-
uals whose foundational reading skills may be more efficient, WM
may be more strongly related to higher level reading skills (e.g.,
comprehension) than to foundational reading skills (e.g., decod-
ing).
Regarding the unique contributions of WM to reading compre-
hension, if WM is critical for text-level linguistic processes such
inference-making, comprehension monitoring, and strategy use
during reading comprehension (Perfetti, 2007), we would expect
that WM would make unique contributions to reading comprehen-
sion after partialing out decoding and vocabulary.
Method
Literature Search
Articles for this meta-analysis were identified in two ways.
First, a computer search of the ProQuest, PsycARTICLES, Psy-
cINFO databases for literature was conducted. We used the earliest
possible start date (1964) through October 2015. Titles, abstracts,
and keywords were searched for the following terms: working
memory AND read
, word, decod
, phonolog
, comprehen
, vo-
cabulary, OR language. The terms read
, decod
, phonolog
, and
comprehen
allowed for inclusion of reading, decode/decoding, pho-
nology/phonological, comprehend/comprehension, and so forth. Sec-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
54 PENG ET AL.
ond, we hand searched citations in prior relevant reviews (Dane-
man & Merikle, 1996;Savage et al., 2007). We searched
unpublished literature through Dissertation and Masters Abstract
indexes in ProQuest, Cochrane Database of Systematic Reviews,
relevant conference programs (e.g., Society for the Scientific
Study of Reading, Conference of Society for Research on Educa-
tional Effectiveness, and Annual Conference of American Educa-
tional Research Association), and by e-mailing researchers likely
to have conducted work in this area. We also contacted several
researchers to request correlation tables not provided in their
reported studies. The initial search yielded 69,007 studies. Two
authors of this study then reviewed all studies by titles and ab-
stracts. After excluding the duplicate 86 articles, and 66,557 irrel-
evant articles, the remaining 2361 articles were closely reviewed
using the specific criteria described below (see Figure 1 for the
flow diagram for the search and inclusion criteria for studies in the
present review).
First, studies had to include at least one quantitative task mea-
suring WM and at least one quantitative task measuring reading.
Based on our definitions of measures on WM and reading (not
necessarily consistent what the authors articulated in the primary
studies), we coded all variables of interest for this study. Specif-
ically, WM measures refer to the tasks that tap processing and
storage simultaneously (e.g., complex span tasks and dual-task
WM). Measures that tap specific executive functions, such as
inhibition, switching, or updating, were not considered to be WM
measures in this meta-analysis. Reading measures refer to the tasks
that tap one of the following skills: phonological coding, decoding
(word recognition and nonword reading), vocabulary, listening
comprehension, reading comprehension, and comprehensive read-
ing that tap at least two of the above-mentioned reading skills. To
be considered as a measure of phonological coding, the task must
involve deletion, blending, counting, segmentation, generation,
judgment, position analysis or replacement of phonemes, onset,
rhymes, and/or syllables in words, or rapid naming of letters/
pictures/objects/numbers. To be considered as a decoding mea-
sure, the test had to comprise reading accuracy/efficiency of
words, nonwords, sentences, or passages in either timed or un-
timed condition. To be considered a measure of reading compre-
hension, studies in which a child completed a cloze test about a
passage/sentence or read a passage/sentence and answered ques-
tions in relation to the text were included. To be considered a
measure of listening comprehension, tests that measured under-
standing of heard sentences or passages by means of oral cloze,
answering questions in relation to the orally presented text, or
retelling the orally presented text were included. To be considered
a measure of vocabulary, tests had to measure expressive or
receptive vocabulary using pictures and measures of synonyms or
antonyms were also included. Table 1 demonstrates the examples
of WM measures and reading measures considered in this study.
Second, studies had to report at least one correlation (r) between
any measure of WM and any measure of reading, or the percentage
of variance (R
2
) in reading accounted for by WM only. The
measures of WM and reading used to calculate the direct correla-
Search Records identified through database search (n = 69,007)
Initial Screening
Records after duplicates (n = 89) and non-relevant (n = 66,557)
Records screened (n = 2,361)
Records excluded (n = 66,646)
Eligibility
Studies deemed
potentially eligible for
inclusion
Included 197 studied included in meta-analysis
Studies did not provide correlation
tables or statistics that can be
transformed to correlations.
2,041 studies excluded
Studies measured verbal WM or
Reading in other languages.
46 studies excluded
Studies only included older adults
(over 50 years old)
10 studies excluded
Studies only included sample with
disabilities
67 studies excluded
Figure 1. Flow diagram for the search and inclusion criteria for studies in the present review.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
55
READING AND WORKING MEMORY: A META-ANALYSIS
tion (not partial correlation) had to be taken at the same time point,
because we were interested in the concurrent direct relation be-
tween reading and WM and how this relation was affected by the
moderators proposed in this study.
Third, we only included studies that focused on reading in
English (e.g., provided measures on verbal WM in English and
reading in English) and thus our sample includes both monolingual
and bilingual individuals. We did not include clinical populations,
because the heterogeneity of disability groups is often not system-
atically studied or reported in many primary studies. Also, because
aging significantly influences memory and verbal knowledge, es-
pecially starting around the 50s (Park et al., 2002), we did not
include studies or findings from studies that are based on adults
over 50 years of age.
Coding Procedure and Interrater Reliability
Studies were coded according to the characteristics of partici-
pants and tasks used to measure WM and reading. In addition to
these variables, we also coded the number of participants (N) used
to obtain each correlation. The latter was needed to weight each
effect size, so that correlations obtained from larger samples were
given more weight in the analysis than those obtained from smaller
samples. The important features of individual studies are provided
in the online Appendix.
Variables were discussed until a consensus was reached be-
tween the first and the third authors. Then, both the first author and
the third author used this coding system to conduct the final coding
of all 197 studies independently. The interrater reliability was 1 for
publication type, 1 for bilingual status, .98 for grade level, .97 for
sample size, .95 for correlation coefficients, .97 for domains of
WM, and .95 for types of reading. Any disagreements were re-
solved by consulting the original article or by discussion.
Missing Data
Not all studies provided sufficient information on the variables
of interest for the present study. In case of insufficient information,
Table 1
Description of Codes and Examples of Response Categories for Domains of WM and Types of Reading
Measure Definition Examples of response categories
Domains of WM
Verbal WM Tasks that tap simultaneous process and storage of verbal
information
Complex Reading Span; Sentence Span; Listening
Recall; Alphabet Recoding; Backward Word
Recall; Continuous Paired Associates Task;
Story Retelling; Letter Span; Semantics
Association; Rhyming Span; Animal Dual Task
Performance
Numerical WM Tasks that tap simultaneous process and storage of numerical
information
Backward Digit Span; Calculation Span; Counting
Span; Auditory Digit Sequencing; Composite of
Counting Recall; Backward Digit Recall
Visuospatial WM Tasks that tap simultaneous process and storage of visual or
spatial information
Backward Block Span, Visual Matrix; Mapping
and Directions; Odd-One-Out; Mr. X; Spatial
Recall; Jigsaw Puzzle; Dot Matrix; Visual
Pattern Test
Composite WM Tasks that tap simultaneous process and storage of information
tapping more than one of the following domains: verbal,
numerical, and visuospatial, or composite score of WM
tasks tapping more than one the following domains: verbal,
numerical, and visuospatial
Operation Span, Digit Sentence Span; Composite
of Listening Recall and Backward Digit Recall
and Counting Span; Letter-Number Sequencing;
Stanford-Binet Fifth Edition Working Memory;
Verbal-Spatial Complex Span; Animal-Color
Span
Types of reading
Phonological coding Tasks that tap the ability to identify and manipulate units of
oral language parts (words, syllables, onsets and rimes, and
phonemes) and phonological codes retrieval efficiency
Identify the rhyme of words; Identify initial sounds
or final sounds in words; Identify medial sounds
in words; Segment words into their component
syllable/sound; delete/add sounds from/to words;
Sound blending; Name
letters/digits/colors/objects rapidly
Decoding Tasks that tap the ability to translate written language into
speech with accuracy and fluency
Real word recognition, Non-word reading; reading
word list; Accuracy/fluency of passage/sentence
reading
Vocabulary Tasks that require individuals to point to a picture
corresponding to a word or explain what a word means
Peabody Picture Vocabulary Test; Wechsler
Abbreviated Scale of Intelligence-Vocabulary;
Nelson Reading Skills Test-Vocabulary; Word
production fluency (e.g., say words that start
with letter B); Extended Range Vocabulary Test
Comprehension Tasks that require individuals to comprehend a passage in
either oral format (listening comprehension) or written
format (reading comprehension).
Nelson Denny Reading Comprehension; Woodcock
Reading Mastery Tests-Reading Comprehension;
Gray Oral Reading Comprehension Tests; The
Peabody Individual Achievement Test- Reading
Comprehension
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
56 PENG ET AL.
authors were contacted to obtain the missing information. How-
ever, if missing data could not be retrieved, especially for data
missing for moderator variables, the study was excluded from the
moderator analysis for which data were missing but was included
in all moderator analyses for which data were provided.
Analytic Strategies
The effect size index used for all outcome measures was Pear-
son’s r, the correlation between reading and WM. We considered
all eligible effect sizes in each study. That is, studies could con-
tribute multiple effect sizes as long as the sample for each effect
size was independent. For studies that reported multiple effect
sizes from the same sample, we accounted for the statistical
dependencies using the random effects robust standard error esti-
mation technique developed by Hedges, Tipton, and Johnson
(2010). This analysis allowed for the clustered data (i.e., effect
sizes nested within samples) by correcting the study standard
errors to take into account the correlations between effect sizes
from the same sample. The robust standard error technique re-
quires that an estimate of the mean correlation () between all the
pairs of effect sizes within a cluster be estimated for calculating
the between-study sampling variance estimate,
2
. In all analyses,
we estimated
2
with ␳⫽.80; sensitivity analyses showed that the
findings were robust across different reasonable estimates of .
Analyses were based on Borenstein, Hedges, Higgins, and Roth-
stein’s (2005) recommendations. Specifically, we converted the
correlation coefficients to Fisher’s Zscale, and all analyses were
performed using the transformed values. The results, such as the
summary effect and its confidence interval, were then converted
back to correlation coefficients for presentation. Also, because we
hypothesized that this body of research reports a distribution of
correlation coefficients with significant between-studies variance,
as opposed to a group of studies that attempts to estimate one true
correlation, a random-effects model was appropriate for the current
study (Lipsey & Wilson, 2001). Weighted, random-effects metare-
gression models using Hedges et al.’s (2010) corrections were run
with ROBUMETA in Stata (Hedberg, 2014) to summarize corre-
lation coefficients and to examine potential moderators.
Specifically, we first estimated only the overall weighted mean
correlation between WM and reading. Then, subgroup analyses
were used to examine the relation between reading and WM for
each subgroup of each moderator. Metaregression analyses were
used to examine whether domains of WM, types of reading, and
grade level moderated the relation between reading and WM. For
the moderation analysis, all moderators were entered into the
model simultaneously, with publication type (peer reviewed vs.
other types of publications) and bilingual status as the covariates in
the model as well. Because the moderators were all categorical, we
created dummy coded variables to examine the comparisons
among categories (Cohen, Cohen, West, & Aiken, 2013).
For the analysis on the unique contributions of WM to reading
comprehension, we calculated the partial correlations based on
correlation matrices of studies that provided the correlations
among (a) WM, decoding, and reading comprehension; (b) WM,
vocabulary, and reading comprehension; or (c) WM, decoding,
vocabulary, and reading comprehension. For each study, the cor-
relation matrix was recorded along with the means, standard de-
viations, and number of observations. For example, if a study
reported 2 measures of decoding, 2 measures of reading compre-
hension, and 2 measures of WM, then the full 2 22
correlation matrix was recorded, producing 6 partial correlations
between WM and reading comprehension, partialing out decoding.
We then synthesized these partial correlations to indicate the
unique contribution of WM to reading comprehension, partialing
out decoding or vocabulary or both. Because the way we calcu-
lated partial correlations produces many effect sizes nested within
a sample, we accounted for the statistical dependencies using the
random effects robust standard error estimation technique devel-
oped by Hedges et al. (2010) as mentioned earlier.
Publication bias (the problem of selective publication, in which
the decision to publish a study is influenced by its results) was
examined using the method of Egger, Davey Smith, Schneider, and
Minder (1997) and funnel plot. We did not find significant publi-
cation bias based on Egger et al.’s (1997) publication bias statistics
(i.e., the standard errors of correlations did not significantly predict
correlations among studies with ROBUMETA in Stata, ps.11),
except for the correlation between phonological coding and WM,
p.01. Further funnel plot analyses showed reasonable symmetry
in all reported correlations (the significant Egger’s test for the
correlation between phonological coding and WM may be due to
two outliers in the funnel plot). Taken together, Egger’s test and
funnel plot suggest that there was little influence of publication
bias in the data and thus the original dataset was used in all
reported analyses.
