Children's Reading Comprehension Ability: Concurrent Prediction by Working Memory, Verbal Ability, and Component Skills

Abstract

The authors report data from a longitudinal study that addresses the relations between working memory capacity and reading comprehension skills in children aged 8, 9, and 11 years. At each time point, the authors assessed children's reading ability, vocabulary and verbal skills, performance on 2 working memory assessments (sentence-span and digit working memory), and component skills of comprehension. At each time point, working memory and component skills of comprehension (inference making, comprehension monitoring, story structure knowledge) predicted unique variance in reading comprehension after word reading ability and vocabulary and verbal ability controls. Further analyses revealed that the relations between reading comprehension and both inference making and comprehension monitoring were not wholly mediated by working memory. Rather, these component skills explained their own unique variance in reading comprehension.

Full text

Children’s Reading Comprehension Ability: Concurrent Prediction by
Working Memory, Verbal Ability, and Component Skills
Kate Cain
University of Essex Jane Oakhill
University of Sussex
Peter Bryant
University of Oxford
The authors report data from a longitudinal study that addresses the relations between working memory
capacity and reading comprehension skills in children aged 8, 9, and 11 years. At each time point, the
authors assessed children’s reading ability, vocabulary and verbal skills, performance on 2 working
memory assessments (sentence-span and digit working memory), and component skills of comprehen-
sion. At each time point, working memory and component skills of comprehension (inference making,
comprehension monitoring, story structure knowledge) predicted unique variance in reading compre-
hension after word reading ability and vocabulary and verbal ability controls. Further analyses revealed
that the relations between reading comprehension and both inference making and comprehension
monitoring were not wholly mediated by working memory. Rather, these component skills explained
their own unique variance in reading comprehension.
Text comprehension is a complex task that draws on many
different cognitive skills and processes. Unfortunately, our knowl-
edge of the unique contribution that these different skills and
processes make to reading comprehension development is limited
because the majority of work in this field has focused on a single
component skill (see Hannon & Daneman, 2001; Saarnio, Oka, &
Paris, 1990, for a discussion of this point). In this article, we are
concerned with how different language skills and processing re-
sources are related to children’s reading comprehension level,
between the ages of 7–10 years.
The language skills that we focus on are higher level language
skills involved in the integration of information across sentences
and ideas in a text, namely, inference and integration, comprehen-
sion monitoring, and knowledge about text structure. These skills
are important for comprehension because they help the reader to
construct an integrated and coherent model of a text’s meaning.
However, young children’s reading comprehension is strongly
predicted by other lower level language skills, such as word
reading accuracy and verbal and semantic skills. Our aim here is to
determine whether the higher level skills make independent con-
tributions to the prediction of reading comprehension over and
above these other language skills, within this age range.
The other focus of this article is the relation among processing
resources (working memory), reading comprehension, and the
higher level language skills. We examine if there is a direct
relation between verbally mediated working memory capacity and
reading comprehension after controlling for verbal and semantic
skills and the extent to which working memory resources underpin
the relations between the higher level language skills and reading
comprehension. First we discuss the relation between working
memory resources and text comprehension and review the evi-
dence for a direct relation between working memory and text
comprehension in children. Then we consider how each higher
level language skill is related to reading comprehension and the
cognitive operations involved in each.
Working Memory and Skilled Text Comprehension
Text comprehension involves the formation of a meaning-based
representation of the text, often called a mental model or a situa-
tion model (Gernsbacher, 1990; Johnson-Laird, 1983; Kintsch,
1998). The processes of integration and inference are important to
the construction of an integrated and coherent model of a text.
Integration between adjacent clauses is necessary to establish local
coherence, and inferences about different events, actions, and
states are required to make the text cohere as a whole (Graesser,
Singer, & Trabasso, 1994; Long & Chong, 2001). These processes
require that the relevant information, either from the text or world
knowledge, is both available and accessible. Working memory
serves as a buffer for the most recently read propositions in a text,
enabling their integration to establish coherence, and holds infor-
mation retrieved from long-term memory to facilitate its integra-
tion with the currently active text (Cooke, Halleran, & O’Brien,
1998; Graesser et al., 1994). Working memory capacity is corre-
lated with college students’ performance on standardized assess-
ments of comprehension skill (Daneman & Carpenter, 1980,
Kate Cain, Department of Psychology, University of Essex, Colchester,
United Kingdom; Jane Oakhill, Department of Psychology, University of
Sussex, Falmer, Brighton, United Kingdom; Peter Bryant, Department of
Experimental Psychology, University of Oxford, Oxford, United Kingdom.
Correspondence concerning this article should be addressed to Kate
Cain, Department of Psychology, University of Essex, Wivenhoe Park,
Colchester CO4 3SQ, United Kingdom. E-mail: kcain@essex.ac.uk
Journal of Educational Psychology Copyright 2004 by the American Psychological Association, Inc.
2004, Vol. 96, No. 1, 31–42 0022-0663/04/$12.00 DOI: 10.1037/0022-0663.96.1.31
31

