Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Running head: READING COMPREHENSION AND WORKING MEMORY
i
Working memory and reading comprehension: The relationship between central executive
functioning and discourse-level reading ability
Jaimée Catherine Taylor
33071318
Thesis Supervised by Dr Bethanie Gouldthorp.
Word Count (From title to end of discussion): 9712
This thesis is presented in partial fulfilment of the requirements for the Bachelor of Arts
(Psychology) Honours degree, Murdoch University, 2017.
Address for correspondence: Dr Bethanie Gouldthorp
School of Psychology and Exercise Science
Murdoch University
South Street
Murdoch, WA, 6015
Email: [email protected]
Telephone: +61 8 9360 2348
Fax: +61 8 9360 6492
Jaimee Taylor
School of Psychology and Exercise Science
Murdoch University
South Street
Murdoch, WA, 6015
Email: [email protected]
Telephone: +61 4 0729 7785
READING COMPREHENSION AND WORKING MEMORY
ii
Declaration
I declare that this thesis is my own account of my research and contains as its main
content work that has not previously been submitted for a degree at any tertiary education
institution.
Jaimée Catherine Taylor
READING COMPREHENSION AND WORKING MEMORY
iii
Table of Contents
Title Page i
Declaration ii
Table of contents iii
List of Figures v
List of Tables vi
Acknowledgement vii
Abstract 1
Introduction 2
Reading comprehension and working memory 4
Individual differences in reading comprehension 5
Aims and Hypotheses 9
Method 10
Participants 10
Materials 10
Reading Integration Test 10
NIH Toolbox Cognition Battery 12
Nelson-Denny Reading Test 13
Apparatus 14
Procedure 15
Reading Integration Test 15
NIH Toolbox Cognition Battery 16
Nelson-Denny Reading Test 18
Results 18
Reaction time 18
Accuracy 20
Speed-accuracy trade-off effect 20
Discourse-level reading comprehension and cognitive processes 21
Propositional-level reading comprehension and cognitive
processes
22
Propositional-level versus discourse-level reading
comprehension
24
Discussion 24
READING COMPREHENSION AND WORKING MEMORY
iv
Facilitation effect 25
Relationship between cognitive processes and discourse-level
comprehension
27
Relationship between cognitive processes and propositional-level
comprehension
31
Discourse-level versus propositional-level reading
comprehension
33
Limitations 35
Implications 36
Future Research 37
Conclusion 39
References 41
Appendices 50
Appendix A Ethics Approval Letter 50
Appendix B Information Letter 51
Appendix C Consent Form 54
Appendix D Detailed Procedure 55
Appendix E Debrief Sheet 57
Appendix F Instructions for NIH Toolbox Cognition Battery
Example
59
Appendix G Example stimuli from Reading Integration Task 61
Appendix H Statistical Output 62
Appendix I Statistical Calculations 72
Appendix J Project summary 73
READING COMPREHENSION AND WORKING MEMORY
v
List of Figures
Figure 1 Baddeley’s 2000 Working Memory Model 5
Figure 2 Example stimuli of the RIT depicting the different levels
of context for target words
16
Figure 3 Partial Correlations for Predictors in a Multiple Regression
Model predicting Discourse-Level and Propositional-
Level Reading Comprehension Ability
24
READING COMPREHENSION AND WORKING MEMORY
vi
List of Tables
Table 1 Example of Target Type (Word or Non-Word) as a
Function of Level of Context
12
Table 2 Mean RT (ms) and Mean Accuracy (%) for Target (Words)
as a Function of Sentence Context
19
Table 3 Unstandarised (B) and standardised errors (SE B) as well
as standardised (β) Regression Coefficients and Part
Correlations (sr²) for Predictors in a Multiple Regression
Model Predicting Discourse-level Reading
Comprehension.
22
Table 4 Unstandarised (B) and standardised errors (SE B) as well
as standardised (β) Regression Coefficients and Part
Correlations (sr²) for Predictors in a Multiple Regression
Model Predicting Propositional-level-Level Reading
Comprehension.