Results
Based on our inclusion criteria, 197 studies involving 260 in-
dependent samples, 29,629 participants, and 2026 correlations
between WM and reading were included for the final analysis.
Overall, the size of the relation between reading and WM was
moderate, r.29 (Cohen, 1988), and significant, 95% CI [.27,
.31]. Next, we examined the relation between reading and WM for
each subcategory of each moderator, and whether domains of WM,
types of reading, and grade level affected the relation between
reading and WM. (Table 2 shows the number of effect sizes for
different moderators).
Moderation Effects of Working Memory Domains
We coded WM tasks as verbal WM (1237 correlations), numer-
ical WM (388 correlations), visuospatial WM (167 correlations),
and composite WM (234 correlations; i.e., WM tasks involving
two domains, or WM scores derived from WM tasks tapping at
least two different domains). As Table 3 shows, the average
correlation between reading (including all reading skills) and WM
for each of the four WM domains was significant: verbal WM, r
.32, 95% CI [.29, .34]; numerical WM, r.26, 95% CI [.23, .29];
visuospatial WM, r.21, 95% CI [.16, .26]; composite WM, r
.26, 95% CI [.22, .31]. As Table 4 shows, after controlling for
publication type, bilingual status, types of reading, and grade level,
verbal WM was more strongly related to reading than were nu-
merical WM and visuospatial WM, ␤⫽.05/.13, t2.45/5.05,
ps.05,
2
.02. Numerical WM was more strongly related to
reading than was visuospatial WM, ␤⫽.08, t2.64, p.01,
2
.02. Taken together, the findings showed that the relation
between reading and WM was affected by domains of WM.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
57
READING AND WORKING MEMORY: A META-ANALYSIS
Moderation Effects of Types of Reading
The relations of types of reading to WM were based on the
following data: phonological coding (193 correlations), decoding
(425 correlations), vocabulary (358 correlations), and comprehen-
sion (953 correlations). As Table 3 shows, the average correlations
between WM and each of the reading skills were significant:
phonological coding, r.34, 95% CI [.29, .37]; decoding, r
.28, 95% CI [.25, .31]; vocabulary, r.26, 95% CI [.24, .29];
comprehension, r.31, 95% CI [.28, .34]. As Table 4 shows,
after controlling for publication type, bilingual status, domains of
WM, and grade level, phonological coding, decoding, vocabulary,
and comprehension were related to WM to a similar degree,
␤⫽⫺.02–.02, t⫽⫺.92–1.01, ps.05,
2
.02. In summary,
the relation between reading and WM did not vary as a function of
types of reading.
Regarding phonological coding, we further investigated whether
WM was differentially related to phoneme manipulations, syllable
manipulations, and rapid naming. As Table 5 shows, after control-
ling for publication type, bilingual status, grade level, and domains
of WM, WM was related to phoneme manipulation, syllable ma-
nipulation, and rapid naming to similar degrees, ␤⫽⫺.09 –.09,
t0 –.90, ps.05,
2
.02.
With respect to decoding, we examined whether different mea-
sures of decoding affected the size of the relation between decod-
ing and WM. We focused on comparisons between word recog-
nition and nonword reading and between sentence/passage reading
and word list reading. As Table 5 shows, after controlling publi-
cation type, bilingual status, domains of WM, and grade level,
word recognition showed a stronger relation with WM than did
nonword reading, ␤⫽.09, t3.07, p.003,
2
.01, and word
list reading showed a stronger relation with WM than did sentence/
passage reading, ␤⫽⫺.08, t⫽⫺2.26, p.03,
2
.01.
Regarding vocabulary, we further investigated whether expres-
sive vocabulary and receptive vocabulary are differentially related
to WM. As Table 5 shows, after controlling publication type,
bilingual status, grade level, and domains of WM, WM was related
to expressive vocabulary and receptive vocabulary to a similar
degree, ␤⫽.003, t.11, p.91,
2
.02.
With respect to comprehension, we examined whether the rela-
tion between WM and comprehension was moderated by the
following factors: listening comprehension versus reading com-
prehension, timed reading comprehension versus untimed reading
comprehension, and expository text comprehension versus narra-
tive text comprehension. Results showed that after controlling
publication type, bilingual status, domains of WM, and grade
level, there were no difference in the size of effects for listening
comprehension versus reading comprehension, ␤⫽.01, t.25,
p.80,
2
.02, for timed reading comprehension versus
untimed reading comprehension, ␤⫽⫺.01, t⫽⫺.45, p.66,
2
.02, or for expository text comprehension versus narrative
text comprehension, ␤⫽⫺.12, t⫽⫺1.22, p.24,
2
.03.
Moderation Effects of Grade Level
We coded two groups for the grade level factor: before 4th grade
and at/beyond 4th grade. As Table 3 shows, the average relation
between reading and WM was r.32, 95% CI [.28, .35] before
4th grade and r.27, 95% CI [.25, .30] at/beyond 4th grade.
Table 4 shows, after controlling for publication type, bilingual
status, domains of WM, and types of reading, the relation between
reading and WM was stronger before 4th grade than it was at/
beyond 4th grade, ␤⫽.06, t2.57; p.01,
2
.02.
Next, we investigated whether this effect of grade level varies
across types of reading. As Table 5 shows, after controlling for
publication type, bilingual status, domains of WM, and types of
phonological coding measures, grade level did not influence the
relation of phonological coding and WM, ␤⫽⫺.02, t⫽⫺.30,
p.77,
2
.01. After controlling for publication type, bilingual
status, domains of WM, and types of decoding measures, grade
level did not influence the relation between decoding and WM,
␤⫽.005, t.15; p.88,
2
.01. However, after controlling
for publication type, bilingual status, domains of WM, and types of
vocabulary measures, vocabulary in the early grades showed a
stronger relation with WM than it did in the later grades, ␤⫽.10,
t2.84; p.01,
2
.02. After controlling for publication type,
bilingual status, domains of WM, and types of comprehension
measures (i.e., listening comprehension vs. reading comprehen-
sion), comprehension in the early grades showed a stronger rela-
tion with WM than it did in the later grades, ␤⫽.06, t2.05; p
.04,
2
.02. To sum up, WM showed a stronger relation with
comprehension and vocabulary before 4th grade than at/beyond
Table 2
The Number of Effects Sizes on the Relation Between WM and Different Types of Reading
Across Moderators
Measure Phonological coding Decoding Vocabulary Comprehension
Domains of WM
Verbal 98 224 222 650
Numerical 48 105 77 143
Visuospatial 20 47 29 57
Composite 27 49 30 103
Grade level
Before 4th grade 103 132 101 208
At/Beyond 4th grade 66 235 245 616
Bilingual status
Native English speakers 158 339 286 875
Bilingual 35 86 72 78
Publication type
Peer-reviewed 132 305 222 582
Non–peer-reviewed 61 120 136 371
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
58 PENG ET AL.
4th grade, but the relations of WM and phonological coding and
decoding did not differ across grades.
Next, we investigated whether grade level interacted with types
of comprehension tasks when considering the relation between
WM and comprehension. After controlling for publication type,
bilingual status, and domains of WM, reading comprehension
showed a stronger relation with WM than did listening compre-
hension before 4th grade, ␤⫽.09, t2.23; p.03,
2
.02. In
contrast, there was no difference in the relations of WM to reading
comprehension and listening comprehension at/beyond 4th grade,
␤⫽.04, t1.39; p.17,
2
.02. After controlling for
publication type, bilingual status, and domains of WM, timed
reading comprehension was more strongly related to WM than was
untimed reading comprehension before 4th grade, ␤⫽.17, t
2.24; p.03,
2
.02. In contrast, there was no difference in the
relation of WM for timed and untimed reading comprehension
at/beyond 4th grade, ␤⫽.01, t.41; p.68,
2
.02. After
controlling for publication type and domains of WM, narrative text
comprehension and expository text comprehension showed com-
parable relations to WM at/beyond 4th grade, ␤⫽.12, t1.32;
p.21,
2
.03 (there is no variance on bilingual status
at/beyond 4th grade for this analysis, thus bilingual status was not
controlled in this model). Because we only obtained 8 data points
(no data points for expository text comprehension), we did not
analyze the comparison between narrative text comprehension and
expository text comprehension before 4th grade.
We also examined whether the effect of domains of WM on
the relation between reading and WM is affected by grade level.
As shown in Table 6, before 4th grade, after controlling for
publication type, bilingual status, and types of reading, reading
was related to different domains of WM to a similar degree,
␤⫽.02 .08, t.47–1.83; ps.05,
2
.02, except that
composite WM showed a stronger relation with reading than did
numerical WM and visuospatial WM, ␤⫽.13/.15, t2.90/
3.30; ps.01,
2
.02. At/beyond 4th grade, after controlling
for publication type, bilingual status, and types of reading,
verbal WM showed a stronger relation with reading than did
visuospatial WM and composite WM, ␤⫽.14/.11, t4.40/
Table 3
Relation Between Reading and WM
Measure
Number of
correlations Correlation Correlation 95% CI
Between-study sampling
variance (
2
)
Main average correlation 2026 .29 [.27, .31] .02
Publication type
1. Peer-reviewed 1318 .31 [.29, .34] .02
2. Non–peer-reviewed 708 .24 [.21, .27] .02
Bilingual status
1. Native English speaker 1749 .29 [.27, .31] .02
2. Bilingual 277 .30 [.24, .36] .03
Grade level
1. Before 4th grade 566 .32 [.28, .35] .02
2. At/Beyond 4th grade 1231 .27 [.25, .30] .02
Domains of WM
1. Verbal WM 1237 .32 [.29, .34] .02
2. Numerical WM 388 .26 [.23, .29] .02
3. Visuospatial WM 167 .21 [.16, .26] .04
4. Composite WM 234 .26 [.22, .31] .03
Types of reading
1. Phonological coding 193 .34 [.29, .37] .02
a. Phoneme manipulation 63 .33 [.27, .38] .01
b. Syllable manipulation 27 .35 [.22, .47] .04
c. Rapid naming 86 .27 [.24, .32] .02
2. Decoding 425 .28 [.25, .31] .02
a. Word recognition 324 .29 [.26, .32] .02
b. Non-word reading 84 .25 [.21, .30] .01
c. Word list reading accuracy 306 .29 [.26, .32] .01
d. Sentence/passage reading accuracy 119 .24 [.19, .28] .02
3. Vocabulary 358 .26 [.24, .29] .02
a. Expressive vocabulary 135 .27 [.23, .32] .03
b. Receptive vocabulary 220 .26 [.23, .29] .02
4. Comprehension 953 .31 [.28, .34] .02
a. Reading comprehension 733 .31 [.28, .34] .02
b. Listening comprehension 220 .31 [.27, .35] .02
c. Narrative text 126 .37 [.24, .50] .05
d. Expository text 28 .22 [.12, .31] .01
e. Timed reading comprehension 183 .29 [.25, .34] .02
f. Untimed reading comprehension 550 .31 [.27, .34] .02
Note. Verbal WM WM task that involves simultaneous verbal information storage and processing; Numerical WM WM task that involves
simultaneous numerical information storage and processing; Visuospatial WM WM task that involves simultaneous visual or spatial information storage
and processing; Composite WM WM task that involves simultaneous storage and processing of information across at least two domains of verbal,
numerical, and visuospatial domains, or composite scores derived from WM tasks tapping at least two domains of verbal, numerical, and visuospatial
domains; CI confidence interval. Between-study sampling variance (
2
) for this model is .02.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
59
READING AND WORKING MEMORY: A META-ANALYSIS
2.61; ps.05,
2
.02, and numerical WM showed a stronger
relation with reading than did visuospatial WM, ␤⫽.10, t2.
63; p.01,
2
.02. To sum up, verbal WM showed stronger
relations with reading than did visuospatial WM, but the con-
trast between these two WM domains is especially distinct
at/beyond 4th grade.
Unique Contributions of Working Memory to
Reading Comprehension
We examined whether WM made unique contributions to read-
ing comprehension. There were 15 studies that provided correla-
tion tables to calculate 134 correlations on reading comprehension
and WM, partialing out only decoding; 9 studies that provided
correlation tables to calculate 126 correlations on reading compre-
hension and WM, partialing out only vocabulary; and 5 studies that
provided correlation tables to calculate 128 correlations on reading
comprehension and WM, partialing out both decoding and vocab-
ulary.