1983). In addition, working memory capacity is related to skills
important for comprehension, such as the resolution of pronouns,
memory for facts, and the inference of unknown word meanings
from context (Daneman & Carpenter, 1980; Daneman & Green,
1986; Masson & Miller, 1983). Consequently, resource accounts
of comprehension give a central role to processing resources such
as working memory.
Working Memory and Children’s Reading Comprehension
There is a strong relation between working memory (i.e., those
tasks that require the simultaneous storage and processing of
symbolic information) and children’s reading comprehension. In
contrast, comprehension does not correlate with tasks that simply
involve the passive storage of information (Leather & Henry,
1994; Oakhill, Yuill, & Parkin, 1986; Swanson & Berninger, 1995;
Yuill, Oakhill, & Parkin, 1989). The relation between working
memory and comprehension skill has been found with tasks that
require the processing and storage of words (de Beni, Palladino,
Pazzaglia, & Cornoldi, 1998), sentences (Seigneuric, Ehrlich,
Oakhill, & Yuill, 2000), and numbers (Yuill et al., 1989). This
variability in choice of task raises the issue of which type of task
is more appropriate in such investigations. A word- or sentence-
based task is likely to be predictive of reading comprehension
because these tasks are demanding of linguistic skills. A number-
based task, which does not require the processing of words and
sentences, can be used to determine whether there is a more
general relation between working memory and comprehension
skill. Children’s verbal and numerical working memories are both
related to reading comprehension (Oakhill, Cain, & Bryant, 2003;
Seigneuric et al., 2000). However, children’s performance on
working memory tasks that require the manipulation of shapes and
patterns does not explain variance in reading comprehension (Na-
tion, Adams, Bowyer-Crane, & Snowling, 1999; Seigneuric et al.,
2000). Word-, sentence-, and number-based working memory
tasks are readily amenable to verbal coding (unlike spatial tasks),
which might explain their association with comprehension skill.
We refer to working memory tasks that require the processing of
either linguistic or numerical materials collectively as verbally
mediated working memory tasks, to differentiate them from tasks
that require spatial processing.
Nation et al. (1999) suggested that the reported relation between
children’s working memory and their text comprehension is un-
derpinned by verbal and semantic skills. They argued that poor
comprehenders have a specific semantic weakness that restricts
their ability to store verbal information in short-term memory. This
weakness impairs performance on the types of verbally mediated
working memory tasks used in previous research. Similarly,
Stothard and Hulme (1992) proposed that working memory dif-
ferences between good and poor comprehenders would disappear
if differences in verbal IQ (VIQ) were controlled, though they did
not present evidence to support this prediction.
The strong version of this hypothesis is, as yet, unproven.
Working memory capacity assessed by verbally mediated tasks
explains individual differences in children’s reading comprehen-
sion over and above other well-established predictors of compre-
hension, such as decoding, word recognition skill, and vocabulary
knowledge (Swanson & Berninger, 1995; Yuill et al., 1989). Thus,
working memory resources seem to be an important and specific
determinant of children’s reading comprehension level. However,
these working memory tasks may tap into verbal resources (see
above), and both vocabulary knowledge and verbal intelligence are
strong predictors of reading comprehension level (Jensen, 1980;
Oakhill et al., 2003; Sternberg & Powell, 1983). Word knowledge
can affect adults’performance on a measure of verbal working
memory (Dixon, LeFevre, & Twilley, 1988). Clearly, the relation
between children’s verbal skills and their verbally mediated work-
ing memory resources warrants further investigation.
Different assessments of verbal skills and semantic knowledge
include tasks that tap the ability to select a synonym or picture that
matches a particular word in its written or spoken form and tasks
that measure the ability to define words. Investigations into chil-
dren’s reading comprehension and working memory have not
included multiple indicators of verbal ability (but see Dixon et al.,
1988), which is an important omission in the light of the findings
summarized above. In addition, working memory tasks that in-
volve the processing and storage of numbers are weaker predictors
of text comprehension than those with a linguistic content
(Seigneuric et al., 2000), suggesting that the explanatory power of
linguistically based working memory tasks might be considerably
reduced if appropriate verbal controls were included. In this study,
we have included both a sentence-span and a numerical working
memory test as well as assessments of verbal ability to explore the
extent to which the predictive power of the working memory tasks
is mediated by verbal ability.
Component Skills of Reading Comprehension
Many skills may contribute to a child’s reading comprehension
level (Carr, Brown, Vavrus, & Evans, 1990; Palincsar & Brown,
1984; Perfetti, Marron, & Foltz, 1996; Saarnio et al., 1990).
Taxonomies of comprehension abilities often categorize the com-
ponent skills and processes as ones that occur higher or lower in
the language processing chain. For example, word recognition
skills are considered a lower level processing skill. In contrast,
inference making is considered a higher level processing skill
because it aids the construction of the meaning-based representa-
tion of the text (Hannon & Daneman, 2001; Pressley, 2000).
Working memory is a resource that affects an individual’s
ability to carry out many of the processes associated with the
construction of the text representation. Within this resource frame-
work, slow or inaccurate word reading is proposed to affect com-
prehension by using up too much processing capacity with little
remaining for text comprehension processes such as integration
and inference (Curtis, 1980; Hannon & Daneman, 2001; Perfetti,
1985). In accordance with this theory, word reading is the best
predictor of reading comprehension level in the early years (Juel,
Griffith, & Gough, 1986), but other skills become the more im-
portant predictors of comprehension level as word reading ability
develops through experience (Curtis, 1980; Saarnio et al., 1990).
Thus, the relative importance of different skills may change during
the course of development. In this investigation, we controlled for
word reading ability, in addition to verbal skills, in order to
determine whether any of the higher level skills associated with
meaning construction play a unique role in the determination of
comprehension level.
Processing variables and knowledge may each make distinct
contributions to the determination of reading comprehension (Per-
32 CAIN, OAKHILL, AND BRYANT

fetti et al., 1996). Processing failures that could lead to compre-
hension difficulties include inefficient lexical processing, impaired
inference-making skill and comprehension monitoring ability, and
limitations of working memory. Types of knowledge failure that
may lead to comprehension difficulties are impoverished knowl-
edge about word meanings or a specific domain. We consider two
higher level meaning-construction skills that could be classed as
processing variables—inference making and comprehension mon-
itoring—and one that could be classed as a knowledge variable—
knowledge about how narrative texts are structured. Performance
on measures of these skills is related to individual differences in
reading comprehension level, but there have been no studies in-
vestigating the relative contributions of these skills to text com-
prehension within a developmental framework.
Working memory resources are clearly important in the execu-
tion of inference and monitoring skills. A crucial question is
whether individual differences in working memory capacity un-
derlie individual differences in specific comprehension skills that
require the integration of information, such as inference making
and comprehension monitoring (Daneman & Carpenter, 1980,
1983; Seigneuric et al., 2000), or whether deficiencies in these
skills exist in the presence of adequate working memory (and
lexical processing) skills (Perfetti et al., 1996). A brief summary of
each skill and the theoretical basis and empirical evidence for its
relation to reading comprehension and working memory follows.
Inference Making
The construction of a meaning-based representation of a text
involves going beyond the meaning of the text through the gen-
eration of inferences. There are many different classes of infer-
ence, for example, inferences that establish referential coherence,
causal antecedents, and character’s emotional reactions (Graesser
et al., 1994). Inferences may be necessary to establish local co-
herence between adjacent clauses or to establish global coherence
among different events, actions, and states in a text (Graesser et al.,
1994; Long & Chong, 2001). In this study, we were interested in
inferences that were necessary to make sense of a text and that
required either the integration of information among individual
sentences in the text or the integration of general knowledge with
information in the text (see Appendix for examples). The ability to
generate these types of inference is related to both age and reading
comprehension skill (Barnes, Dennis, & Haefele-Kalvaitis, 1996;
Cain & Oakhill, 1999; Cain, Oakhill, Barnes, & Bryant, 2001;
Casteel & Simpson, 1991; Oakhill, 1982, 1984; Omanson, Warren,
& Trabasso, 1978).
Working memory is conceptualized as the work space where
integration and inference take place. Independent assessments of
working memory are related to adults’ability to perform tasks that
rely on these processes, but to date the contribution made by
working memory to the prediction of children’s inference-making
skill (and ultimately their comprehension level) has not been
assessed. However, working memory capacity alone may not be
sufficient to ensure that a crucial inference is generated. The reader
must possess the relevant world knowledge from which an infer-
ence can be drawn (Barnes et al., 1996). Strategic reading behavior
may also play a role in individual differences in inference making.
Adult poor comprehenders make a greater number of knowledge-
based inferences when questions to promote the generation of
crucial inferences are incorporated into the text (Hannon & Dane-
man, 1988). Children’s inference making improves when they are
trained to focus on key words in the text (Yuill & Joscelyne, 1988).
Similarly, children’s reading comprehension benefits when chil-
dren are trained to generate questions to promote the interpretation
of text and to facilitate prediction from text (Palincsar & Brown,
1984).
Comprehension Monitoring
Comprehension monitoring is one aspect of metacognition that
concerns the comprehension of connected prose (Wagoner, 1983).
Measures of comprehension monitoring usually assess a reader’s
ability to detect inconsistencies in text, such as scrambled sen-
tences, contradictory sentences, or statements that conflict with
external information (world knowledge). These error-detection
tasks require readers to evaluate their understanding of the text and
to regulate their reading to resolve any reading problems and to
facilitate their understanding. Hacker (1998) proposed the term
self-regulated reading, rather than comprehension monitoring, to
highlight the importance of both processes.
Performance on error-detection tasks improves with age (Baker,
1984; Markman, 1981), which may be related to children’s devel-
oping information processing capabilities (Ruffman, 1996; Vos-
niadou, Pearson, & Rogers, 1988). Children with reading compre-
hension difficulties are poor at detecting internal inconsistencies in
text (Ehrlich, 1996; Ehrlich, Remond, & Tardieu, 1999). Such
difficulties are more pronounced when the anomalous pieces of
information are nonadjacent (Yuill et al., 1989), indicating that
working memory capacity may influence the application of this
skill. Rubman and Waters (2000) argued that effective compre-
hension monitoring cannot take place unless the reader has ac-
quired the ability to integrate propositions to construct a coherent
representation of a text, which may particularly affect the detection
of internal inconsistencies. However, the detection of internal
inconsistencies can be achieved through the comparison of literal
(explicit) statements in the text, which would not require the reader
to engage in the same sort of constructive processing that is
necessary for inference generation.
The regulation of one’s reading may also involve knowledge of
appropriate repair strategies to deal with comprehension failure
(Wagoner, 1983). If an individual is aware that his comprehension
is inadequate, he can take appropriate steps to remedy the situa-
tion, provided he has the knowledge to do so. Knowledge about
effective repair strategies and the standards used to detect errors in
text improves with age (Baker, 1984; Saarnio et al., 1990). In
addition, task variables such as levels of interest in the task can
affect monitoring performance (de Sousa & Oakhill, 1996).
Understanding Text Structure
Knowledge about the organization of narrative texts increases
throughout middle childhood (Stein & Glenn, 1982). Perfetti
(1994) proposed that a possible source of comprehension failure is
inadequate knowledge about text structures and genres, which may
arise because of insufficient reading experience. Explicit aware-
ness about text structure and the expectations engendered by
certain common features of text may be useful aids for readers,
helping them to invoke relevant background information and sche-
33
CHILDREN’S READING COMPREHENSION ABILITY