23
READING COMPREHENSION AND WORKING MEMORY
vii
Acknowledgements
During the time of writing I received support from many people. I am indebted to my
supervisor, Dr Bethanie Gouldthorp, who was very generous with her time and
knowledge and provided continuous support, motivation and guidance. I would like to
thank my whole family for all their love, support, kind words and encouragement
throughout the year. Specifically, I would like to say a massive thank-you to my mother,
Lynda, who read an earlier draft of this thesis and to my sister, Robyn, who was always
there to help. I would also like to extend my sincerest gratitude to my dearest friends,
Rachael, Kushca and Damon, for their never-ending support, encouragement and
understanding during this intense time. Lastly, I would like to thank my peers and all
those who have assisted with this study. I could not have asked for a better group of people
to have worked with and I am very appreciative of the time and effort that has gone into
making this study possible.
READING COMPREHENSION AND WORKING MEMORY
50
Abstract
Research has repeatedly demonstrated that a relationship exists between reading
comprehension and working memory (WM). However, there are inconsistencies in the
literature regarding which executive functions (EFs) of WM contribute to overall reading
comprehension ability. This study attempted to address why these inconsistencies exist
by investigating to what extent the ability to manipulate, integrate, and update information
quickly and efficiently was related to adults’ overall reading comprehension. Fifty-one
participants, aged between 18 and 56 years without any cognitive impairments, completed
a standardised reading comprehension assessment and a novel computer-based task which
assessed propositional-level and discourse-level reading comprehension, respectively.
Additionally, participants completed a cognition battery which assessed WM, inhibitory
and attentional control, attention, cognitive flexibility and episodic memory. The findings
indicated that context (i.e., low, medium and high) had a significant effect on discourse-
level reading comprehension ability. This was accompanied by a significant relationship
between WM, its central EFs and adults’ propositional-level reading comprehension.
Unexpectedly, WM and the EFs did not significantly account for individual differences
in discourse-level ability comprehension performance. Moreover, inhibition and attention
negatively predicted participants’ discourse-level reading comprehension ability. Sound
understanding of the relationship between WM and its central EFs has important practical
implications for developing training programs aiming to improve children and adult’s
executive skills and academic performance.
Key words: working memory (WM), executive functions (EFs), situation model
construction, adults, propositional-level reading comprehension, discourse-level reading
comprehension.
READING COMPREHENSION AND WORKING MEMORY
51
Working memory and reading comprehension: The relationship between central executive
functioning and discourse-level reading ability
Understanding underlying factors which influence individual differences in
reading comprehension has important implications for society as it is associated with a
broad range of psychosocial, cognitive, economic and health benefits (Schwanenflugel &
Knapp, 2016). A well established predictor of individual differences in reading
comprehension is working memory (WM; Cain, Oakhill, & Bryant, 2004; Kiss, Pisio,
Francois, & Schopflocher, 1998; Nouwens, Groen, & Verhoeven, 2017; Oakhill, Cain, &
Bryant, 2003). Comprehension skills rely on the processing and storage of words and
their meanings, which demand various cognitive resources (e.g., WM; Cain et al., 2004;
Oakhill et al., 2003). Yet, the executive functions (EFs) central to WM, which underlie
individual differences in discourse-level reading comprehension, is less clearly
understood.
Current theories of discourse-level comprehension assert that comprehension
skills rely on the construction of a multi-level representation of a text (Fincher-Kiefer,
1993). These representations consist of two lower levels (i.e., surface and propositional)
and a higher discourse-level (Fincher-Kiefer, 1993; Zwaan & Radvansky, 1998; Zwaan,
Stanfield & Yaxley, 2002). Consider the following sentences: (i) The ranger saw the
eagle in the sky; and (ii) The ranger saw the eagle in the nest (Zwaan et al., 2002). At
surface level, individual words are decoded based on the grammatical information while,
at propositonal level, the two sentences are nearly identical with the eagle’s locality being
the only difference (i.e., sky or nest; Zwaan et al., 2002). This propositional processing
draws on the textual information to construct a fairly basic, amodal meaning (Zwaan et
al., 2002). The discourse-level builds on these two lower levels, with the first sentence
activating a visual experience of an eagle with stretched-out wings and the second
READING COMPREHENSION AND WORKING MEMORY
52
activating a visual experience in which the eagle has its wings folded-in (Zwaan et al.,
2002).