Before we synthesized the partial correlations, we examined
whether this small pool of studies are representative of the
overall studies included in the present review. Results showed
that based on these studies, the correlation between WM and
phonological coding, decoding, vocabulary, and reading com-
prehension is, .27 (95% CI [.22, .33]), .28 (95% CI [.23, .35]),
.24 (95% CI [.18, .29]), and .34 (95% CI [.27, .37]), respec-
tively, which is similar to findings based on all studies included
in the present review.
Next, we synthesized the partial correlations of interest.
When only decoding was partialed out, there was a significant
relation between WM and reading comprehension, r.17, 95%
CI [.10, .24]. When only vocabulary was partialed out, there
was a significant relation between WM and reading compre-
hension, r.23, 95% CI [.16, .29]. When both decoding and
vocabulary were partialed out, WM was no longer related to
reading comprehension at a significant level, r.20, 95% CI
[.01, .40]. Taken together, WM made unique contributions to
reading comprehension after controlling for either decoding or
vocabulary, but not when both decoding and vocabulary were
controlled for.
Discussion
This meta-analysis investigated the relation between reading
and WM, and whether domains of WM, types of reading, and
grade level influence this relation. We found that the relation
between reading and WM was moderate (r.29) and was
significantly influenced by domains of WM and grade level. Spe-
cifically, before 4th grade, different domains of WM were related
to reading to a similar degree. In contrast, verbal WM showed a
stronger relation with reading than did visuospatial WM at/beyond
4th grade. WM showed a stronger relation to reading, especially
for comprehension and vocabulary, before 4th grade than that
at/beyond 4th grade. In general, the size of the WM-reading
relation was not influenced by types of reading. Two exceptions to
this general finding emerged for different types of decoding mea-
sures, where WM and word recognition were more strongly related
than were WM and nonword reading, and where WM was more
strongly related to word list reading accuracy than it was to
sentence/passage reading accuracy. We also found that WM made
unique contributions to reading comprehension when either decod-
ing or vocabulary were controlled for, but not when both were
controlled for.
Table 4
Meta-Regression of the Moderation Analysis on the Relation Between Reading and WM
Measure Beta SE t 95% CI pvalue
Grade level: Before 4th grade vs. at/beyond 4th grade .06 .02 2.57 [.01, .10] .01
Publication type: Peer-reviewed vs. Non–peer-reviewed .06 .02 2.86 [.02, .11] .005
Bilingual status: Native English speaker vs. Bilingual .01 .04 .39 [.08, .06] .70
Domains of WM
Verbal vs. Numerical .05 .02 2.45 [.01, .10] .02
Verbal vs. Visuospatial .13 .03 5.05 [.08, .18] .001
Verbal vs. Composite .06 .03 1.68 [.01, .12] .10
Numerical vs. Visuospatial .08 .03 2.64 [.02, .14] .01
Numerical vs. Composite .003 .04 .09 [.07, .07] .93
Visuospatial vs. Composite .07 .04 2.13 [.14, .01] .04
Types of reading
Decoding vs. Phonological coding .01 .03 .29 [.05, .06] .78
Vocabulary vs. Phonological coding .02 .03 .68 [.08, .04] .50
Comprehension vs. Phonological coding .001 .02 .05 [.05, .04] .96
Vocabulary vs. Decoding .02 .03 .92 [.08, .03] .36
Comprehension vs. Decoding .01 .02 .27 [.04, .03] .79
Comprehension vs. Vocabulary .02 .02 1.01 [.02, .06] .43
Note. All moderators were entered in one model. Several models were run for thorough subgroup comparisons among moderators with more than 2
categories. For the convenience of presentation, subgroup comparisons within Domains of WM and Types of Reading are all listed in the table. Verbal
WM WM task that involves simultaneous verbal information storage and processing; Numerical WM WM task that involves simultaneous numerical
information storage and processing; Visuospatial WM WM task that involves simultaneous visual or spatial information storage and processing;
Composite WM WM task that involves simultaneous storage and processing of information across at least two domains of verbal, numerical, and
visuospatial domains, or composite scores derived from WM tasks tapping at least two domains of verbal, numerical, and visuospatial domains; CI
confidence interval. Between-study sampling variance (
2
) for this model is .02. The italicized variable shows a stronger correlation with WM in a set of
dummy variables.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
60 PENG ET AL.
Table 5
Meta-Regression of the Moderation Analysis on the Relation Between WM and Different Reading Skills
Measure
Phonological coding
Measure
Decoding
Beta SE t 95% CI pvalue Beta SE t 95% CI pvalue
Publication type Publication Type
Peer-reviewed vs. non–peer-reviewed .07 .06 1.27 [.05, .20] .22 Peer-reviewed vs. non–peer-reviewed .09 .03 2.97 [.03, .15] .004
Grade level Grade level
Before 4th grade vs at/beyond 4th grade .02 .06 .30 [.15, .11] .77 Before 4th grade vs at/beyond 4th grade .005 .03 .15 [.07, .06] .88
Bilingual status Bilingual status
Native English speaker vs. Bilingual .04 .09 .45 [.23, .15] .66 Native English speaker vs. Bilingual .04 .04 .99 [.12, .04] .32
Domains of WM Domains of WM
Verbal vs. Numerical .11 .07 1.72 [.02, .26] .10 Verbal vs. Numerical .02 .03 .83 [.07, .03] .41
Verbal vs. Visuospatial .06 .09 .60 [.14, .25] .55 Verbal vs. Visuospatial .08 .04 1.88 [.01, .17] .06
Verbal vs. Composite .02 .11 .21 [.23, .19] .83 Verbal vs. Composite .04 .05 .95 [.13, .05] .34
Numerical vs. Visuospatial .06 .11 .57 [.28, .16] .57 Numerical vs. Visuospatial .10 .04 2.67 [.02, .18] .01
Numerical vs. Composite .14 .10 1.46 [.34, .06] .16 Numerical vs. Composite .02 .04 .51 [.11, .06] .61
Visuospatial vs. Composite .08 .07 1.18 [.21, .06] .25 Visuospatial vs. Composite .13 .05 2.72 [.22, .03] .01
Measures of phonological coding Measures of decoding
Phoneme vs. Syllable .09 .10 .90 [.29, .11] .38 Real words vs. Non-words .09 .03 3.07 [.03, .14] .003
Phoneme vs. Rapid naming 0 .05 0 [.11, .11] .99 Sentence/passage vs. word list .08 .04 2.26 [.16, .01] .03
Syllable vs. Rapid naming .09 .11 .82 [.13, .31] .42
Measure
Vocabulary
Measure
Comprehension
Beta SE t 95% CI p-value Beta SE t 95% CI p-value
Publication type Publication type
Peer-reviewed
vs. non–peer-
reviewed
.01 .03 .18 [.07, .06] .86 Peer-reviewed vs. non–
peer-reviewed
.09/.14/.09 .03/.08/.03 2.78/1.78/2.75 [.02, .15]/[.03, .32]/
[.03, .15]
.01/.10/.01
Grade level Grade level
Before 4th grade
vs at/beyond
4th grade
.10 .03 2.84 [.03, .17] .01 Before 4th grade vs.
at/beyond 4th grade
.06/.01/.12 .03/.09/.04 2.05/.16/2.89 [.002, .12]/[.17, .20]/
[.04, .20]
.04/.87/.004
Bilingual status Bilingual status
a
Native English
speaker vs.
Bilingual
.08 .06 1.41 [.03, .20] .16 Native English speaker
vs. Bilingual
.11//.07 .05//.05 2.39//1.39 [.20, .02]//
[.18, .03]
.02//.17
Domains of WM Domains of WM
Verbal vs.
Numerical
.02 .03 .64 [.09, .05] .52 Verbal vs. Numerical
b
.10/.12/.09 .03/.06/.03 3.38/1.89/3.19 [.04, .17]/[.02, .06]/
[.04, .15]
.001/.08/.002
Verbal vs.
Visuospatial
.09 .05 1.86 [.01, .18] .07 Verbal vs.
Visuospatial
b
.16/.03/.13 .03/.10/.04 4.94/.27/3.79 [.09, .22]/[.19, .24]/
[.06, .21]
.001/.79/.001
Verbal vs.
Composite
.004 .07 .06 [.14, .14] .95 Verbal vs. Composite
c
.08/.02/.07 .04/.08/.04 2.04/.22/1.74 [.002, .15]/[.20, .16]/
[.01, .16]
.04/.83/.08
Numerical vs.
Visuospatial
.07 .05 1.30 [.17, .03] .20 Numerical vs.
Visuospatial
.05/.09/.04 .04/.09/.04 1.29/.04/1.03 [.03, .13]/[.29, .10]/
[.04, .12]
.20/.32/.30
Numerical vs.
Composite
.03 .07 .35 [.17, .12] .73 Numerical vs.
Composite
.03/.14/.02 .04/.05/.05 .63/2.63/.42 [.11, .06]/[.25, .03]/
[.11, .07]
.53/.02/.68
Visuospatial vs.
Composite
.09 .08 1.21 [.24, .06] .23 Visuospatial vs.
Composite
.08/.05/.06 .04/.08/.05 1.87/.55/.26 [.16, .01]/[.22, .12]/
[.16, .03]
.06/.59/.21
(table continues)
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
61
READING AND WORKING MEMORY: A META-ANALYSIS
Table 5 (continued)
Measure
Vocabulary
Measure
Comprehension
Beta SE t 95% CI p-value Beta SE t 95% CI p-value
Measures of
vocabulary
Measures of
comprehension
d
Expressive vs.
Receptive
.003 .03 .11 [.05, .06] .91 Listening vs. Reading/
Expository vs.
Narrative/Timed
Reading
Comprehension vs.
Untimed Reading
Comprehension
.01/.12/.01 .03/.10/.06 .25/1.22/.45 [.05, .06]/[.33, .09]/
[.07, .05]
.80/.24/.66
Note. All moderators were entered in one model. Several models were run for thorough subgroup comparisons among moderators with more than 2 categories. For the convenience of presentation,
subgroup comparisons within Domains of WM are all listed in the table. Verbal WM WM task that involves simultaneous verbal information storage and processing; Numerical WM WM task
that involves simultaneous numerical information storage and processing; Visuospatial WM WM task that involves simultaneous visual or spatial information storage and processing; Composite
WM WM task that involves simultaneous storage and processing of information across at least two domains of verbal, numerical, and Visuospatial domains, or composite scores derived from WM
tasks tapping at least two domains of verbal, numerical, and Visuospatial domains. CI confidence interval. The first group in each group comparison variable is the reference group (e.g., in Verbal
vs. Numerical, Numerical is the reference group in the dummy coding of Domains of WM). Between-study sampling variance (
2
)is.01.03 across models. The italicized variable shows a stronger
correlation with WM in a set of dummy variables.
a
Because there is no variance on ELL, it is not included in the moderation analysis for Expository vs. Narrative.
b
Verbal WM showed stronger relation with comprehension than numerical WM in
models that contains Listening vs. Reading or Timed Reading Comprehension vs. Untimed Reading Comprehension.
c
Verbal WM showed stronger relation with comprehension than numerical WM
in the models that contains Listening vs. Reading.
d
For the model of comprehension, because of collinearity between Listening vs. Reading and Expository vs. Narrative for the moderation analysis
of comprehension, these moderators were run in separate models. Timed Reading Comprehension vs. Untimed Reading Comprehension was separately analyzed as the moderator for reading
comprehension.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
62 PENG ET AL.
Domains of Working Memory and Grade Level
Whether different domains of WM are differentially related to
reading is relevant to the debate on the domain-specificity of WM
in higher cognitive functions. Based on a domain-general view of
WM (e.g., Baddeley, 1986;Engle, 2002), WM is a domain-general
construct, and the relation between WM and reading should not be
influenced by the nature of the materials used to measure WM. In
contrast, domain-specific WM models (e.g., Ericsson & Kintsch,
1995), propose that the relation of WM to reading should show
domain-specificity because of its role in the retrieval and integra-
tion of verbal information from the text and long-term declarative
memory.
Although we found significant relations between reading and
WM regardless of domains (verbal, numerical, visuospatial),
the size of this relation was strongest for verbal WM and
weakest for visuospatial WM (with relations for numerical WM
falling in between). These findings suggest that WM has both
domain-general and domain-specific relations with reading.
That is, the domain-general central executive component in
WM, based on Baddeley’s domain-general WM model, may
play a role in reading performance, but the retrieval of verbal
knowledge in long-term memory and its integration with
language-based information during reading specifically re-
quires verbal WM.