mas to facilitate their construction of a meaning-based
representation.
Paris and colleagues (e.g., Myers & Paris, 1978; Paris & Jacobs,
1984) have found that knowledge about the purpose of reading and
knowledge about the information provided by conventional fea-
tures of text are related to both age and reading comprehension.
For example, older readers and better comprehenders were better
able to explain the sorts of information that may be provided by the
introduction and ending of a text. Children with specific compre-
hension difficulties demonstrate impairments in their ability to
structure stories (Cain, 2003; Cain & Oakhill, 1996) and impov-
erished knowledge about the information contained in certain
features of text (Cain, 1996). If knowledge about narrative struc-
ture is well learned and can be activated with little cost to pro-
cessing capacity, it is plausible that efficient retrieval and use of
such knowledge may reduce the adverse effects of limited pro-
cessing capacity (see Graesser et al., 1994, for a discussion of this
point).
Summary of Aims
Investigations of individual differences in children’s reading
ability have identified a wide range of skills that are related to
reading comprehension level. In this article, we focus on the
relative contribution made by the higher level language skills and
processing resources outlined above to children’s text comprehen-
sion during the period in which they move from beginner to
independent readers. In particular, we are concerned with the
predictive ability of working memory, given its prominent role in
resource accounts of comprehension.
The aims are as follows: (a) to investigate whether there is a
direct relation between children’s working memory capacity (as
measured by sentence-span and digit working memory tasks) and
their reading comprehension ability, or whether any association
between these two variables arises through verbal and semantic
skills; (b) to determine whether higher level component skills
(inference, comprehension monitoring, story structure) play a
unique role in the determination of comprehension level; (c) to
explore whether any relations between the higher level language
skills and children’s reading comprehension are mediated by pro-
cessing resources (working memory). In addition, we wanted to
see whether these patterns of prediction change with age.
Method
Participants
The children participating in this study were assessed as part of a
longitudinal project, in which their progress was measured during the years
when they had their 8th, 9th, and 11th birthdays. We report data from all
three time points in this article. The initial sample comprised 102 7–8-
year-olds. These children were selected from an initial assessment of all of
the 7–8-year-old children attending six schools who agreed to participate
in the study. These schools served socially mixed catchment areas on the
south coast of England. Children who were extremely good readers, or very
poor readers, were excluded. The very good readers (those whose word
reading skills were more than 2 years above their chronological age) were
excluded from the study because we expected that their reading ability
would be beyond the scale of the standardized reading assessment given to
our sample (13 years) by the final test point.
Children who did not speak English as their first language, or who had
any known behavioral, emotional, or learning difficulties, were also ex-
cluded. The final selection ensured a sample comprising children with a
range of word reading and reading comprehension ability. The ratio of
children selected from each of the schools reflected the ratios of total
numbers (within the relevant age group) in those schools.
Assessments
The assessments are described briefly below. Further detail can be
obtained from Kate Cain.
Reading ability. As part of the initial screening process, children were
assessed individually on the Neale Analysis of Reading Ability—Revised
British Edition (Neale, 1989). The Neale Analysis provides measures of
both word reading accuracy (word recognition in context) and reading
comprehension (assessed by ability to answer a series of questions about
each passage).
Vocabulary. The children completed two assessments of vocabulary
knowledge. The Gates–MacGinitie Vocabulary subtest, Level 2, Form K
(MacGinitie & MacGinitie, 1989), was administered as a group test during
the initial selection phase. In this test, the child has to select one out of four
words to go with a picture (7–8-year-olds) or to select a synonym of a word
in a short phrase from one of four options (8–9-year-olds, Level 3;
10–11-year-olds, Levels 5 and 6). This test provides an index of a child’s
ability to read and understand written words in unconnected prose. We
assessed receptive vocabulary using an individually administered test, the
British Picture Vocabulary Scale (BPVS; Dunn, Dunn, Whetton, & Pintil-
lie, 1982). In this test, the administrator says a word, and the child has to
point to one of four pictures—the one that pictures the meaning of the
word. Because the children were from a single year group, raw scores from
this test were used in the analyses below.
Verbal ability. We assessed verbal intelligence using two subtests from
the Wechsler Intelligence Scale for Children—Third UK Edition (Wechs-
ler, 1992), Vocabulary and Similarities. These tests provide measures of
children’s knowledge of word meanings and their general knowledge and
reasoning skills. The percentage of the total possible score obtained was
used in the analyses below, to give equal weighting to both subtests.
Working memory. Children completed two assessments of working
memory, one that involved the processing and storage of digits and one that
involved the processing and storage of sentences and words. The digit
working memory task was developed by Yuill et al. (1989) to be analogous
to a reading-span task but without the requirement of sentence compre-
hension. The task includes both a storage and a processing component and
discriminates good and poor comprehenders in a way that digit- and
word-span tasks do not (Yuill et al., 1989). Children read groups of three
digits out loud (the processing component) and are required to remember
the final digit in each group for later recall (the storage component), in the
order of presentation. At the easiest level, children read two groups of
digits and have two final digits to recall, for example, “5–2–8,”“3–7–1,”
recall “8–1.”Thus, the task cannot be performed simply by memorizing
the final digit from each group. Increasing the number of groups of digits
to be read and therefore the number of final digits to be recalled increased
the processing and storage demands of the task. The other working memory
task was a sentence-span task, in which children were read short sentences
that were missing their final word. The sentences were relatively con-
strained, for example, “The color of grass is . . . .”The child’s task was to
complete the sentence and to remember the supplied word for later recall,
in the correct order.
We refer to these tasks as the digit working memory task and the
sentence-span task, respectively. At Time 1, children completed three trials
each of two, three, and four groups of digits or sentences. At Times 2 and
3, they completed three trials each, of three, four, and five groups of digits
or sentences. Two practice items preceded the experimental trials at each
level of difficulty.
34 CAIN, OAKHILL, AND BRYANT