This discourse-level processing refers to higher-level multi-modal
representations, or ‘situation models’, of a situation described in a text (i.e., characters,
places and actions; Davoudi & Moghadam, 2015; Zwaan et al., 2002). Cook and Myers
(2004) suggest that discourse-level processing may influence reading comprehension by
facilitating the integration of the target concept with the preceding text. This often relies
on knowledge-based inferences (i.e., the activation of additional information from
background knowledge during the reading process; Ababneh & Ramadan, 2013; Davoudi
& Moghadam, 2015; Yeari & van den Broek, 2015; Zwaan et al., 2002). Knowledge-
based inferences provide an explanation why readers can create the visual experience of
an eagle with out-stretched or folded wings, although this information is not described in
the text (Yeari & van den Broek, 2015; Zwaan et al., 2002).
There is a lack of consensus in the literature regarding how to measure situation
models. While different components utilised in the construction of a situation model (e.g.,
inference generation, metaphor comprehension and integration) have been explored by
various authors (e.g., Beeman et al., 1994; Cook & Myers, 2004; Federmeier, 2007;
Zwaan & Radvansky, 1998; Zwaan et al., 2002), not many have attempted to directly
measure overall situation model construction. Thus, experimental tasks theoretically
employing discourse-level comprehension processing, such as the integration of
information across sentences, are relied upon to measure individual differences in
situation model construction (Gouldthorp & Coney, 2011). For example, Gouldthorp and
Coney (2011) suggest that facilitation of target words in their Reading Integration Task
(RIT) is greater when readers are provided with three context sentences (i.e., “I am
sweet”, “I am spread on toast” and “I am made by insects”) than only one or two context
READING COMPREHENSION AND WORKING MEMORY
53
sentences. Thus, this measure can be used to examine individual differences in readers’
abilities to construct a coherent situation model by integrating information from world-
knowledge (i.e., reader’s prior knowledge) across context sentences (Gouldthorp &
Coney, 2011).
Individual differences in the ability to construct meaning across sentences (termed
sequencing ability) has been examined in studies such as Gouldthorp, Katsipis and
Mueller (2017). Specifically, Gouldthorp et al found that sequencing ability facilitated
children’s discourse-level comprehension skills. This supports Zwaan et al’s (2002) claim
that discourse-level reading ability is fundamental for gaining a comprehensive
understanding of a text. Moreover, the construction of situation models is associated with
WM’s central executive component, relying on the ability to hold incoming information
in memory while retrieving existing knowledge from long-term memory (LTM; Davoudi
& Moghadam, 2015; García-Madruga, Vila, Gόmez-Veiga, Duque & Elosúa, 2014).
Reading comprehension and working memory
WM is widely associated with comprehension performance as it determines an
individual’s ability to hold incoming information and knowledge from LTM (García-
Madruga et al., 2014) and to integrate these into a coherent situation model (Cantin,
Gnaedinger, Gallaway, Hesson-McInnis & Hund, 2016; Davoudi & Moghadam, 2015;
Gouldthorp et al., 2017). WM is defined as a multi-component system which briefly
retains information while simultaneously carrying out other mental processing operations,
after which information is encoded into LTM or is forgotten (Baddeley, 2012; Davoudi
& Moghadam, 2015; Engel de Abreu et al., 2014; García-Madruga et al., 2014; Nouwens,
Groen, & Verhoeven, 2016). Different WM models are depicted in literature (e.g.,
Baddeley’s Multiple Component Model, the Object-Oriented Episodic Record and the
Embedded-Process Model; Chein & Fiez, 2010). Although these (and other) models
READING COMPREHENSION AND WORKING MEMORY
54
conceptualise WM differently, they all posit a central EF or a similar attentional control
mechanism (Chein & Fiez, 2010; McCabe, Roediger, McDaniel, Balota, & Hambrick,
2010), of which Baddeley’s 2000 WM model is most clearly articulated (Baddeley, 2012;
Chein & Fiez, 2010; Figure 1).
The executive component is described as a flexible, limited storage capacity,
attentional control system (Iglesias-Sarmiento, Carriedo-López, & Rodríguez-Rodríguez,
2015). It is particularly important in explaining individual differences in complex
cognitive tasks (Iglesias-Sarmiento et al., 2015). These often involve planning and control
(e.g., comprehension, decision-making and problem solving; Iglesias-Sarmiento et al.,
2015).