That the role of WM in reading is both domain-general and
domain-specific may also have implications for understanding the
relation between reading and WM from a developmental perspec-
tive. That is, as children are rapidly acquiring verbal knowledge/
skills (e.g., decoding and vocabulary) when they are learning to
read (i.e., before 4th grade), domain-general aspects of WM may
Table 6
Meta-Regression of the Moderation Analysis on the Relation Between Reading and WM Before 4th Grade and at/Beyond 4th Grade
Measure Beta SE t 95% CI pvalue
Before 4th grade
Publication type
Peer-reviewed vs. Non–peer-reviewed .02 .04 .39 [.07, .11] .70
Bilingual status
Native English speaker vs. Bilingual .03 .05 .71 [.06, .13] .48
Domains of WM
Verbal vs. Numerical .06 .04 1.49 [.02, .14] .14
Verbal vs. Visuospatial .08 .05 1.68 [.01, .17] .10
Verbal vs. Composite .08 .05 1.70 [.17, .02] .10
Numerical vs. Visuospatial .02 .05 .43 [.07, .11] .67
Numerical vs. Composite .14 .05 2.89 [.24, .04] .005
Visuospatial vs. Composite .16 .05 3.27 [.25, .06] .002
Types of reading
Decoding vs. Phonological coding .02 .04 .44 [.09, .06] .66
Vocabulary vs. Phonological coding .01 .05 .31 [.10, .08] .76
Comprehension vs. Phonological coding .01 .04 .35 [.06, .09] .73
Vocabulary vs. Decoding .01 .04 .22 [.09, .07] .82
Comprehension vs. Decoding .02 .03 .57 [.05, .08] .57
Comprehension vs. Vocabulary .02 .03 .48 [.05, .08] .64
At\Beyond 4th grade
Publication type
Peer-reviewed vs. Non–peer-reviewed .08 .03 3.07 [.03, .13] .003
Bilingual status
Native English speaker vs. Bilingual .04 .05 .95 [.14, .05] .35
Domains of WM
Verbal vs. Numerical .05 .03 1.84 [.004, .10] .07
Verbal vs. Visuospatial .15 .03 4.76 [.09, .21] .001
Verbal vs. Composite .09 .04 2.21 [.01, .17] .03
Numerical vs. Visuospatial .10 .04 2.64 [.03, .17] .01
Numerical vs. Composite .04 .04 .94 [.05, .13] .35
Visuospatial vs. Composite .06 .04 1.40 [.14, .02] .16
Types of reading
Decoding vs. Phonological coding .01 .04 .33 [.06, .08] .75
Vocabulary vs. Phonological coding .04 .04 1.00 [.12, .04] .32
Comprehension vs. Phonological coding .02 .03 .52 [.07, .04] .60
Vocabulary vs. Decoding .04 .03 1.16 [.11, .03] .25
Comprehension vs. Decoding .02 .02 .64 [.06, .03] .53
Comprehension vs. Vocabulary .02 .02 .84 [.02, .07] .40
Note. All moderators were entered in one model. Several models were run for thorough subgroup comparisons among moderators with more than 2
categories. For the convenience of presentation, subgroup comparisons within Domains of WM and Types of Reading are all listed in the table. Verbal
WM WM task that involves simultaneous verbal information storage and processing; Numerical WM WM task that involves simultaneous numerical
information storage and processing; Visuospatial WM WM task that involves simultaneous visual or spatial information storage and processing;
Composite WM WM task that involves simultaneous storage and processing of information across at least two domains of verbal, numerical, and
visuospatial domains, or composite scores derived from WM tasks tapping at least two domains of verbal, numerical, and visuospatial domains; CI
confidence interval. Between-study sampling variance (
2
) for this model is .02. The italicized variable shows a stronger correlation with WM in a set of
dummy variables.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
63
READING AND WORKING MEMORY: A META-ANALYSIS
be influential and predictive of reading. In contrast, at/beyond 4th
grade when typically developing readers have built up relatively
solid foundational lexical representations and verbal knowledge,
more domain-specific aspects of WM would be expected to play a
greater role in reading performance. Indeed, our findings are
consistent with this bidirectional hypothesis.
This bidirectional hypothesis is also consistent with data from
studies that suggest learning to read may help shape verbal mem-
ory. For example, Melby-Lervåg and Hulme (2010) found that
training phonological awareness or vocabulary among 7-year-olds
not only improved phonological awareness and vocabulary, but
also had significant transfer effects to verbal STM. In another
training study, Park, Ritter, Lombardino, Wiseheart, and Sherman
(2014) found that explicit phonemic awareness training improved
word reading as well as verbal STM and verbal WM. In a recent
review, Demoulin and Kolinsky (2016) suggest that learning to
read in English improves phonemic representations that could
influence both spoken word recognition and verbal memory per-
formance. Learning to read might help connect the phonological,
orthographic, and semantic representations of words. To the extent
that this multiple mapping improves the quality of lexical repre-
sentations in long-term memory, the encoding of to-be remem-
bered items should be strengthened in verbal WM and verbal
knowledge retrieval in verbal WM should become more efficient.
All of these findings together with those from the present meta-
analysis suggest that the effects of domains of WM on reading
should be interpreted from a developmental perspective.
Types of Reading and Grade Level
The intrinsic cognitive load theory suggests that relatively com-
plex reading tasks consume more WM resources than reading tasks
with relatively simpler structures (Chandler & Sweller, 1991).
According to this view, complex reading skills such as compre-
hension are hypothesized to show stronger correlations with WM
than foundational reading skills such as phonological coding,
decoding, and vocabulary. In contrast, the dual process theory
suggests that the involvement of WM in a reading task is largely
determined by the efficiency with which the task can be performed
(Evans & Stanovich, 2013). In this view, word reading and access
to word meaning is expected to be more effortful during early
learning, but may become more automatized with experience and,
therefore, come to rely less on WM. Other reading tasks, particu-
larly those involving new materials, reasoning, and complex inte-
gration of information ought to require WM and have less potential
for more automatic forms of processing.
The overall pattern of findings did not completely support
predictions based on the intrinsic cognitive load theory. Specifi-
cally, we found that phonological coding, decoding, vocabulary,
and comprehension were related to WM to a similar degree. Even
within each reading skill, we did not find that seeming complexity
of reading skills affected the relation of those skills with WM. For
example, within phonological coding, even though phoneme ma-
nipulation seems to have more concurrent storage and processing
features than rapid naming, these two phonological coding skills
were not differentially related to WM. Regarding vocabulary,
although expressive vocabulary is more difficult than receptive
vocabulary, they were related to WM to a similar degree. With
respect to the relation between comprehension and WM, we found
no difference in the relations of WM to expository versus narrative
text comprehension.
The findings were generally consistent with some of the predic-
tions derived from the dual process theory. For example, WM was
more strongly related to word list reading accuracy than it was to
sentence/passage reading accuracy. This may be because in sen-
tence/passage reading tasks, the syntactic and semantic context
supports word recognition in a way that is not present in reading
lists of unrelated words. We also found that WM was more
strongly related to word reading than to nonword reading, which is
in line with the idea that GPC rules can be applied to decoding
nonwords, but cannot be routinely applied to the decoding of real
words especially when those lists include low frequency words and
words with irregular spelling-to-sound correspondence or both
(e.g., pint). That said, because of limited information from the
primary studies, we unable to pinpoint the word frequency and
regularity in the real word reading measurement. Future studies
should further investigate whether the frequency and regularity are
the key factors that determine the differences of WM involvement
in real word reading and nonword reading.
The findings that grade level influenced the relation between
reading and WM further support some of the predictions derived
from the dual process theory. Specifically, the relation between
reading and WM was stronger before 4th grade than it was at/
beyond 4th grade. Moreover, WM showed stronger relations with
comprehension and vocabulary before 4th grade than at/beyond
4th grade. These findings suggest that in the early grades, children
whose foundational reading skills are not yet efficient, may draw
more on WM during reading performance, particularly for com-
prehension. In contrast, at/beyond 4th grade, individuals are more
efficient decoders and they have acquired stronger verbal knowl-
edge (e.g., vocabulary knowledge). During reading they can effi-
ciently retrieve words and their meanings, requiring less involve-
ment of WM.
Because of the reading development sequence, we predicted that
there might be an interaction between grade level and different
aspects of comprehension. Specifically, we suggested that WM
may be more strongly related to reading comprehension than to
listening comprehension before 4th grade because WM may be
needed for both decoding (which is not yet efficient) and compre-
hension during reading, whereas listening comprehension requires
comprehension, but not decoding. At 4th grade and beyond, we
suggested that reading comprehension and listening comprehen-
sion might show comparable relations with WM because experi-
enced readers would require WM primarily for the comprehension
component in both reading and listening. The findings were con-
sistent with this hypothesis, which is in line with the dual process
theory.
Based on the dual process theory, we predicted that WM might
be more strongly related to timed reading comprehension versus
untimed reading comprehension. Similar to the findings above for
reading comprehension versus listening comprehension, we found
that WM was more strongly related to timed versus untimed
reading comprehension, but only for children before 4th grade.
This may be because before 4th grade, children are still acquiring
foundational decoding and vocabulary skills and thus timed read-
ing comprehension places greater cognitive load on the reading
system for these younger children than does untimed reading
comprehension.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
64 PENG ET AL.
We also hypothesized that before 4th grade, narrative text com-
prehension and expository text comprehension would show com-
parable relations with WM, because individuals may find both
types of texts challenging due to limited foundational reading
skills (i.e., decoding and vocabulary). We predicted that WM
would be more strongly related to expository text comprehension
versus narrative comprehension at/beyond 4th grade. This predic-
tion is based on the idea that although older readers have largely
mastered decoding and have acquired vocabulary knowledge that
facilitates narrative comprehension, expository texts contain unfa-
miliar content-specific words (e.g., academic vocabulary) and have
less familiar text structures. Tasks that are novel or less familiar
are thought to require more effortful processing, which requires
more WM. Although we did not have sufficient data points (8 data
points) before 4th grade to test all of these predictions derived
from the dual process theory, we found that narrative text com-
prehension and expository text showed comparable relations with
WM at/beyond 4th grade, which is inconsistent with the predic-
tion. This finding may make sense if one considers studies that
have investigated the features of narrative and expository texts at
different grade levels. Narrative texts in the later grades often
increase difficulty by increasing passage length and sentence com-
plexity. Moreover, text complexity is also determined by cohesion
(e.g., explicit referential overlap and causal relationships, operat-
ing between sentences, groups of sentences, paragraphs, and chap-
ters; Graesser, McNamara, & Louwerse, 2003;Kintsch, 1998) and
compared with expository texts, narrative texts often have low
cohesion (McNamara et al., 2012). Therefore, it may be that
specific, but different features of narrative and expository texts at
the higher grade levels require that readers to draw on WM to a
similar extent across the two types of text.
Working Memory’s Unique Contributions to
Reading Comprehension
Another contribution of the present review is that we investi-
gated WM’s unique contributions to reading comprehension. Re-
sults showed that when only decoding or only vocabulary was
controlled for, there was a significant, though small relation be-
tween WM and reading comprehension. However, when both
decoding and vocabulary were controlled for, the relation between
WM and reading comprehension failed to reach significance.
These findings suggest the relation of WM to reading comprehen-
sion involves both decoding and vocabulary, which is in line with
the Simple View of Reading (SVR) that claims that the product of
decoding and oral language (often indexed by vocabulary) ac-
counts for reading comprehension (e.g., Adlof et al., 2006). These
findings, however, are not consistent with models of reading
comprehension that suggest that other comprehension-related lin-
guistic processes (e.g., strategies, comprehension monitoring and
inference-making), in addition to decoding and vocabulary, require
WM (Perfetti, 2007).
Thus, our findings may suggest either other linguistic processes
in reading comprehension do not involve WM (which is unlikely),
or WM influences other linguistic processes via decoding and
vocabulary. Research using the Direct and Inferential Mediation
(DIME) Model has shown that vocabulary and world knowledge
have large direct effects on reading comprehension, but also indi-
rect effects on reading comprehension, via inferencing and, in
some studies, via comprehension strategies (e.g., Ahmed et al.,
2016;Cromley & Azevedo, 2007;Cromley, Snyder-Hogan, &
Luciw-Dubas, 2010). Moreover, Fuchs et al. (2015) investigated
whether WM combined with reading skills (decoding and lan-
guage comprehension) training produce synergistic effects on
reading comprehension compared to reading skills training only
among young children. Their results showed that compared with
controls, the hybrid training group significantly improved in both
WM and reading comprehension, and the skill-based training
group significantly improved in reading comprehension, but not
WM. However, there was no difference between the hybrid train-
ing group and the skill training only group on reading comprehen-
sion. All these findings, together with ours, may suggest that WM
is important for reading comprehension, but WM may contribute
to reading comprehension indirectly through decoding and lan-
guage skills (e.g., vocabulary, inferencing, and comprehension
strategies). Future studies may further test this hypothesis through
the use of longitudinal designs that investigate developmental
associations between WM, decoding and reading comprehension
(e.g., Kim, 2016) starting in the preschool (i.e., prereading) years.