Inference and integration skill. At Time 1, we assessed children’s
inference and integration skills using Oakhill’s (1982) constructive inte-
gration task. In this task, the children listen to eight short (three-line) texts
and are then asked to state whether the given sentences were ones that
occurred in the texts they had been read. An example text is provided in the
Appendix.
For each text, there were four test sentences of three types: two sentences
that had actually been presented (literal information), one sentence that
integrated information from adjacent sentences in a manner that was
consistent with the overall meaning (valid inferences), and one that com-
bined text information in a similar manner, but in a way that was not
compatible with the overall meaning of the text (invalid inferences). Thus,
the child’s performance on the recognition sentences indicates not only
memory for the surface form of the text (e.g., recognition memory for the
original sentences) but also the degree to which the child has gone beyond
the meaning of individual sentences by combining this information to
construct an integrated representation of the text’s meaning. We used the
number of correct acceptances of original sentences and valid inferences
minus the number of acceptances of invalid inferences (to control for
guessing) as an index of the child’s ability to integrate information. The
maximum possible score was 24.
We were concerned that the “false memory”paradigm was subject to
problems of guessing and response bias. Thus, to assess the children’s
inference and integration skills at Times 2 and 3, we used a task developed
by Cain and Oakhill (1999). Children read three short stories and answered
six questions that tapped both literal and inferential information after each
one. Different stories were used at Times 2 and 3. An example text and
questions are provided in the Appendix. For each text, there were two
literal questions to measure memory for facts in the text and four inference
questions. Two of these inference questions assessed children’s ability to
integrate information between two sentences, and two questions assessed
their ability to integrate general knowledge with information in the text to
fill in missing details. Both types of inference questions require a deeper
level of understanding than do the literal questions because the inference
questions demand the integration of information in the text and/or the
supplementation of text information with general knowledge, whereas the
literal questions can be answered by making use of more superficial
information in the text and do not require that the reader go beyond the
propositional content. This task enabled us to look separately at the
contribution of literal and inferential skills to comprehension. The maxi-
mum possible score at Times 2 and 3 was 6 for the literal questions and 12
for the inference questions (combined).
Comprehension monitoring. Children read short stories, some of
which contained conflicting information, to assess their ability to monitor
their own comprehension. At Time 1, the children’s task was to read the
stories out loud and underline any parts that did not make sense. They were
then asked to explain why they had underlined those particular parts. At
Times 2 and 3, the children completed the task silently in small groups. The
distance between the anomalies in the stories was increased to ensure that
the task was of a suitable level of difficulty at each time point. At Times
1 and 2, there were four stories containing one inconsistency (two lines
containing contradictory information) and two fully consistent stories. The
maximum score was 18 at each time point. At Time 3, there were three
stories, each containing two inconsistencies and two fully consistent sto-
ries. The maximum score was 8.
Knowledge of story structure. We used two different measures of
knowledge about story structure, the titles task and the story anagram task.
In the titles task (Cain, 1996), children were required to explain the
information contained in story titles. At Time 2, they also had to explain
the purpose of story beginnings and endings, and at Time 3, they were
additionally asked to explain the information contained in titles relating to
different story genres and to specify the type (i.e., the genre) of the story.
The sentence anagram task has been used to assess the development of
children’s understanding of story structure (Stein & Glenn, 1982). In our
version, children were given three short stories, which had been cut up into
their constituent sentences, and the sentences were randomized. Their task
was to arrange the sentences in the correct order, so that the story made
sense. The stories comprised 6 sentences at Time 1, 8 sentences at Time 2,
and 12 sentences at Time 3. The scores used in the analyses are concor-
dance ratings, to reflect the degree to which the arranged sequence matched
the correct sequence of sentences. The maximum possible score is 1, which
reflects perfect concordance between the reordered sentences and the target
order.
Results and Discussion
Some children did not complete every experimental task at each
time point because of absence or because they moved away from
the area. The data presented here include only those children who
completed every task at a particular time point: Time 1, N⫽100;
Time 2, N⫽92; Time 3, N⫽80. (These sample sizes are adequate
for reliable multiple regression analyses to be conducted, ensuring
more than the minimum 10 data points per predictor variable, as
suggested by Tabachnick & Fidell, 1989.)
Mean Scores Obtained for All Variables at Each Time
Point
The mean scores obtained on each assessment at each time point
are presented in Table 1. The reliability of the different experi-
mental measures was assessed by calculating Chronbach’s alpha
over items. For most assessments, the reliability coefficient was
good or very good (.60–.80). However, the measures of inference
and integration produced rather low alpha levels: The alpha for the
test used at Time 1 was .48 (over 24 items), and at Time 2, it was
.51 (over 12 items).
There was a divergence between word reading ability and text
comprehension across time. At Time 1, the age-equivalent scores
for both word reading accuracy and comprehension were in line
with the mean chronological age. However, at Times 2 and 3, word
reading performance was above chronological age, whereas the
mean reading comprehension score was below chronological age.
Extremely poor word readers were excluded in the initial selection
process. The divergence in word reading and text comprehension
indicated that the remaining children all had the initial skills to
become fluent and able word readers but that their word reading
ability did not automatically enable them to develop good text
comprehension skills.
Interrelations Between Variables at Each Time Point
Tables 2, 3, and 4 show the correlations (one-tailed) between the
standardized assessments and experimental measures taken at each
time point. Because of the large number of correlations, a signif-
icance level of .01 was adopted.
Reading comprehension, working memory, and component
skills. As predicted, at each time point, consistent correlations
were found between reading comprehension and the component
measures of comprehension and also working memory. Although
the two working memory tasks were significantly correlated with
each other, the sentence-span measure was more strongly corre-
lated with the component skills and also with reading comprehen-
sion. Indeed, the digit working memory task was only correlated
with reading comprehension at Time 2. We expected the sentence-
35
CHILDREN’S READING COMPREHENSION ABILITY

span measure to be more highly correlated with measures of
language ability because it involved sentence comprehension. In
previous work, strong relations between the digit working memory
task and reading comprehension level have been found (Seigneuric
et al., 2000; Yuill et al., 1989) so it is surprising that it was not
such a good predictor of text comprehension in this study.
The majority of our comprehension skills were related to each
other throughout the study. The exceptions were the lack of a
significant relation between inference making and monitoring at
Time 1 and also inference making and the story anagram task at
Times 1 and 3. These two tasks were rather different in nature, so
we did not expect them to be highly correlated. The inference task
required the reader to go beyond the literal content of the text,
whereas the comprehension monitoring task required a more su-
perficial type of processing, specifically, the comparison of two
explicit (literal) pieces of information. The story anagram task, by
contrast, relied to a certain extent on knowledge about story
structure. These component skills were more highly correlated
with reading comprehension than with each other.
Although comprehension monitoring was correlated with the
sentence-span working memory task at each time point, the infer-
ence and integration measure was not correlated with either work-
ing memory task at Time 1. The different inference measures used
at Time 1 versus Times 2 and 3 may have made different demands
on working memory, and there were strong correlations with the
sentence-span task at both these later time points. We return to this
point in the General Discussion.
Reading comprehension and word reading ability. The corre-
lation between word reading and text comprehension was greater
at Time 2, when the children were aged 8–9 years, than at Time 3.
The absence of a correlation at Time 1 may be a consequence of
the initial screening procedure: Children with very poor word
reading were not included in the study. Indeed, the standard
deviation of word reading accuracy is lower at Time 1 than at the
Table 1
Means and Standard Deviations for Measures Taken at Time 1, Time 2, and Time 3
Measure
Time 1
(N⫽100) Time 2
(N⫽92) Time 3
(N⫽80)
MSDMSDMSD
Chronological age (in years) 7.53 3.30 8.62 3.15 10.64 3.23
Neale word reading accuracy (in years) 7.81 6.32 9.45 12.45 11.62 14.85
Neale reading comprehension (in years) 7.31 11.24 8.54 15.81 9.26 17.69
Verbal IQ (prorated to 50) 52.13 10.95 67.25 9.81
Gates–MacGinitie 34.36 4.65 32.11 6.52 28.08 7.21
BPVS (standardized scores) 102.90 9.50 105.20 8.80 115.00 13.00
Inference and integration 14.80 3.76 8.39 1.83 6.33 2.09
Literal questions 5.17 0.84 5.40 0.76
Comprehension monitoring 14.58 3.11 14.62 2.76 3.61 1.82
Story anagram 0.80 0.16 0.87 0.08 0.92 0.11
Story titles 2.94 1.16 3.34 1.28 8.11 2.51
Working memory: Sentence span 11.40 3.01 16.77 4.18 21.60 4.98
Working memory: Digit task 10.70 3.06 16.39 5.77 21.61 4.76
Note. The apparent decrease over time in mean scores for the integration and inference and the comprehension
monitoring measures reflect the different maximum scores possible. Neale ⫽Neale Analysis of Reading
Ability—Revised; BPVS ⫽British Picture Vocabulary Scale; Gates–MacGinitie ⫽Gates–MacGinitie Vocab-
ulary subtest.
Table 2
Correlations Between Measures Taken at Time 1
Variable 1 2 34567891011
1. Neale word reading accuracy —.140 .235** .686*** .114 .142 .098 .206 .057 .015 .115
2. Neale reading comprehension —.415*** .222 .440*** .422*** .488*** .402*** .556*** .372*** .057
3. VIQ —.337*** .420*** .259** .271** .521*** .441*** .119 ⫺.056
4. Gates–MacGinitie —.257** .054 .156 .314*** .102 .060 .084
5. BPVS —.203 .406*** .348*** .326*** .251** .000
6. Integration —.165 .220 .240** .066 .011
7. Comprehension monitoring —.355*** .384*** .250** .016
8. Story anagram —.389*** .222 ⫺.008
9. Story titles —.317*** .108
10. Working memory: Sentence span —.356***
11. Working memory: Digit task —
Note. N ⫽100. Data excluded listwise. Neale ⫽Neale Analysis of Reading Ability—Revised; VIQ ⫽verbal IQ; Gates–MacGinitie ⫽Gates–MacGinitie
Vocabulary subtest; BPVS ⫽British Picture Vocabulary Scale.
** p⬍.01. *** p⬍.001.
36 CAIN, OAKHILL, AND BRYANT