Individual differences in reading comprehension
Literature suggests that WM capacity (i.e., the ability to maintain task-relevant
information in a highly active state; Meinz & Hambrick, 2010) explains individual
performance in reading comprehension (Daneman & Carpenter, 1980; Dutke & von
Hecker, 2011; Turner & Engle, 1989). This is evident as individuals with greater WM
capacity (i.e., information storage) performed better on comprehension tasks than those
with lower WM capacity (Dutke & von Hecker, 2011; Turner & Engle, 1989).
Figure 1. Baddeley's 2000 Working Memory Model (Baddeley, 2012, p. 16).
READING COMPREHENSION AND WORKING MEMORY
55
Consequently, Daneman and Carpenter (1980) suggests that skilled readers rely on fewer
cognitive processes in comparison to less skilled readers when completing the same task.
For example, skilled comprehenders could skip intermediate steps in some or all stages
of decoding, inferencing and integrating incoming information (Daneman and Carpenter,
1980).
Emerging research suggests that EFs are crucial in becoming skilled in reading,
problem solving and controlling goal-directed behaviour, particularly during early
childhood (Nouwens et al., 2016). EFs refer to the ability to plan, organise and monitor
deliberate, goal-orientated behaviours (National Institutes of Health & Northwestern
University (NIH & NU), 2016a). This enable readers to manipulate thoughts and
behaviours, monitor novel information, inhibit behavioural responses and maintain focus
on completing challenging tasks (Cantin et al., 2016, Diamond, 2013; Engle de Abreu et
al., 2014). Types of EFs include inhibition, cognitive flexibility, and attention.
Inhibition potentially influences reading comprehension by allowing readers to
suppress automatic reactions and ignore irrelevant information (Cain, 2006; Cantin et al.,
2016; Diamond, 2013; Engel de Abreu et al., 2014; Nouwens et al., 2016). Cognitive
flexibility allows readers to adapt WM representations of events according to the situation
presented in the text (Diamond, 2013; Engel de Abreu et al., 2014). Moreover, it enables
readers to coordinate reading strategies and processes to achieve success in reading
(Gnaedinger, Hund & Hesson-McInnis, 2016). Lastly, attention aids readers in directing
attentional resources towards task-relevant information to achieve goal-orientated tasks,
thereby facilitating reading ability (García-Madruga et al., 2014; Yogev‐Seligmann,
Hausdorff, & Giladi, 2008).
An alternative cognitive process which differs from WM and EFs is episodic
memory, which has been considered necessary for successful reading comprehension
READING COMPREHENSION AND WORKING MEMORY
56
performance (Mumper, 2013). Episodic memory relates to a reader’s ability to reactivate
information stored in LTM (Mumper, 2013). Information stored in LTM includes
contextual information that is either essential for preserving coherence or related to the
information currently active in memory (Mumper, 2013). Essentially, these assist in
developing knowledge-based inferences described by Ababneh and Ramadan (2013),
Yeari and van den Broek (2015) and Zwaan et al (2002).
While the literature strongly supports that WM accounts for individual differences
in reading comprehension ability, most of this research has focused on surface and
propositional-level comprehension processes (Zwaan et al., 2002). Moreover, other
researchers claim that WM capacity does not distinguish between good and poor reading
comprehenders (Daneman & Carpenter, 1980). This may be due to the digit span and
word span measures (to assess WM capacity) not effectively tapping into the WM
processing component (Daneman & Carpenter, 1980). This is further demonstrated by
inconsistency in the research findings concerning EFs of WM and discourse-level reading
ability (Follmer, 2017).
Explicitly, some research has demonstrated that EFs are associated with
differences in discourse-level reading comprehension ability (Follmer, 2017), while
others (e.g., Georgiou & Das, 2016) indicate that WM is not heavily involved in situation
model construction. This difference is possibly explained by the opposing perspectives
of various authors (Follmer, 2017). For example, some authors concentrate on modality-
specific WM stores (i.e., the phonological loop), whilst others focus the central executive,
or attentional control system (Follmer, 2017). It has been suggested that the central
executive component is of more importance than the modality-specific WM stores as it is
more directly involved in the processing and manipulation of incoming textual
information (Gouldthorp et al., 2017).