Limitations and Implications for Future Studies
The conclusions from this meta-analysis were derived from the
combined results of 197 studies conducted among more than
29,000 individuals. Despite the scale of our literature search and
the final sample size for our study, we note the following limita-
tions. First, we focused predominantly on samples that included
typically developing participants. In our literature review, we
found and coded 67 studies that included individuals with disabil-
ities. However, the disabilities in these studies are rather hetero-
geneous including learning disabilities, reading disabilities, intel-
lectual disabilities, Autism, Schizophrenia, Williams Syndrome,
hearing impairment, Parkinson’s disease, and other disorders. Be-
cause different types of disabilities may affect WM to a different
degree (Alloway, Gathercole, Kirkwood, & Elliott, 2009;Koshino
et al., 2005;Vicari, Carlesimo, & Caltagirone, 1995), we decided
not to analyze the data by disability group.
In particular, we made the decision not to include individuals
with reading disabilities in the analysis because we were con-
cerned that their inclusion might lead to less clarity in determining
the relation between reading and WM even disaggregating data for
this subgroup. Specifically, reading disability groups in primary
studies can be quite heterogeneous, sometimes consisting of indi-
viduals with comorbid reading and mathematics disabilities
(Cirino, Fuchs, Elias, Powell, & Schumacher, 2015;Peng &
Fuchs, 2016), individuals with dyslexia (Hulme & Snowling,
2009;Snowling, 2000), and individuals with specific comprehen-
sion deficits (e.g., Cain & Oakhill, 2006;Lervåg & Aukrust,
2010). Different subgroups of individuals with reading disabilities
may have different cognitive or skill deficit profiles that could
influence the relation of WM and reading. It is suggested that
children with comorbid reading and mathematics disabilities tend
to show more comprehensive and severe cognitive and reading
deficits (e.g., Cirino et al., 2015;Peng & Fuchs, 2016), dyslexia is
closely related to phonological coding and decoding deficits (e.g.,
Snowling, 2000), and specific comprehension deficits are related
to difficulties in vocabulary and oral language as well as in
executive functions (e.g., Cain & Oakhill, 2006;Cutting et al.,
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
65
READING AND WORKING MEMORY: A META-ANALYSIS
2013;De Beni & Palladino, 2000). In the studies we reviewed that
included reading disabilities, few provided detailed information
about whether their samples had comorbid reading and mathemat-
ics disabilities, dyslexia, or specific comprehension deficits. Thus,
we determined that the inclusion of individuals with reading dis-
abilities in this meta-analysis would not facilitate fine-grained
conclusions on the relation between reading and WM. Future
studies are needed to specifically investigate the relation between
reading and WM among specific reading disability groups.
A second limitation is that we focused on reading in English.
Whether the findings can be generalized to reading in other lan-
guages warrants further investigations. For example, different al-
phabetic languages have varying orthographic depth, which may
affect the relation between reading and WM. According to the
orthographic depth hypothesis (Frost & Katz, 1992), in languages
with shallow orthographies where every word can be decoded with
GPC rules, beginning readers can just apply GPC rules to reading
any word. In contrast, in deep orthographies where there are
regular and irregular words, beginning readers need to apply GPC
rules as well as other skills such as analogies and morphological
knowledge to reading words. Therefore, for languages with rela-
tively deep orthographies such as English, beginning readers may
have more difficulty decoding words, thus requiring more WM in
early reading. In contrast, for languages with relatively shallow
orthographies such as Finnish, beginning readers may have fewer
problems decoding words, thus relying on WM less in early
reading performance.
Written systems in different languages may also affect the
relation between reading and WM. Compared to reading in alpha-
betic languages, reading in logographic languages may require
WM resources differently. For example, Chinese is one logo-
graphic language with characters as the basic reading unit. Com-
pared with words in English, characters are more visually com-
plex. Moreover, it is common to see characters with different
meanings that sound the same and look alike (Peng, Tao, & Li,
2013). Thus, reading Chinese characters relies heavily on seman-
tics and individuals must memorize a large number of characters to
build a strong character-semantic route for fluent reading (Shu,
McBride-Chang, Wu, & Liu, 2006). These characteristics of the
Chinese written system may make reading in Chinese more WM
taxing and draw on different domains of WM (e.g., visuospatial
WM) at different reading stages (e.g., Peng et al., 2013;Tan et al.,
2001). To sum up, future research is needed to systematically
investigate whether orthographic depth within alphabetic lan-
guages and different language written systems affect the relation
between reading and WM.
A third limitation is that we were unable to systematically
study how time constraints in the measurement of WM might
affect the relation between WM and reading. This is because the
majority of studies did not report time constraints on their WM
tasks and there was large variation in the time constraints on
WM assessment across studies when such timing was reported.
That being said, we suggest two ways to help address the
time-delimited nature of WM tasks in future studies. First, we
found that preload WM span tasks (e.g., backward digit recall)
have relatively consistent time constraints across studies, are
used across a wide age range (Pickering & Gathercole, 2001),
and require less task difficulty adjustment for different popu-
lations compared to complex span tasks (e.g., Adams, Bourke,
& Willis, 1999). Thus, for studies that want to strictly control
for timing aspects of WM assessment, especially among young
children, preload WM span tasks may be a good candidate.
Second, complex span tasks usually use the span (storage) as
the indicator of WM capacity, which could be inflated by WM
tasks with fewer time constraints as individuals will allot more
time/attention resources to rehearsing the to-be-remembered
items. The combination of processing efficiency and storage
may better reflect the WM capacity in complex span tasks (e.g.,
Bayliss, Jarrold, Gunn, & Baddeley, 2003;Winke, 2005).
Another limitation of the present review is that we were unable
to address the effect of strategy use during WM tasks on the
relation between WM and reading. There are substantial individual
differences in strategy use during WM tasks (Morrison, Rosen-
baum, Fair, & Chien, 2016;Peng & Fuchs, 2017), and strategy use
is often considered a primary driver of the predictive utility of WM
performance (e.g., McNamara & Scott, 2001;Turley-Ames &
Whitfield, 2003). The dominant view from previous research is
that strategy use in WM tasks introduces “noise” that weakens the
relation between WM and performance on other tasks, and so
restricting strategy use in WM tasks would strengthen the relation
between WM and performance on other tasks (e.g., Bailey, Dun-
losky, & Kane, 2008;Dunlosky & Kane, 2007;Friedman &
Miyake, 2004;St Clair-Thompson, 2007;Turley-Ames & Whit-
field, 2003).
However, recent research on specific strategy use in WM tasks
suggests that strategy use efficiency during WM tasks may actu-
ally reflect WM capacity and this is what links WM to other
cognitive tasks (e.g., Olesen, Westerberg, & Klingberg, 2004;
Robison & Unsworth, 2017;Unsworth & Spillers, 2010). More
specifically, two hypotheses have been proposed to explain how a
specific strategy can link WM and other cognitive tasks. The
strategy mediation hypothesis claims that the relation between
WM and cognitive tasks is “fully” mediated by effective strategies
used in the cognitive tasks (Gonthier & Thomassin, 2015). That is,
strategy use in a cognitive task reflects WM capacity and using a
specific effective strategy in a cognitive task mediates the relation
between WM and performance on that task. In contrast, the strat-
egy affordance hypothesis claims that the relation between WM
and performance of other tasks is affected by the similarity of
strategy used in both WM tasks and other tasks (Bailey et al.,
2008). That is, if the strategies used to improve WM performance
also can be used to improve performance on other tasks, then using
these strategies in WM tasks is likely to improve the relation
between WM and performance on other tasks (Peng & Fuchs,
2017). That said, only a few studies directly tested the two hy-
potheses. More studies are needed to investigate different strate-
gies use in WM and reading tasks as to better understand which
strategies influence the relation between reading and WM (Mc-
Cabe, Redick, & Engle, 2016).
Implications for Theory and Practice
This study is the first meta-analysis that systematically and
comprehensively investigated the relation between reading and
WM and the moderators of that relation. Findings of this study
have implications for theories of WM and reading as well as for
reading instruction and WM training. Specifically, we applied
three cognitive theories (i.e., domain-specificity theory/debate of
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
66 PENG ET AL.
WM, intrinsic cognitive load theory, dual process theory) to de-
termine candidates for variables that might affect the relation
between reading and WM (e.g., domains of WM, types of reading,
and grade level). In contrast to predictions based on the intrinsic
cognitive load theory, the seeming complexity of reading tasks was
not the critical factor in determining the size of the relation
between WM and reading. Several of the findings align with the
dual process theory, suggesting that development (i.e., reading
experience) plays an important role in the relation between reading
and WM. The development also affects the domain-specific nature
of the relation of WM to reading.
By integrating the results of the meta-analysis in the context of
the dual process theory, debates about domain-specificity of WM,
and empirical findings that learning to read may help shape verbal
memory (Demoulin & Kolinsky, 2016), we tentatively propose a
“Working Memory-Reading Development” model. Based on this
model, WM, especially the domain-general central executive com-
ponent (Baddeley, 1986), should be heavily involved in reading in
the early stages of reading. As reading experience accumulates,
lexical and verbal knowledge is consolidated in long-term mem-
ory, and readers come to rely more on direct retrieval of lexical/
verbal knowledge from long-term memory to perform a variety of
reading tasks. As children are developing foundational reading
skills and attempting to read for understanding, WM resources
may be allocated to integrate verbal knowledge and procedures to
meet the demands of reading tasks, strengthening verbal WM and
the impact of verbal WM on reading in the process. In this model,
the relation between reading and WM varies as a function of
development: WM primarily exerts an impact on reading early on,
with reading also shaping the further development of verbal WM
in particular.
Our findings also have implications for reading instruction.
Although WM did not contribute to reading comprehension after
controlling for both decoding and vocabulary, it should be noted
that WM still makes made unique contributions to reading com-
prehension after controlling for only decoding or only vocabulary.
Moreover, as mentioned earlier, vocabulary and world knowledge
have large direct effects on reading comprehension, but also indi-
rect effects on reading comprehension, via inferencing and com-
prehension strategies (e.g., Ahmed et al., 2016;Cromley &
Azevedo, 2007;Cromley, Snyder-Hogan, & Luciw-Dubas, 2010).
These findings, together with those of the present meta-analysis,
suggest that to facilitate early reading comprehension, instruction
in both decoding and vocabulary (i.e., to link orthography, pho-
nology, and semantics at the word-level) may reduce WM load
during reading (Perfetti, 2007). The findings also suggest that
instructional procedures that reduce WM load during the acquisi-
tion of word codes and their meanings may also facilitate reading
comprehension in young readers.
Our findings may also have implications for WM training re-
search. Specifically, WM showed stronger relations with reading
in early grades versus later grades. We suggested that this finding
may reflect younger children’s greater need to rely more on WM
during reading, because their word decoding and semantic retrieval
processes are less efficient than those of more experienced readers.
Thus, WM training in early grades may have the strongest effects
on reading. This suggestion is also in line with the age effects
found in cognitive training. That is, cognitive training may be most
effective for younger individuals (e.g., Peng & Miller, 2016;Wass,
Scerif, & Johnson, 2012).
However, it is important to note that although WM is correlated
with reading, the strength of these correlations is moderate, rang-
ing from .22 to .37. Translating these correlations into variances,
WM accounts for relatively low amounts of variance in reading
performance, ranging from 5% to 14% with an average around 9%.
These numbers indicate that even if WM training can produce
transfer effects on reading performance, the transfer effects would
likely be small, suggesting that training WM alone may be insuf-
ficient for improving reading performance (Jacob & Parkinson,
2015). Given our findings that verbal WM and reading may
become more important to each other in the process of reading
development, we propose that it may be best to choose verbal WM
tasks as the WM training tasks and combine this training with
reading instruction to maximize the WM training effects on read-
ing. That said, the effectiveness of hybrid interventions that com-
bine WM and skills specific (i.e., reading) instruction has yet to be
determined.
Conclusion
In summary, the current meta-analysis investigated the relation
between reading and WM and the main findings provide new
information for the field as follows: (a) WM showed moderate
relations with reading, and, in contrast to several hypotheses about
the relations of WM to reading, these relations were as strong for
more foundational reading skills as they were for comprehension;
(b) after controlling for both decoding and general language (vo-
cabulary), WM was no longer related to reading comprehension,
which is generally consistent with the Simple View of Reading; (c)
the relation of WM to reading fit a domain-general account early
in development, but became more domain-specific with greater
reading experience; and (d) there were several other grade-related
moderation in the relation of reading and WM, including a stronger
relation of WM to reading comprehension versus listening com-
prehension in younger versus older readers. In general, these
grade-related findings suggest that the relation between reading
and WM needs to be considered within a developmental context.