other time points, indicating that this variable was more con-
strained at the earliest time point. The weaker correlation at Time
3 may have arisen because the children at this stage in the study
were aged 10–11 years and were becoming more independent
readers. The average word reading age was higher than mean
chronological age, indicating that, for the majority of readers in
this study, word reading was fairly fluent by Time 3 and may have
had less of an influence on text comprehension (e.g., Curtis, 1980;
Saarnio et al., 1990).
Relation between working memory and reading comprehension.
At each time point, reading comprehension was significantly cor-
related with the sentence-span measure of working memory. Sen-
tence span was also significantly correlated with vocabulary, word
reading ability, and/or VIQ at different time points in the study. It
was important to determine whether working memory explained
variance in reading comprehension scores over and above the
contributions made by other verbal skills.
We conducted three fixed-order hierarchical multiple regression
analyses for each time point, with Neale reading comprehension as
the dependent variable. The order of the variables was selected a
priori to test our experimental predictions. To control for the
relation between reading comprehension and lexical–verbal skills,
we entered Gates–MacGinitie Sight Vocabulary scores, BPVS
Receptive Vocabulary scores, and VIQ scores first. Word reading
accuracy scores (Neale Analysis) were also entered at this stage
because word reading ability is a significant determinant of read-
ing comprehension. (We do not report the individual contributions
made by each control variable because this was not an aim of the
article.) VIQ was not assessed at Time 2, so the Time 1 values
were used in the Time 2 analysis. The sentence-span measure of
working memory was more highly correlated with reading com-
prehension than was the digit task. The two working memory
measures were very significantly correlated at each time point, and
more important, both relied on verbally mediated working memory
resources (Nation et al., 1999). Therefore, we entered the sum of
the two working memory scores at the next step to determine
whether working memory explained additional variance in com-
prehension. (Parallel analyses conducted with just the sentence-
span scores produced the same pattern of results.) The results are
summarized in Table 5 (top two rows).
At each time point, the combined working memory measure
explained significant variance in reading comprehension over and
Table 3
Correlations Between Measures Taken at Time 2
Variable 1 2 3 4 5 6 7 8 9 10 11
1. Neale word reading accuracy —.397*** .571*** .364*** .050 .003 .254** .274** .132 .321*** .132
2. Neale reading comprehension —.520*** .602*** .521*** .272** .478*** .344*** .449*** .498*** .344***
3. Gates–MacGinitie —.637*** .322*** .130 .390*** .425*** .328*** .421*** .170
4. BPVS —.419*** .221 .379*** .214 .392*** .492*** .197
5. Inference making —.321** .269** .250** .398*** .417*** .240**
6. Literal questions —.058 .070 .175 .180 .208
7. Comprehension monitoring —.292** .484*** .403*** .193
8. Story anagram —.289** .350*** .206
9. Story titles —.309** .181
10. Working memory: Sentence span —.313***
11. Working memory: Digit task —
Note. N ⫽92. Data excluded listwise. Neale ⫽Neale Analysis of Reading Ability—Revised; Gates–MacGinitie ⫽Gates–MacGinitie Vocabulary subtest;
BPVS ⫽British Picture Vocabulary Scale.
** p⬍.01. *** p⬍.001.
Table 4
Correlations Between Measures Taken at Time 3
Variable 1 2 3456 78 9101112
1. Neale word reading —.194 .246 .459*** .372*** .102 ⫺.141 .412*** .177 .214 .082 .081
2. Neale reading comprehension —.588*** .627*** .511*** .612*** .388** .535*** .299** .470*** .451*** .091
3. VIQ —.707*** .595*** .454*** .184 .405** .209 .418*** .364*** .074
4. Gates–MacGinitie —.674*** .475*** .215 .582*** .180 .413*** .428*** .275**
5. BPVS —.392*** .149 .501*** .155 .268** .248 ⫺.031
6. Inference making —.342** .262** .131 .292** .336** .039
7. Literal questions —.074 ⫺.093 .163 .242 .083
8. Comprehension monitoring —.316** .290** .371*** .191
9. Story anagram —.260** .160 .077
10. Story titles —.360*** .224
11. Working memory: Sentence span —.442***
12. Working memory: Digit task —
Note. N ⫽80. Data excluded listwise. Neale ⫽Neale Analysis of Reading Ability—Revised; VIQ ⫽verbal IQ; Gates–MacGinitie ⫽Gates–MacGinitie
Vocabulary subtest; BPVS ⫽British Picture Vocabulary Scale.
** p⬍.01. *** p⬍.001.
37
CHILDREN’S READING COMPREHENSION ABILITY