READING COMPREHENSION AND WORKING MEMORY
57
Inconsistency in the literature exists in terms of determining which EFs have been
directly implicated with individual differences in reading comprehension (Follmer,
2017). That is, cognitive switching (and to a lesser extent, attention) has been shown to
successfully predict reading comprehension performance (Arrington, Kulesz, Francis,
Fletcher, & Barnes, 2014; Cantin et al., 2016; Engel de Abreu et al., 2014; García-
Madruga, Gómez-Veiga, & Vila, 2016; Gerst, Cirino, Fletcher, & Yoshida, 2017;
Nouwens et al., 2016). However, Nouwens and colleagues (2016) found that planning did
not predict performance on a comprehension task. This discrepancy in findings may be
due to sampling differences as some studies assessed older children (Georgiou & Das,
2016) and adults (Chrysochoou, Bablekou & Tsigilis, 2011), while others assessed
younger children (Nouwens et al., 2017; García-Madruga et al., 2016). Likewise, some
measures of EFs may not be scientifically valid (i.e., fail to measure what they purport to
measure), as conveyed by a lack of psychometric information of some EF tasks
(Arrington et al., 2014; Christopher et al., 2012).
Additionally, these discrepancies may be attributed to either (or both) the
underlying model of WM and the vastly different WM measures used by different authors
(Follmer, 2017). For example, studies conducted by Arrington et al (2014) and Cantin et
al (2016) suggest that participants with high reading comprehension performance achieve
higher scores on different WM measures. This is inconsistent with Georgiou and Das
(2016). They claimed that WM does not affect reading ability as different measures of
WM may not accurately assess WM in general (i.e., the listening span and digit span
backward tasks). In addition, they suggest that the nature of the reading task may itself
influence the outcome of the study (Georgiou & Das, 2016).
Nevertheless, WM and its central EFs are considered crucial in underpinning a
wide range of higher cognitive abilities, including reasoning, learning and comprehension
READING COMPREHENSION AND WORKING MEMORY
58
(Arrington et al., 2014; Cantin et al., 2016; Chein & Fiez, 2010; Collette & Van der
Linden, 2002; Engel de Abreu et al., 2014; García-Madruga et al., 2014; 2016; Gerst et
al., 2017; Kiss et al., 1998; Nouwens et al., 2016; Oakhill et al., 2003). Unfortunately, the
existing literature has not systematically evaluated the different EF processes to establish
their contributions to discourse-level reading comprehension ability. Thus, this thesis
addresses the underlying theory of EFs and considers how these contribute to reading
comprehension ability at propositional and discourse levels of processing.
Aims and Hypotheses
The purpose of this study was to elucidate why there are inconsistencies in the
literature regarding the role of WM in reading comprehension ability. Specifically, the
aim of the study was to examine how central EFs in WM (i.e., the ability to manipulate,
integrate and update information quickly and efficiently regardless of modality) may
account for individual variation in the ability to produce higher-level (discourse) meaning
during reading. Thus, the following research question was posed:
Are individual differences in reading comprehension and situation model
construction partly explained by differences in functioning of specific EFs central
to WM?
To investigate this question, a within-group study was conducted. Adult
participants completed three tasks: the Nelson-Denny Comprehension Test (a
standardised measure of surface and propositional-level reading comprehension ability;
Brown, Fishco, & Hanna, 1993); the NIH Cognitive battery, consisting of a series of
cognitive tasks (Gershon et al., 2013); and an adapted Reading Integration Task (RIT)
which assesses discourse-level reading comprehension (as indicated by the ability to
integrate information from world-knowledge across multiple sentences to produce
implicit meaning; Gouldthorp & Coney, 2011).
READING COMPREHENSION AND WORKING MEMORY
59
Three hypotheses guided this study. First, it was hypothesised that the
presentation of three complete sentences (i.e., high-context condition) would produce
greater facilitation for target words which will indicate stronger, discourse-level
comprehension skills. Second, it was hypothesised that performance on the EF subtests
would account for a significant proportion of variance on the outcome variable of
discourse-level reading comprehension ability (i.e., high-context facilitation effect from
the RIT). Third, it was hypothesised that the EF cognitive flexibility would contribute a
significant proportion of variance on both propositional and discourse levels of reading
comprehension ability in comparison to inhibition and attention, WM capacity and
episodic memory.
Method
Participants
Fifty-one participants (13 males and 38 females) aged between 18 and 56 years
(M = 26.94, SD = 9.77), who have been (or are currently) enrolled in an English-speaking
tertiary education institution, were recruited for this study. All participants self-reported
to predominantly speak English, have normal or corrected-normal vision and hearing and
lacked cognitive impairment. This study was approved by Murdoch University’s Human
Research Ethics Committee (approval number 2017/047; Appendix A).