References
References marked with an asterisk indicate studies included in the
meta-analysis.
Abu-Rabia, S. (2005). Social aspects and reading, writing, and working
memory skills in Arabic, Hebrew, English, and Circassian: The quadri-
lingual case of Circassians. Language, Culture and Curriculum, 18,
27–58. http://dx.doi.org/10.1080/07908310508668732
Abu-Rabia, S., & Siegel, L. S. (2002). Reading, syntactic, orthographic,
and working memory skills of bilingual Arabic-English speaking Cana-
dian children. Journal of Psycholinguistic Research, 31, 661– 678. http://
dx.doi.org/10.1023/A:1021221206119
Abu-Rabia, S., & Siegel, L. S. (2003). Reading skills in three orthogra-
phies: The case of trilingual Arabic–Hebrew–English-speaking Arab
children. Reading and Writing, 16, 611– 634. http://dx.doi.org/10.1023/
A:1025838029204
Adams, A. M., Bourke, L., & Willis, C. (1999). Working memory and
spoken language comprehension in young children. International Jour-
nal of Psychology, 34, 364 –373. http://dx.doi.org/10.1080/00207
5999399701
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
67
READING AND WORKING MEMORY: A META-ANALYSIS
Adlof, S. M., Catts, H. W., & Little, T. D. (2006). Should the simple view
of reading include a fluency component? Reading and Writing, 19,
933–958. http://dx.doi.org/10.1007/s11145-006-9024-z
Ahmed, Y., Francis, D. J., York, M., Fletcher, J. M., Barnes, M., & Kulesz,
P. (2016). Validation of the direct and inferential mediation (DIME)
model of reading comprehension in grades 7 through 12. Contemporary
Educational Psychology, 44, 68 – 82. http://dx.doi.org/10.1016/j
.cedpsych.2016.02.002
Alexander, J. (1993). The relationship of automaticity, metacognition, and
working memory in normal and learning disabled readers (Doctoral
dissertation). Retrieval from The University of British Columbia.
Alloway, T. P., Gathercole, S. E., Kirkwood, H., & Elliott, J. (2009). The
cognitive and behavioral characteristics of children with low working
memory. Child Development, 80, 606 – 621. http://dx.doi.org/10.1111/j
.1467-8624.2009.01282.x
Andersson, U. (2010). The contribution of working memory capacity to
foreign language comprehension in children. Memory, 18, 458 – 472.
http://dx.doi.org/10.1080/09658211003762084
Andrews, G., Birney, D., & Halford, G. S. (2006). Relational processing
and working memory capacity in comprehension of relative clause
sentences. Memory & Cognition, 34, 1325–1340. http://dx.doi.org/10
.3758/BF03193275
Arnell, K. M., Stokes, K. A., MacLean, M. H., & Gicante, C. (2010).
Executive control processes of working memory predict attentional blink
magnitude over and above storage capacity. Psychological Research, 74,
1–11. http://dx.doi.org/10.1007/s00426-008-0200-4
Arrington, C. N., Kulesz, P. A., Francis, D. J., Fletcher, J. M., & Barnes,
M. A. (2014). The contribution of attentional control and working
memory to reading comprehension and decoding. Scientific Studies of
Reading, 18, 325–346. http://dx.doi.org/10.1080/10888438.2014
.902461
Babayig
˘it, S. (2015). The relations between word reading, oral language,
and reading comprehension in children who speak English as a first (L1)
and second language (L2): A multigroup structural analysis. Reading
and Writing, 28, 527–544. http://dx.doi.org/10.1007/s11145-014-9536-x
Baddeley, A. D. (1979). Working memory and reading. In P. A. Kolers,
M. E. Wrolstad, & H. Bouma (Eds.), Processing of visible language (pp.
355–370). New York, NY: Springer US. http://dx.doi.org/10.1007/978-
1-4684-0994-9_21
Baddeley, A. D. (1986). Working memory. New York, NY: Oxford Uni-
versity Press.
Baddeley, A., Logie, R., Nimmo-Smith, I., & Brereton, N. (1985). Com-
ponents of fluent reading. Journal of Memory and Language, 24, 119 –
131. http://dx.doi.org/10.1016/0749-596X(85)90019-1(85)90019-1
Baek, S. (2012). Processing of English relative clauses by adult L2
learners (Doctoral dissertation). Retrieval from University of Illinois at
Urbana-Champaign.
Bagner, D. M., Melinder, M. R., & Barch, D. M. (2003). Language
comprehension and working memory language comprehension and
working memory deficits in patients with schizophrenia. Schizophrenia
Research, 60, 299 –309. http://dx.doi.org/10.1016/S0920-9964(02)
00280-3
Bailey, H., Dunlosky, J., & Kane, M. J. (2008). Why does working
memory span predict complex cognition? Testing the strategy affor-
dance hypothesis. Memory & Cognition, 36, 1383–1390. http://dx.doi
.org/10.3758/MC.36.8.1383
Baker, L. (1985). Working memory and comprehension: A replication.
Bulletin of the Psychonomic Society, 23, 28 –30. http://dx.doi.org/10
.3758/BF03329770
Bargh, J. A. (1994). The four horsemen of automaticity: Intention, aware-
ness, efficiency, and control as separate issues. In R. S. Wyer & T. K.
Srull (Eds.), Handbook of social cognition: Vol. 1, basic processes (2nd
ed., pp. 1– 40). Hillsdale, NJ: Erlbaum.
Barnes, M. A., Klein, A., Swank, P., Starkey, P., McCandliss, B., Flynn,
K.,...Roberts, G. (2016). Effects of tutorial interventions in mathe-
matics and attention for low-performing preschool children. Journal of
Research on Educational Effectiveness, 9, 577– 606. http://dx.doi.org/10
.1080/19345747.2016.1191575
Bayliss, D. M., Jarrold, C., Gunn, D. M., & Baddeley, A. D. (2003). The
complexities of complex span: Explaining individual differences in
working memory in children and adults. Journal of Experimental Psy-
chology: General, 132, 71–92. http://dx.doi.org/10.1037/0096-3445.132
.1.71
Berninger, V. W., Abbott, R. D., Swanson, H. L., Lovitt, D., Trivedi, P.,
Lin, S. J. C.,...Amtmann, D. (2010). Relationship of word- and
sentence-level working memory to reading and writing in second, fourth,
and sixth grade. Language, Speech, and Hearing Services in Schools, 41,
179 –193. http://dx.doi.org/10.1044/0161-1461(2009/08-0002)
Best, R. M., Floyd, R. G., & Mcnamara, D. S. (2008). Differential com-
petencies contributing to children’s comprehension of narrative and
expository texts. Reading Psychology, 29, 137–164. http://dx.doi.org/10
.1080/02702710801963951
Borenstein, M., Hedges, L., Higgins, J., & Rothstein, H. (2005). Compre-
hensive meta-analysis (Version 2.2.040) [Computer software]. Engle-
wood, NJ: Biostat.
Bourke, L., & Adams, A. M. (2003). The relationship between working
memory and early writing assessed at the word, sentence and text level.
Educational and Child Psychology, 20, 19 –36.
Bowey, J. A., Cain, M. T., & Ryan, S. M. (1992). A reading-level design
study of phonological skills underlying fourth-grade children’s word
reading difficulties. Child Development, 63, 999 –1011. http://dx.doi
.org/10.2307/1131249
Bradley, L., & Bryant, P. E. (1978). Difficulties in auditory organisation as
a possible cause of reading backwardness. Nature, 271, 746 –747. http://
dx.doi.org/10.1038/271746a0
Bresnahan, B. (2006). Component processes in the predictors of reading
achievement: Direct and indirect effects (Doctoral dissertation). Re-
trieval from ProQuest Dissertation. (Accession No. 3238536)
Cain, K., & Bignell, S. (2014). Reading and listening comprehension and
their relation to inattention and hyperactivity. British Journal of Educa-
tional Psychology, 84, 108 –124. http://dx.doi.org/10.1111/bjep.12009
Cain, K., & Oakhill, J. (2006). Profiles of children with specific reading
comprehension difficulties. British Journal of Educational Psychology,
76, 683– 696. http://dx.doi.org/10.1348/000709905X67610
Cain, K., Oakhill, J., & Bryant, P. (2004). Children’s reading comprehen-
sion ability: Concurrent prediction by working memory, verbal ability,
and component skills. Journal of Educational Psychology, 96, 31– 42.
http://dx.doi.org/10.1037/0022-0663.96.1.31
Cain, K., Oakhill, J., & Lemmon, K. (2004). Individual differences in the
inference of word meanings from context: The influence of reading
comprehension, vocabulary knowledge, and memory capacity. Journal
of Educational Psychology, 96, 671– 681. http://dx.doi.org/10.1037/
0022-0663.96.4.671
Caplan, D., Dede, G., Waters, G., Michaud, J., & Tripodis, Y. (2011).
Effects of age, speed of processing, and working memory on compre-
hension of sentences with relative clauses. Psychology and Aging, 26,
439 – 450. http://dx.doi.org/10.1037/a0021837
Caplan, D., & Waters, G. (2005). The relationship between age, process-
ing speed, working memory capacity, and language comprehension.
Memory, 13, 403– 413. http://dx.doi.org/10.1080/09658210344000459
Cates, D. M. (2015). Predictors of reading comprehension skill in deaf
and hearing bilingual readers (Doctoral dissertation). Retrieval from
ProQuest Dissertation. (Accession No. 3706567)
Chall, J. S. (1983). Learning to read: The great debate. New York, NY:
McGraw-Hill.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
68 PENG ET AL.
Chandler, P., & Sweller, J. (1991). Cognitive load theory and the format of
instruction. Cognition and Instruction, 8, 293–332. http://dx.doi.org/10
.1207/s1532690xci0804_2
Chang, I. (2004). The impact of cognitive and ecological variables on
children’s reading achievement (Doctoral dissertation). Retrieval from
ProQuest Dissertation. (Accession No. 3125821)
Chapman, S. (1995). The role of individual differences in memory pro-
cessing on recall of text. (Master dissertation). Retrieval from Carleton
University.
Chavez, N. M. (2002). Individual differences in verbal working memory,
visuo-spatial working memory, and metacognition: Learning from text
in a hypertext environment (Doctoral dissertation). Retrieval from Pro-
Quest Dissertation. (Accession No. 3056101)
Cherry, K. E., & Park, D. C. (1993). Individual difference and contextual
variables influence spatial memory in younger and older adults. Psy-
chology and Aging, 8, 517–526. http://dx.doi.org/10.1037/0882-7974.8
.4.517
Chiappe, D. L., & Chiappe, P. (2007). The role of working memory in
metaphor production and comprehension. Journal of Memory and Lan-
guage, 56, 172–188. http://dx.doi.org/10.1016/j.jml.2006.11.006
Christopher, M. E., Miyake, A., Keenan, J. M., Pennington, B., DeFries,
J. C., Wadsworth, S. J.,...Olson, R. K. (2012). Predicting word reading
and comprehension with executive function and speed measures across
development: A latent variable analysis. Journal of Experimental Psy-
chology: General, 141, 470 – 488. http://dx.doi.org/10.1037/a0027375
Cirino, P. T., Fuchs, L. S., Elias, J. T., Powell, S. R., & Schumacher, R. F.
(2015). Cognitive and mathematical profiles for different forms of
learning difficulties. Journal of Learning Disabilities, 48, 156 –175.
http://dx.doi.org/10.1177/0022219413494239
Clarke, P. J., Snowling, M. J., Truelove, E., & Hulme, C. (2010). Ame-
liorating children’s reading-comprehension difficulties: A randomized
controlled trial. Psychological Science, 21, 1106 –1116. http://dx.doi
.org/10.1177/0956797610375449
Cohen, J. (1988). Statistical power analysis for the behavioral sciences
(2nd ed.). Hillsdale, NJ: Erlbaum.
Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2013). Applied multiple
regression/correlation analysis for the behavioral sciences. London,
UK: Routledge.
Coltheart, V., & Leahy, J. (1992). Children’s and adults’ reading of
nonwords: Effects of regularity and consistency. Journal of Experimen-
tal Psychology: Learning, Memory, and Cognition, 18, 718 –729. http://
dx.doi.org/10.1037/0278-7393.18.4.718
Cormier, P., & Dea, S. (1997). Distinctive patterns of relationship of
phonological awareness and working memory with reading develop-
ment. Reading and Writing, 9, 193–206. http://dx.doi.org/10.1023/A:
1007932721290
Cromley, J. G., & Azevedo, R. (2007). Testing and refining the direct and
inferential mediation model of reading comprehension. Journal of Ed-
ucational Psychology, 99, 311–325. http://dx.doi.org/10.1037/0022-
0663.99.2.311
Cromley, J. G., Snyder-Hogan, L. E., & Luciw-Dubas, U. A. (2010).