above the contribution made by the other variables. Further to the
findings of Seigneuric et al. (2000), there was a specific relation
between reading comprehension and verbally mediated working
memory in 8-, 9-, and 11-year-olds after the contributions made by
word reading and verbal ability had been taken into account.
Relation between reading comprehension and component skills.
Our second aim was to determine whether any of our higher level
meaning construction skills made an independent contribution to
the prediction of reading comprehension ability after controlling
for word reading, vocabulary knowledge, and verbal ability. This
was a particularly strong test for a specific relation between
reading comprehension skill and individual component skills of
comprehension because of the variance that the component skills
shared with verbal skills. Three sets of fixed-order hierarchical
multiple regression analyses were conducted, one for each time
point. These analyses were identical to the first set, except that
performance on a particular component skill, rather than the work-
ing memory score, was entered after the control variables. These
analyses are summarized in Table 5 (rows 3–6).
At each time point, inference-making ability, comprehension
monitoring, and knowledge about story titles explained unique
variance in reading comprehension, after the contribution made by
the control measures had been taken into account. Performance on
the story structure task explained significant variance when chil-
dren were aged 7–8 years and 10–11 years, but not when they
were aged 8–9 years. In general, both knowledge and processing
variables made specific contributions to reading comprehension
for 8–11-year-olds, over and above lower level lexical and verbal
skills.
The role of working memory in relation to reading comprehen-
sion and component skills. Seigneuric et al. (2000) speculated
that individual differences in children’s working memory capacity
might explain individual differences in specific reading compre-
hension skills, such as inference making and comprehension mon-
itoring. In contrast, Perfetti et al. (1996) proposed that readers
might demonstrate impaired inference making or comprehension
monitoring despite possessing adequate working memory re-
sources. To test whether working memory is the skill that under-
pins the relations between comprehension and component skills
and whether any such relation is consistent across this age range,
we conducted a third set of fixed-order hierarchical multiple re-
gressions. Reading comprehension was the dependent variable.
Word reading accuracy, vocabulary, and verbal ability were en-
tered first, followed by the working memory scores. At the final
step, the score of one of the component comprehension tasks was
entered. (This analysis was not performed for the story structure
variable at Times 1 and 2 because this variable did not explain
significant unique variance in reading comprehension.) If working
memory capacity mediates the relations between reading compre-
hension skill and the specific component skills of comprehension,
the processing skills should not explain further variance in reading
comprehension scores after the contribution made by working
memory has been taken into account. Working memory was not
expected to predict performance on the knowledge variables.
The results are summarized in Table 6. Inference-making skill
predicted significant and considerable variance in reading compre-
hension after the contribution of working memory was accounted
for, at each time point. The same relation was found for compre-
hension monitoring. Although these tasks required the integration
and comparison of information where the reader had to process
and store information simultaneously, performance on them was
not wholly determined by working memory capacity. The Time 1
findings were as expected because working memory was not
correlated with the components of comprehension at that point. At
Times 2 and 3, working memory was significantly correlated with
inference making and comprehension monitoring, but these latter
variables made their own unique contribution to the prediction of
reading comprehension. A different pattern was found for the
knowledge variables. This pattern was predicted, because the story
structure measures assessed use of existing knowledge, rather than
processing, which we did not expect to be highly constrained by
working memory capacity.
General Discussion
The data presented here add to our knowledge about working
memory, reading comprehension, and component skills of com-
prehension in several important ways. First, we have established
that working memory capacity explains unique variance in reading
comprehension between the ages of 8–11 years, after the contri-
butions made by word reading skill and verbal ability have been
taken into account. Second, we have demonstrated that specific
component skills of comprehension contribute unique variance to
reading comprehension after word reading and verbal ability con-
trols. Third, we have shown that the relations among inference
making, comprehension monitoring, and reading comprehension
are not wholly explained by variance that the component skills
share with working memory. Finally, we have found a consistent
pattern of prediction of reading comprehension level at each time
point. These findings and their implications for our understanding
of comprehension development are discussed below.
Some researchers have proposed that verbally mediated assess-
ments of working memory are heavily dependent on vocabulary
knowledge and verbal ability (Nation et al., 1999; Stothard &
Hulme, 1992). In this study, the working memory task that re-
quired the manipulation of linguistic material, the sentence-span
task, was more highly correlated with reading comprehension and
the component skills than was the digit task (see also Seigneuric et
al., 2000). This finding is not surprising, given the word and
Table 5
Summary of Fixed-Order Hierarchical Multiple Regression
Analyses (⌬R
2
) With Reading Comprehension as the Dependent
Variable and Working Memory and Component Skills as Criteria,
Controlling for Word Reading, Vocabulary, and Verbal IQ
Step Variable Time 1 Time 2 Time 3
1–4 Neale word reading, Gates–MacGinitie,
BPVS, Verbal IQ .263*** .437*** .459***
5 Working memory .069* .055* .052*
5 Inference and integration .089** .086*** .092***
5 Comprehension monitoring .097*** .048** .046*
5 Story anagram .028, ns .015, ns .031*
5 Story titles .144*** .034* .046*
Note. Neale ⫽Neale Analysis of Reading Ability—Revised; Gates–
MacGinitie ⫽Gates–MacGinitie Vocabulary subtest; BPVS ⫽British
Picture Vocabulary Scale.
*p⬍.05. ** p⬍.01. *** p⬍.001.
38 CAIN, OAKHILL, AND BRYANT

sentence comprehension elements of this task, and supports, in
part, the claims of Nation et al. (1999). However, the significant
correlation between the two working memory assessments at each
time point is consistent with the interpretation that these tasks tap
a common construct. Furthermore, the results of the regression
analyses with their stringent verbal ability controls lead us to reject
the strong version of the verbal ability hypothesis proposed by
Nation et al. and to conclude that verbal skills alone cannot
account for the relation between working memory and text
comprehension.
Skills that affect reading comprehension can be categorized as
occurring higher or lower in the language processing chain. This
system can be useful for identifying the earliest point of difficulty,
if we assume that text processing is bottom-up so that a lower level
impairment will lead to an impairment in the execution of a higher
level skill. Our results are in line with this theoretical position: At
each time point, performance on measures of word recognition and
verbal skills explained a sizable proportion of the variance in
reading comprehension. However, we were interested in whether
lower level processes wholly account for individual differences in
text comprehension or whether higher level processing and knowl-
edge variables could play an independent role. The determination
of comprehension level and the predictive ability of these higher
level component skills of comprehension were not fully explained
by these particularly influential lower level skills (word reading
ability and verbal skills). The measures of processing skills (e.g.,
inference making and comprehension monitoring) and processing
capacity (e.g., working memory) explained unique variance in
reading comprehension ability even after the contribution made by
lower level skills had been taken into account.
The relations between reading comprehension and processing
skills that are important to meaning construction, namely, infer-
ence generation and comprehension monitoring, were not fully
determined by a reader’s working memory capacity. It is important
to identify the types of processes and representations that are
involved in the execution of these tasks and the other factors that
may influence performance on these tasks.
Consider the modest correlations among our measures of com-
prehension monitoring and integration and inference making. Al-
though working memory capacity is important to both tasks, these
measures seem to tap different skills. The monitoring materials
required the reader to identify anomalous pieces of information,
and this comparison is of quite superficial (or literal) information,
which would not require the reader to engage in the sort of
constructive processing involved in the generation of inferences.
Such differences in the operations involved in performing such
tasks may account for the modest correlations.
One factor that may have affected performance on these two
tasks is knowledge about how to generate inferences and monitor
one’s comprehension. Comprehension monitoring can be im-
proved if children are told about the nature of the errors in the text
and if they are instructed in the use of mental imagery when
reading (see Paris, Wasik, & Turner, 1991, for a review). It is also
facilitated when children are required to construct a (pictorial)
storyboard of the text (Rubman & Waters, 2000). Poor compre-
henders make a greater number of inferences (assessed by re-
sponses to questions) when they are instructed to look for clues
when reading in order to establish specific locations and events
that have been left implicit (Yuill & Joscelyne, 1988). It is more
likely that these procedures were successful because they helped
children to read strategically and process text constructively, rather
than developed their working memory skills. Indeed, comprehen-
sion in general can be improved with instruction in how to read
strategically (Paris, Cross, & Lipson, 1984). The implications for
comprehension instruction are favorable: When both component
skills and working memory resources appear to be inadequate,
instructing children in effective ways to process text may help to
circumvent any impairment associated with limited working mem-
ory capacity.
Different patterns of relation between working memory and our
measures of integration and inference were apparent. Ericsson and
Kintsch (1995) proposed that reading span taps into the processes
necessary to store and retrieve the situation model and also world
knowledge from long-term memory. These processes are neces-
sary to perform the inference task used at Times 2 and 3, and that
task was correlated with working memory capacity at both time
points. In contrast, the Time 1 constructive integration measure
was not significantly correlated with concurrent measures of work-
ing memory. In a study of adults’comprehension, Hannon and
Daneman (2001) found a differentiation between integration and
inference-making skills and the processing skills associated with
them. They developed a tool to assess component processes of
comprehension, namely, memory for the text, text inferencing,
knowledge access, and knowledge integration. Different aspects of
this measure accounted for unique variance in different experi-
mental assessments of specific comprehension skills, such as mea-
sures of inference and integration, comparable with those used in
the current study. Thus, others have also found that different types
of inference and integration skills rely on different processing
skills, an issue that clearly warrants further investigation in the
study of comprehension development.
Knowledge about narrative text structure was positively related
to reading comprehension level, even after other measures of
verbal ability and language skill had been taken into account.
Knowledge can help the reader to organize and relate events in a
text, which then benefits her memory and understanding. The
compensatory effects of prior content knowledge on memory for
text are well established. Recht and Leslie (1988) found that
memory for a text about baseball was influenced by readers’prior
Table 6
Summary of Fixed-Order Hierarchical Multiple Regression
Analyses (⌬R
2
) With Reading Comprehension as the Dependent
Variable and Component Skills as Criteria, Controlling for
Word Reading, Vocabulary, Verbal IQ, and Working Memory
Step Variable Time 1 Time 2 Time 3
1–5 Neale word reading, Gates–MacGinitie,
BPVS, Verbal IQ, Working memory .332*** .492*** .511***
6 Inference and integration .087*** .056** .069***
6 Comprehension monitoring .073*** .030* .033*
6 Story anagram .026*
6 Story titles .104*** .025** .035*
Note. Neale ⫽Neale Analysis of Reading Ability—Revised; Gates–
MacGinitie ⫽Gates–MacGinitie Vocabulary subtest; BPVS ⫽British
Picture Vocabulary Scale.
*p⬍.05. ** p⬍.01. *** p⬍.001.
39
CHILDREN’S READING COMPREHENSION ABILITY