Reading comprehension of scientific text: A domain-specific test of the
direct and inferential mediation model of reading comprehension. Jour-
nal of Educational Psychology, 102, 687–700. http://dx.doi.org/10.1037/
a0019452
Currie, N. K., & Cain, K. (2015). Children’s inference generation: The
role of vocabulary and working memory. Journal of Experimental Child
Psychology, 137, 57–75. http://dx.doi.org/10.1016/j.jecp.2015.03.005
Cutting, L. E., Clements-Stephens, A., Pugh, K. R., Burns, S., Cao, A.,
Pekar, J. J.,...Rimrodt, S. L. (2013). Not all reading disabilities are
dyslexia: Distinct neurobiology of specific comprehension deficits.
Brain Connectivity, 3, 199 –211. http://dx.doi.org/10.1089/brain.2012
.0116
Da Fontoura, H. A., & Siegel, L. S. (1995). Reading, syntactic, and
working memory skills of bilingual Portuguese-English Canadian chil-
dren. Reading and Writing, 7, 139 –153. http://dx.doi.org/10.1007/
BF01026951
Dahlin, K. I. (2011). Effects of working memory training on reading in
children with special needs. Reading and Writing, 24, 479 – 491. http://
dx.doi.org/10.1007/s11145-010-9238-y
Daneman, M. (1991). Working memory as a predictor of verbal fluency.
Journal of Psycholinguistic Research, 20, 445– 464. http://dx.doi.org/10
.1007/BF01067637
Daneman, M., & Carpenter, P. A. (1980). Individual differences in work-
ing memory and reading. Journal of Verbal Learning & Verbal Behav-
ior, 19, 450 – 466. http://dx.doi.org/10.1016/S0022-5371(80)90312-6
Daneman, M., & Green, I. (1986). Individual differences in comprehend-
ing and producing words in context. Journal of Memory and Language,
25, 1–18. http://dx.doi.org/10.1016/0749-596X(86)90018-5
Daneman, M., & Hannon, B. (2001). Using working memory theory to
investigate the construct validity of multiple-choice reading comprehen-
sion tests such as the SAT. Journal of Experimental Psychology: Gen-
eral, 130, 208 –223. http://dx.doi.org/10.1037/0096-3445.130.2.208
Daneman, M., & Merikle, P. M. (1996). Working memory and language
comprehension: A meta-analysis. Psychonomic Bulletin & Review, 3,
422– 433. http://dx.doi.org/10.3758/BF03214546
De Beni, R., & Palladino, P. (2000). Intrusion errors in working memory
tasks: Are they related to reading comprehension ability? Learning and
Individual Differences, 12, 131–143. http://dx.doi.org/10.1016/S1041-
6080(01)00033-4
DeDe, G., Caplan, D., Kemtes, K., & Waters, G. (2004). The relationship
between age, verbal working memory, and language comprehension.
Psychology and Aging, 19, 601– 616. http://dx.doi.org/10.1037/0882-
7974.19.4.601
Demoulin, C., & Kolinsky, R. (2016). Does learning to read shape verbal
working memory? Psychonomic Bulletin & Review, 23, 703–722. http://
dx.doi.org/10.3758/s13423-015-0956-7
Dixon, P., LeFevre, J. A., & Twilley, L. C. (1988). Word knowledge and
working memory as predictors of reading skill. Journal of Educational
Psychology, 80, 465– 472. http://dx.doi.org/10.1037/0022-0663.80.4
.465
Dreyer, L. G., & Katz, L. (1992). An examination of “the simple view of
reading”. National Reading Conference Yearbook, 41, 169 –175.
Dunlap, R. D. (1997). An assessment of visuo-spatial and phonological
components of working memory (Doctoral dissertation). Retrieval from
Texas Tech University.
Dunlosky, J., & Kane, M. J. (2007). The contributions of strategy use to
working memory span: A comparison of strategy assessment methods.
Quarterly Journal of Experimental Psychology: Human Experimental
Psychology, 60, 1227–1245. http://dx.doi.org/10.1080/17470210600
926075
Echols, L. D., West, R. F., Stanovich, K. E., & Zehr, K. S. (1996). Using
children’s literacy activities to predict growth in verbal cognitive skills:
A longitudinal investigation. Journal of Educational Psychology, 88,
296 –304. http://dx.doi.org/10.1037/0022-0663.88.2.296
Egger, M., Davey Smith, G., Schneider, M., & Minder, C. (1997). Bias in
meta-analysis detected by a simple, graphical test. British Medical
Journal, 315, 629 – 634. http://dx.doi.org/10.1136/bmj.315.7109.629
Ehri, L. C. (1992). Reconceptualizing the development of sight word
reading and its relationship to recoding. Mahwah, NJ: Erlbaum.
Ellis Weismer, S., Evans, J., & Hesketh, L. J. (1999). An examination of
verbal working memory capacity in children with specific language
impairment. Journal of Speech, Language, and Hearing Research, 42,
1249 –1260. http://dx.doi.org/10.1044/jslhr.4205.1249
Engle, R. W. (2002). Working memory capacity as executive attention.
Current Directions in Psychological Science, 11, 19 –23. http://dx.doi
.org/10.1111/1467-8721.00160
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
69
READING AND WORKING MEMORY: A META-ANALYSIS
Engle, R. W., Cantor, J., & Carullo, J. J. (1992). Individual differences in
working memory and comprehension: A test of four hypotheses. Journal
of Experimental Psychology: Learning, Memory, and Cognition, 18,
972–992. http://dx.doi.org/10.1037/0278-7393.18.5.972
Engle, R. W., Carullo, J. J., & Collins, K. W. (1991). Individual differ-
ences in working memory for comprehension and following directions.
The Journal of Educational Research, 84, 253–262. http://dx.doi.org/10
.1080/00220671.1991.10886025
Engle, R. W., & Kane, M. J. (2003). Executive attention, working memory
capacity, and a two-factor theory of cognitive control. Psychology of
Learning and Motivation, 44, 145–199. http://dx.doi.org/10.1016/
S0079-7421(03)44005-X
Engle, R. W., Nations, J. K., & Cantor, J. (1990). Is “working memory
capacity” just another name for word knowledge? Journal of Educa-
tional Psychology, 82, 799 – 804. http://dx.doi.org/10.1037/0022-0663
.82.4.799
Ericsson, K. A., & Kintsch, W. (1995). Long-term working memory.
Psychological Review, 102, 211–245. http://dx.doi.org/10.1037/0033-
295X.102.2.211
Etmanskie, J. M., Partanen, M., & Siegel, L. S. (2016). A longitudinal
examination of the persistence of late emerging reading disabilities.
Journal of Learning Disabilities, 49, 21–35. http://dx.doi.org/10.1177/
0022219414522706
Evans, J. S. B., & Stanovich, K. E. (2013). Dual-process theories of higher
cognition advancing the debate. Perspectives on Psychological Science,
8, 223–241. http://dx.doi.org/10.1177/1745691612460685
Ewers, C. A. (1988). An information processing analysis of working
memory in relation to listening comprehension in young children (Doc-
toral dissertation). Retrieval from ProQuest Dissertation. (Accession No.
8821342)
Felbinger, L. C. (1997). A component skills analysis of college-level
reading comprehension (Doctoral dissertation). Retrieval from ProQuest
Dissertation. (Accession No. 9731253)
Freed, E. M. (2010). Does administration format influence the construct
validity of SAT comprehension passages? (Doctoral dissertation). Re-
trieval from ProQuest Dissertation. (Accession No. 1475841)
Friedman, N. P., & Miyake, A. (2004). The reading span test and its
predictive power for reading comprehension ability. Journal of Memory
and Language, 51, 136 –158. http://dx.doi.org/10.1016/j.jml.2004.03
.008
Friedman, N. P., & Miyake, A. (2004). The relations among inhibition and
interference control functions: A latent-variable analysis. Journal of
Experimental Psychology: General, 133, 101–135. http://dx.doi.org/10
.1037/0096-3445.133.1.101
Frost, R., & Katz, L. (1992). Orthography, phonology, morphology and
meaning. Amsterdam, the Netherlands: Elsevier.
Fuchs, D., Peng, P., Elleman, A., Kearns, D., Fuchs, L., & Miller, A.
(2015). Explaining the value of working memory training when com-
bined with skills-based instruction in reading comprehension for young
at-risk students. SREE Spring Conference, Washington, DC.
Fuchs, L. S., Compton, D. L., Fuchs, D., Powell, S. R., Schumacher, R. F.,
Hamlett, C. L.,...Vukovic, R. K. (2012). Contributions of domain-
general cognitive resources and different forms of arithmetic develop-
ment to pre-algebraic knowledge. Developmental Psychology, 48, 1315–
1326. http://dx.doi.org/10.1037/a0027475
Fuchs, L. S., Fuchs, D., Compton, D. L., Hamlett, C. L., & Wang, A. Y.
(2015). Is word-problem solving a form of text comprehension? Scien-
tific Studies of Reading, 19, 204 –223. http://dx.doi.org/10.1080/
10888438.2015.1005745
Fuchs, L. S., Fuchs, D., Compton, D. L., Powell, S. R., Seethaler, P. M.,
Capizzi, A. M.,...Fletcher, J. M. (2006). The cognitive correlates of
third-grade skill in arithmetic, algorithmic computation, and arithmetic
word problems. Journal of Educational Psychology, 98, 29 – 43. http://
dx.doi.org/10.1037/0022-0663.98.1.29
Fuhs, M. W., Farran, D. C., & Nesbitt, K. T. (2015). Prekindergarten
children’s executive functioning skills and achievement gains: The util-
ity of direct assessments and teacher ratings. Journal of Educational
Psychology, 107, 207–221. http://dx.doi.org/10.1037/a0037366
Gathercole, S. E., & Pickering, S. J. (2000). Assessment of working
memory in six-and seven-year-old children. Journal of Educational
Psychology, 92, 377–390. http://dx.doi.org/10.1037/0022-0663.92.2.377
Georgiou, G. K., & Das, J. P. (2016). What component of executive
functions contributes to normal and impaired reading comprehension in
young adults? Research in Developmental Disabilities, 49 –50, 118 –
128. http://dx.doi.org/10.1016/j.ridd.2015.12.001
Gholamain, M. (1992). The role of orthography and cognitive factors in
the concurrent development of basic reading skills in bilingual Persian-
English speaking children (Doctoral dissertation). Retrieved from Pro-
Quest Dissertations Publishing. (MM78421)
Gillespie, M., James, A. N., Federmeier, K. D., & Watson, D. G. (2014).
Verbal working memory predicts co-speech gesture: Evidence from
individual differences. Cognition, 132, 174 –180. http://dx.doi.org/10
.1016/j.cognition.2014.03.012
Goff, D. A., Pratt, C., & Ong, B. (2005). The relations between children’s
reading comprehension, working memory, language skills and compo-
nents of reading decoding in a normal sample. Reading and Writing, 18,
583– 616. http://dx.doi.org/10.1007/s11145-004-7109-0
Gonthier, C., & Thomassin, N. (2015). Strategy use fully mediates the
relationship between working memory capacity and performance on
Raven’s matrices. Journal of Experimental Psychology: General, 144,
916 –924. http://dx.doi.org/10.1037/xge0000101
Gottardo, A., Stanovich, K. E., & Siegel, L. S. (1996). The relationships
between phonological sensitivity, syntactic processing, and verbal work-
ing memory in the reading performance of third-grade children. Journal
of Experimental Child Psychology, 63, 563–582. http://dx.doi.org/10
.1006/jecp.1996.0062
Gough, P. B., & Tunmer, W. E. (1986). Decoding, reading, and reading
disability. Remedial and Special Education, 7, 6 –10. http://dx.doi.org/
10.1177/074193258600700104
Graesser, A. C., McNamara, D. S., & Louwerse, M. M. (2003). What do
readers need to learn in order to process coherence relations in narrative
and expository text. In A. P. Sweet & C. E. Snow (Eds.), Rethinking
reading comprehension (pp. 82–98). New York, NY: Guilford.