knowledge about the sport, not by their reading ability. In the same
way that individuals with good knowledge about baseball will be
better able to invoke specific schemas to help structure the events
in a text about that sport, prior knowledge about the structures and
event patterns common to narratives may facilitate comprehension
in general by aiding the construction of meaning-based
representations.
We found that even from a young age, children’s skills that
foster meaning construction make an important contribution to the
determination of comprehension level over and above the contri-
bution made by word-level and verbal skills. These findings lead
us to conclude that working memory should be regarded as one of
several factors that can influence comprehension ability and com-
prehension development. Clearly, neither good verbal skills nor
good working memory resources are in themselves sufficient for
the application of processes such as inference making and com-
prehension monitoring that are used in the construction of repre-
sentations of text. Instead, further research is needed to establish
which particular aspects of the comprehension-fostering skills
need to be taught and included in future curricula.
References
Baker, L. (1984). Spontaneous versus instructed use of multiple standards
for evaluation comprehension: Effects of age, reading proficiency, and
type of standard. Journal of Experimental Child Psychology, 38, 289–
311.
Barnes, M. A., Dennis, M., & Haefele-Kalvaitis, J. (1996). The effects of
knowledge availability and knowledge accessibility on coherence and
elaborative inferencing in children from six to fifteen years of age.
Journal of Experimental Child Psychology, 61, 216–241.
Cain, K. (1996). Story knowledge and comprehension skill. In C. Cornoldi
& J. V. Oakhill (Eds.), Reading comprehension difficulties: Processes
and remediation (pp. 167–192). Mahwah, NJ: Erlbaum.
Cain, K. (2003). Text comprehension and its relation to coherence and
cohesion in children’s fictional narratives. British Journal of Develop-
mental Psychology, 21, 335⫺351.
Cain, K., & Oakhill, J. (1996). The nature of the relation between com-
prehension skill and the ability to tell a story. British Journal of Devel-
opmental Psychology, 14, 187–201.
Cain, K., & Oakhill, J. V. (1999). Inference making and its relation to
comprehension failure. Reading and Writing, 11, 489–503.
Cain, K., Oakhill, J. V., Barnes, M. A., & Bryant, P. E. (2001). Compre-
hension skill, inference making ability, and the relation to knowledge.
Memory and Cognition, 29, 850–859.
Carr, T. H., Brown, T. L., Vavrus, L. G., & Evans, M. A. (1990). Cognitive
skill maps and cognitive skill profiles: Componential analysis of indi-
vidual differences in children’s reading efficiency. In T. H. Carr & B. A.
Levy (Eds.), Reading and its development: Component skills approaches
(pp. 1–55). London: Academic Press.
Casteel, M. A., & Simpson, G. B. (1991). Textual coherence and the
development of inferential generation skills. Journal of Research in
Reading, 14, 116–129.
Cooke, A. E., Halleran, J. G., & O’Brien, E. J. (1998). What is readily
available during reading? A memory-based view. Discourse Processes,
26, 109–129.
Curtis, M. E. (1980). Development of components of reading skills. Jour-
nal of Educational Psychology, 72, 656–669.
Daneman, M., & Carpenter, P. A. (1980). Individual differences in working
memory and reading. Journal of Verbal Learning and Verbal Behavior,
19, 450–466.
Daneman, M., & Carpenter, P. A. (1983). Individual differences in inte-
grating information between and within sentences. Journal of Experi-
mental Psychology: Learning, Memory, and Cognition, 9, 561–584.
Daneman, M., & Green, I. (1986). Individual differences in comprehending
and producing words in context. Journal of Memory and Language, 25,
1–18.
de Beni, R., Palladino, P., Pazzaglia, F., & Cornoldi, C. (1998). Increases
in intrusion errors and working memory deficit of poor comprehenders.
Quarterly Journal of Experimental Psychology: Human Experimental
Psychology, 51A, 305–320.
de Sousa, I., & Oakhill, J. (1996). Do levels of interest have an effect on
children’s comprehension monitoring performance? British Journal of
Educational Psychology, 66, 471–482.
Dixon, P., LeFevre, J., & Twilley, L. C. (1988). Word knowledge and
working memory as predictors of reading skill. Journal of Educational
Psychology, 80, 465–472.
Dunn, L. M., Dunn, L. M., Whetton, C., & Pintillie, D. (1982). British
Picture Vocabulary Scale. Windsor, United Kingdom: NFER-Nelson.
Ehrlich, M. F. (1996). Metacognitive monitoring in the processing of
anaphoric devices in skilled and less-skilled comprehenders. In C. Cor-
noldi & J. V. Oakhill (Eds.), Reading comprehension difficulties: Pro-
cesses and remediation (pp. 221–249). Mahwah, NJ: Erlbaum.
Ehrlich, M. F., Remond M., & Tardieu, H. (1999). Processing of anaphoric
devices in young skilled and less skilled comprehenders: Differences in
metacognitive monitoring. Reading and Writing, 11, 29–63.
Ericsson, K. A., & Kintsch, W. (1995). Long-term working memory.
Psychological Review, 102, 211–245.
Gernsbacher, M. A. (1990). Language comprehension as structure build-
ing. Hillsdale, NJ: Erlbaum.
Graesser, A. C., Singer, M., & Trabasso, T. (1994). Constructing infer-
ences during narrative text comprehension. Psychological Review, 101,
371–395.
Hacker, D. J. (1998). Self-regulated comprehension during normal reading.
In D. J. Hacker, J. Dunlosky, & A. C. Graesser (Eds.), Metacognition in
educational theory and practice (pp. 165–191). Mahwah, NJ: Erlbaum.
Hannon, B., & Daneman, M. (1988). Facilitating knowledge-based infer-
ences in less-skilled readers. Contemporary Educational Psychology,
23, 149⫺172.
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.
Jensen, A. (1980). Bias in mental testing. New York: Free Press.
Johnson-Laird, P. N. (1983). Mental models. Cambridge, England: Cam-
bridge University Press.
Juel, C., Griffith, P. L., & Gough, P. B. (1986). Acquisition of literacy: A
longitudinal study of children in first and second grade. Journal of
Educational Psychology, 78, 243–255.
Kintsch, W. (1998). Comprehension: A paradigm for cognition. New
York: Cambridge University Press.
Leather, C. V., & Henry, L. A. (1994). Working memory span and
phonological awareness tasks as predictors of early reading ability.
Journal of Experimental Child Psychology, 58, 88–111.
Long, D. L., & Chong, J. L. (2001). Comprehension skill and global
coherence: A paradoxical picture of poor comprehenders’abilities. Jour-
nal of Experimental Psychology: Learning, Memory, and Cognition, 27,
1424–1429.
MacGinitie, W. H., & MacGinitie, R. K. (1989). Gates–MacGinitie Read-
ing Tests. Chicago: Riverside.
Markman, E. M. (1981). Realizing that you don’t understand: Elementary
school children’s awareness of inconsistencies. Child Development, 50,
643–655.
Masson, M. E. J., & Miller, J. A. (1983). Working memory and individual
differences in comprehension and memory of text. Journal of Educa-
tional Psychology, 75, 314–318.
40 CAIN, OAKHILL, AND BRYANT