Guan, Q. (2007). The activation and long-term memory of predictive
inferences: The role of working memory constraint and text elaboration
(Doctoral dissertation). Retrieval from ProQuest Dissertation. (Acces-
sion No. 3301551)
Gustafson, M. S. (2001). The role of conceptual flexibility in reading
comprehension: The moment-to-moment case of selecting meaning
(Doctoral dissertation). Retrieval from ProQuest Dissertation. (Acces-
sion No. 3010554)
Gutiérrez-Clellen, V. F., Calderón, J., & Ellis Weismer, S. (2004). Verbal
working memory in bilingual children. Journal of Speech, Language,
and Hearing Research, 47, 863– 876. http://dx.doi.org/10.1044/1092-
4388(2004/064)
Hambrick, D. Z., & Engle, R. W. (2002). Effects of domain knowledge,
working memory capacity, and age on cognitive performance: An in-
vestigation of the knowledge-is-power hypothesis. Cognitive Psychol-
ogy, 44, 339 –387. http://dx.doi.org/10.1006/cogp.2001.0769
Hamilton, S. T., & Oakhill, J. V. (2014). Establishing coherence across
sentence boundaries: An individual differences approach. Language,
Cognition and Neuroscience, 29, 1240 –1248. http://dx.doi.org/10.1080/
01690965.2013.863368
Hannon, B., & Daneman, M. (2001). A new tool for measuring and
understanding individual differences in the component processes of
reading comprehension. Journal of Educational Psychology, 93, 103–
128. http://dx.doi.org/10.1037/0022-0663.93.1.103
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
70 PENG ET AL.
Hannon, B., & Frias, S. (2012). A new measure for assessing the contri-
butions of higher level processes to language comprehension perfor-
mance in preschoolers. Journal of Educational Psychology, 104, 897–
921. http://dx.doi.org/10.1037/a0029156
Hansen, N. M. (2015). Development of fraction knowledge: A longitudinal
study from third through sixth grade (Doctoral dissertation). Retrieval
from ProQuest Dissertation. (Accession No. 3718335)
Hargrave, J. L. (2004). The relationship between executive functions and
broad written language skills in students ages 12 to 14 years old
(Doctoral dissertation). Retrieval from The University of Texas at Aus-
tin.
Harrison, G. L., Goegan, L. D., Jalbert, R., McManus, K., Sinclair, K., &
Spurling, J. (2016). Predictors of spelling and writing skills in first-and
second-language learners. Reading and Writing, 29, 69 – 89. http://dx
.doi.org/10.1007/s11145-015-9580-1
Hawkins, C. W. (1995). The differential effects of word recognition and
listening comprehension on reading comprehension of narrative and
expository text in grades one and three (Doctoral dissertation). Retrieval
from ProQuest Dissertation. (Accession No. 9542602)
He, Y. (2007). Influence of reading proficiency on text representation in L1
and L2 (Unpublished Dissertation). University of Illinois, Urban-
Champaign, IL.
Hedberg, E. C. (2014). ROBUMETA: Stata module to perform robust
variance estimation in meta-regression with dependent effect size esti-
mates. Statistical Software Components S457219. Boston, MA: Boston
College Department of Economics.
Hedges, L. V., Tipton, E., & Johnson, M. C. (2010). Robust variance
estimation in meta-regression with dependent effect size estimates. Re-
search Synthesis Methods, 1, 39 – 65. http://dx.doi.org/10.1002/jrsm.5
Henry, L., & MacLean, M. (2003). Relationships between working mem-
ory, expressive vocabulary and arithmetical reasoning in children with
and without intellectual disabilities. Educational and Child Psychology,
20, 51– 63.
Henry, L., & Winfield, J. (2010). Working memory and educational
achievement in children with intellectual disabilities. Journal of Intel-
lectual Disability Research, 54, 354 –365. http://dx.doi.org/10.1111/j
.1365-2788.2010.01264.x
Hester, E. (2000). Phonological production, reading decoding, and work-
ing memory in third graders (Doctoral dissertation). Retrieval from
ProQuest Dissertation. (Accession No. 9978802)
Hester, E., & Hodson, B. W. (2004). The role of phonological represen-
tation in decoding skills of young readers. Child Language Teaching and
Therapy, 20, 115–133. http://dx.doi.org/10.1191/0265659004ct266oa
Hitch, G. J., Towse, J. N., & Hutton, U. (2001). What limits children’s
working memory span? Theoretical accounts and applications for scho-
lastic development. Journal of Experimental Psychology: General, 130,
184 –198. http://dx.doi.org/10.1037/0096-3445.130.2.184
Holsgrove, J. V., & Garton, A. F. (2006). Phonological and syntactic
processing and the role of working memory in reading comprehension
among secondary school students. Australian Journal of Psychology, 58,
111–118. http://dx.doi.org/10.1080/00049530600730476
Hoover, W. A., & Gough, P. B. (1990). The simple view of reading.
Reading and Writing, 2, 127–160. http://dx.doi.org/10.1007/
BF00401799
Hopp, H. (2015). Individual differences in the second language processing
of object–subject ambiguities. Applied Psycholinguistics, 36, 129 –173.
http://dx.doi.org/10.1017/S0142716413000180
Howes, A. A. (1999). Exploration of a tactile working memory component
in foreign language learning among university students (Doctoral dis-
sertation). Retrieval from ProQuest Dissertation. (Accession No.
9952419)
Hulme, C., & Snowling, M. J. (2009). Developmental disorders of lan-
guage learning and cognition. Hoboken, NJ: Wiley.
Jacob, R., & Parkinson, J. (2015). The potential for school-based interven-
tions that target executive function to improve academic achievement: A
review. Review of Educational Research, 85, 512–552. http://dx.doi.org/
10.3102/0034654314561338
Jean, M. W. L. (2005). L2 vocabulary knowledge: A predictor of L2 word
recognition skills in EL1 and ESL upper-elementary children? (Doctoral
dissertation). Retrieval from ProQuest Dissertation. (Accession No.
0494073020)
Johnston, A. M. (2007). The role of attentional control in reading com-
prehension (Doctoral dissertation). Retrieval from ProQuest Disserta-
tion. (Accession No. 9780494254349)
Jongejan, W., Verhoeven, L., & Siegel, L. S. (2007). Predictors of reading
and spelling abilities in first-and second-language learners. Journal of
Educational Psychology, 99, 835– 851. http://dx.doi.org/10.1037/0022-
0663.99.4.835
Jurden, F. H. (1995). Individual differences in working memory and
complex cognition. Journal of Educational Psychology, 87, 93–102.
http://dx.doi.org/10.1037/0022-0663.87.1.93
Just, M. A., & Carpenter, P. A. (1992). A capacity theory of comprehen-
sion: Individual differences in working memory. Psychological Review,
99, 122–149. http://dx.doi.org/10.1037/0033-295X.99.1.122
Kail, R., & Hall, L. K. (2001). Distinguishing short-term memory from
working memory. Memory & Cognition, 29, 1–9. http://dx.doi.org/10
.3758/BF03195735
Karasinski, C., & Ellis Weismer, S. (2010). Comprehension of inferences
in discourse processing by adolescents with and without language im-
pairment. Journal of Speech, Language, and Hearing Research, 53,
1268 –1279. http://dx.doi.org/10.1044/1092-4388(2009/09-0006)
Keenan, J. M., Betjemann, R. S., & Olson, R. K. (2008). Reading com-
prehension tests vary in the skills they assess: Differential dependence
on decoding and oral comprehension. Scientific Studies of Reading, 12,
281–300. http://dx.doi.org/10.1080/10888430802132279
Kemper, S., & Sumner, A. (2001). The structure of verbal abilities in
young and older adults. Psychology and Aging, 16, 312–322. http://dx
.doi.org/10.1037/0882-7974.16.2.312
Kemtes, K. A. (1999). Decomposing adults’ sentence comprehension: The
roles of age, working memory, inhibitory functioning, and perceptual
speed (Doctoral dissertation). Retrieval from ProQuest Dissertation.
(Accession No. 9920350)
Kim, K. H. (2007). A structural equation model of the factors impacting
problem success of algebra word problems: Reading, math, working
memory and conceptual knowledge (Doctoral dissertation). Retrieval
from ProQuest Dissertation. (Accession No. 3270438)
Kim, S. (2012). Visuals in the mind’s eye: Investigating graph compre-
hension in students with dyslexia (Doctoral dissertation). Retrieval from
ProQuest Dissertation. (Accession No. 3585186)
Kim, Y. S. G. (2016). Direct and mediated effects of language and
cognitive skills on comprehension of oral narrative texts (listening
comprehension) for children. Journal of Experimental Child Psychol-
ogy, 141, 101–120. http://dx.doi.org/10.1016/j.jecp.2015.08.003
Kintsch, W. (1998). Comprehension: A paradigm for cognition. Cam-
bridge, UK: Cambridge University Press.
Koltum, H. S. (2003). Narrative comprehension and production abilities of
children with attention-deficit/hyperactivity disorder (Doctoral disserta-
tion). Retrieval from University of Toronto.
Koshino, H., Carpenter, P. A., Minshew, N. J., Cherkassky, V. L., Keller,
T. A., & Just, M. A. (2005). Functional connectivity in an fMRI working
memory task in high-functioning autism. NeuroImage, 24, 810 – 821.
http://dx.doi.org/10.1016/j.neuroimage.2004.09.028
Kulesz, P. A. (2014). The effects of reader characteristics, text features,
and comprehension processes on reading comprehension (Doctoral dis-
sertation). Retrieval from University of Houston.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
71
READING AND WORKING MEMORY: A META-ANALYSIS
Lafrance, A., & Gottardo, A. (2005). A longitudinal study of phonological
processing skills and reading in bilingual children. Applied Psycholin-
guistics, 26, 559 –578. http://dx.doi.org/10.1017/S0142716405050307
Lafrance, C. A. (2003). French and English word reading and word
decoding performance in bilingual children: How are phonological
processing skills related? (Doctoral dissertation). Retrieval from Wilfrid
Laurier University.
Laing, S. P., & Kamhi, A. G. (2002). The use of think-aloud protocols to
compare inferencing abilities in average and below-average readers.
Journal of Learning Disabilities, 35, 437– 447. http://dx.doi.org/10
.1177/00222194020350050401
Lanfranchi, S., & Swanson, H. L. (2005). Short-term memory and work-
ing memory in children as a function of language-specific knowledge in
English and Spanish. Learning and Individual Differences, 15, 299 –319.
http://dx.doi.org/10.1016/j.lindif.2005.05.003
La Pointe, L. B., & Engle, R. W. (1990). Simple and complex word spans
as measures of working memory capacity. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 16, 1118 –1133. http://
dx.doi.org/10.1037/0278-7393.16.6.1118
Larkin, A. A. (1997). Semantic priming and working memory capacity: A
test of distinctive roles in reading comprehension (Doctoral disserta-
tion). Retrieval from ProQuest Dissertation. (Accession No. 9721406)
Laumann, L. L. (1999). Adult age differences in vocabulary acquisition as
a function of individual differences in working memory and prior knowl-
edge (Doctoral dissertation). Retrieval from ProQuest Dissertation. (Ac-
cession No. 9967200)
Lee, C. S., Huggins, A. C., & Therriault, D. J. (2014). A measure of
creativity or intelligence? Examining internal and external structure
validity evidence of the Remote Associates Test. Psychology of Aesthet-
ics, Creativity, and the Arts, 8, 446 – 460. http://dx.doi.org/10.1037/
a0036773
Lee, S. E. (2014). The impact of working memory training on third-grade
students’ reading fluency and reading comprehension performance
(Doctoral dissertation). Retrieval from ProQuest Dissertation. (Acces-
sion No. 3684696)
Lervåg, A., & Aukrust, V. G. (2010). Vocabulary knowledge is a critical
determinant of the difference in reading comprehension growth between
first and second language learners. Journal of Child Psychology and
Psychiatry, 51, 612– 620. http://dx.doi.org/10.1111/j.1469-7610.2009
.02185.x
Lesaux, N. K., Lipka, O., & Siegel, L. S. (2006). Investigating cognitive
and linguistic abilities that influence the reading comprehension skills of
children from diverse linguistic backgrounds. Reading and Writing, 19,
99 –131. http://dx.doi.org/10.1007/s11145-005-4713-6
Lipsey, M. W., & Wilson, D. B. (2001). The way in which intervention
studies have “personality” and why it is important to meta-analysis.
Evaluation & the Health Professions, 24, 236 –254.
Livesay, K. L. (1998). Multiple constraints and individual differences in
semantic and syntactic processing (Doctoral dissertation). Retrieval
from ProQuest Dissertation. (Accession No. 9913657)
Long, L. L., & Shaw, R. J. (2000). Adult age differences in vocabulary
acquisition. Educational Gerontology, 26, 651– 664. http://dx.doi.org/10
.1080/03601270050200644
Low, P. B., & Siegel, L. S. (2005). A comparison of the cognitive
processes underlying reading comprehension in native English and ESL
speakers. Written Language and Literacy, 8, 131–155.
Ludlum, J. W. (2002). Effects of massed and distributed practice on
verbal working memory performan