Myers, M., & Paris, S. G. (1978). Children’s metacognitive knowledge
about reading. Journal of Educational Psychology, 70, 680–690.
Nation, K., Adams, J. W., Bowyer-Crane, C. A., & Snowling, M. J. (1999).
Working memory deficits in poor comprehenders reflect underlying
language impairments. Journal of Experimental Child Psychology, 73,
139–158.
Neale, M. D. (1989). The Neale Analysis of Reading Ability—Revised
British Edition. Windsor, United Kingdom: NFER-Nelson.
Oakhill, J. V. (1982). Constructive processes in skilled and less-skilled
comprehenders. British Journal of Psychology, 73, 13–20.
Oakhill, J. V. (1984). Inferential and memory skills in children’s compre-
hension of stories. British Journal of Educational Psychology, 54, 31–
39.
Oakhill, J. V., Cain, K., & Bryant, P. E. (2003). The dissociation of
single-word reading and text comprehension: Evidence from component
skills. Language and Cognitive Processes, 18, 443⫺468.
Oakhill, J. V., Yuill, N. M., & Parkin, A. (1986). On the nature of the
difference between skilled and less-skilled comprehenders. Journal of
Research in Reading, 9, 80–91.
Omanson, R. C., Warren, W. M., & Trabasso, T. (1978). Goals, inferences,
comprehension and recall of stories by children. Discourse Processes, 1,
337–354.
Palincsar, A. S., & Brown, A. L. (1984). Reciprocal teaching of
comprehension-fostering and comprehension-monitoring activities.
Cognition and Instruction, 1, 117–175.
Paris, S. G., Cross, D. R., & Lipson, M. Y. (1984). Informed strategies for
learning: A program to improve children’s reading awareness and com-
prehension. Journal of Educational Psychology, 76, 1239–1252.
Paris, S. G., & Jacobs, J. E. (1984). The benefits of informed instruction for
children’s reading awareness and comprehension skills. Child Develop-
ment, 55, 2083–2093.
Paris, S. G., Wasik, B. A., & Turner, J. C. (1991). The development of
strategic readers. In R. Barr, M. L. Kamil, P. B. Mosenthal, & P. D.
Pearson (Eds.), Handbook of reading research (Vol. 2, pp. 609–640).
New York: Longman.
Perfetti, C. A. (1985). Reading ability. Oxford, England: Oxford University
Press.
Perfetti, C. A. (1994). Psycholinguistics and reading ability. In M. A.
Gernsbacher (Ed.), Handbook of psycholinguistics (pp. 849–894). San
Diego, CA: Academic Press.
Perfetti, C. A., Marron, M. A., & Foltz, P. W. (1996). Sources of compre-
hension failure: Theoretical perspectives and case studies. In C. Cornoldi
& J. V. Oakhill (Eds.), Reading comprehension difficulties: Processes
and remediation (pp. 137–165). Mahwah, NJ: Erlbaum.
Pressley, M. (2000). What should comprehension instruction be the in-
struction of? In M. L. Kamil, P. B. Mosenthal, P. D. Pearson, & R. Barr
(Eds.), Handbook of reading research (Vol. 3, pp. 545–561). Mahwah,
NJ: Erlbaum.
Recht, D. R., & Leslie, L. (1988). Effect of prior knowledge on good and
poor readers’memory of text. Journal of Educational Psychology, 80,
16–20.
Rubman, C. N., & Waters, H. S. (2000). A, B seeing: The role of
constructive processes in children’s comprehension monitoring. Journal
of Educational Psychology, 92, 503–514.
Ruffman, K. E. (1996). Reassessing children’s comprehension-monitoring
skills. In C. Cornoldi & J. V. Oakhill (Eds.), Reading comprehension
difficulties: Processes and intervention (pp. 33–67). Mahwah, NJ:
Erlbaum.
Saarnio, D. A., Oka, E. R., & Paris, S. G. (1990). Developmental predictors
of children’s reading comprehension. In T. H. Carr & B. A. Levy (Eds.),
Reading and its development: Component skills approaches (pp. 57–79).
New York: Academic Press.
Seigneuric, A., Ehrlich, M.-F., Oakhill, J. V., & Yuill, N. M. (2000).
Working memory resources and children’s reading comprehension.
Reading and Writing, 13, 81–103.
Stein, N. L., & Glenn, C. G. (1982). Children’s concept of time: The
development of a story schema. In W. J. Friedman (Ed.), The develop-
mental psychology of time (pp. 255–282). New York: Academic Press.
Sternberg, R. J., & Powell, J. S. (1983). Comprehending verbal compre-
hension. American Psychologist, 38, 878–893.
Stothard, S. E., & Hulme, C. (1992). Reading comprehension difficulties in
children: The role of language comprehension and working memory
skills. Reading and Writing, 4, 245–256.
Swanson, H. L., & Berninger, V. (1995). The role of working memory in
skilled and less skilled readers’comprehension. Intelligence, 21, 83–
108.
Tabachnick, B. G., & Fidell, L. S. (1989). Using multivariate statistics
(2nd ed.). New York: Harper Collins.
Vosniadou, S., Pearson, P. D., & Rogers, T. (1988). What causes children’s
failures to detect inconsistencies in text? Representation versus compar-
ison difficulties. Journal of Educational Psychology, 80, 27–39.
Wagoner, S. A. (1983). Comprehension monitoring: What it is and what
we know about it. Reading Research Quarterly, 23, 328–346.
Wechsler, D. (1992). Wechsler Intelligence Scale for Children—Third UK
Edition. London: Psychological Corporation.
Yuill, N., & Joscelyne, T. (1988). Effect of organizational cues and
strategies on good and poor comprehenders’story understanding. Jour-
nal of Educational Psychology, 80, 152–158.
Yuill, N. M., Oakhill, J. V., & Parkin, A. J. (1989). Working memory,
comprehension ability and the resolution of text anomaly. British Jour-
nal of Psychology, 80, 351–361.
(Appendix follows)
41
CHILDREN’S READING COMPREHENSION ABILITY

Appendix
Examples of Inference and Integration Measures
Time 1: Text
The mouse ate the food.
The food was bread.
The mouse looked for some cheese.
Time 1: Recognition Sentences
The food was bread. (original)
The mouse looked for some cheese. (original)
The mouse ate the bread. (valid inference)
The food was some cheese. (invalid inference)
Time 3: Extract of Text
Mum was just in the middle of a job when Jenny walked in. “Take off
those wet clothes,”Mum said. “I was just sorting out the blue items
to do first. I can put your jumper in with them now. It will be ready
to wear again by Monday.”Jenny went upstairs to dry and change out
of her wet clothes. But she left a puddle of water in the kitchen by the
fridge where she had been standing. Mum looked for the cleaning
equipment. She found the bucket in the cupboard under the stairs.
Question to test recall of literal information: Where was the puddle of
water?
Question to test integration of information between two sentences:
Where did Mum look for the cleaning equipment?
Question to test integration of general knowledge with information in the
text: What job was Mum doing when Jenny got home?
Note. Time 1 inference and integration materials are from “Constructive
Processes in Skilled and Less-Skilled Comprehenders,”by J. V. Oakhill,
1982, British Journal of Psychology, 73, pp. 19–20. Copyright 1982 by the
British Psychological Society. Reprinted with permission.
Received May 10, 2002
Revision received April 8, 2003
Accepted April 18, 2003 䡲
42 CAIN, OAKHILL, AND BRYANT