Materials
RIT (adapted from Gouldthorp & Coney, 2011). This measure assessed
individual differences in relation to the ability to integrate provided information across
sentences with world-knowledge to produce discourse-level meaning (Gouldthorp &
Coney, 2011). The task required participants to make lexical decisions regarding
centrally-presented targets, primed by centrally-presented sentences (consisting of three
to eight words) which establish context (Gouldthorp & Coney, 2011). Previous work
READING COMPREHENSION AND WORKING MEMORY
60
using these stimuli established that the presentation of three complete sentences (i.e.,
high-context condition) generated greater facilitation for targets, which reflected
discourse-level comprehension (Gouldthorp & Coney, 2011). Target words were short
(3 to 5 letters), highly concrete and highly imageable nouns as determined by the MRC
Psycholinguistic database (see Gouldthorp & Coney, 2011). Moreover, the stimuli have
been evaluated to discount alternative explanations of these facilitation effects (e.g.,
summation priming from processing individual words rather than the whole sentence;
Gouldthorp & Coney, 2011).
The RIT employed a 4x2 repeated-measures design, with the independent
variables (IV) of level of context (i.e., neutral = neutral sentence only, low = one context
sentence, medium = two context sentences and high = three context sentences) and target
type (word or non-word; Gouldthorp & Coney, 2011; refer to Table 1). The dependent
variables (DV) were: reaction time (RT) to stimuli (i.e., whether target was a word or a
non-word); and accuracy of responses (Gouldthorp & Coney, 2011). To facilitate a
concise measure of discourse-level comprehension ability for this study, the RIT was
adapted from Gouldthorp and Coney (2011) to 80 trials (with 20 trials per condition).
These trials were without the repetition of stimuli and with an equal distribution of target
words and non-words for each condition. Additional example stimuli are presented in
Appendix G. Administration of this task took approximately 20 minutes, with a rest-break
opportunity after every tenth trial.
READING COMPREHENSION AND WORKING MEMORY
61
Table 1
Example of Target Type (Word or Non-Word) as a Function of Level of Context.
NIH Toolbox Cognition Battery (Gershon et al., 2013). This measure was
designed for use in epidemiological studies and clinical trials for ages 3 to 85 years and
include the following domains: cognition, emotion, motor and sensation (Gershon et al.,
2013). Heaton and colleagues (2014) assessed the NIH Cognition Battery’s reliability and
validity in an adult sample. The cognition battery was considered reliable and valid, with
high test-retest reliability (r = .86 - .92) and strong convergent (r = .78 - .90) and
discriminant (r = .19 - .39) validities (Heaton et al., 2014; Weintraub et al., 2014). The
measure consisted of four instruments which assessed WM capacity, episodic memory
and three central EFs (i.e., inhibition, attention and cognitive flexibility; Gershon et al.,
Target Context First Sentence Second Sentence Third Sentence
Word
HONEY Neutral “This is a neutral
sentence.”
Low “I taste sweet.”
Medium “I can be put on
toast.”
High
“I am made by an
insect.”
Non-word
KIVES Neutral “This is a neutral
sentence.”
Low “I am placed on
tables.”
Medium “I am used to eat.”
High “I can cut.”
READING COMPREHENSION AND WORKING MEMORY
62
2013). Each subtask had automated instructions which were presented on an iPad screen.
This measure required approximately 30 minutes to complete, with each subtask taking
between 90 seconds to 7 minutes to complete (Gershon et al., 2013).
The Flanker Inhibitory Control and Attention Test measured inhibitory and
attentional control (Gershon et al., 2013). Participants were asked to complete four
practice trials, accompanied by 20 test-trials (Gershon et al., 2013). The List Sorting WM
Test assessed WM capacity and consisted of two conditions: 1-List condition which
contained a series of object-pictures (i.e., food OR animal) which were to be size-ordered
from smallest to biggest; and 2-List condition which contained a series of object-pictures
(i.e., food AND animal) which were to be categorically and size-ordered (i.e., food first
then animal), from smallest to biggest (Gershon et al., 2013). A maximum of seven
picture-objects (i.e., food and/or animal objects) was presented to participants (Gershon
et al., 2013).
The Dimensional Change Card Sort Test measured cognitive flexibility using a
mixed block of 30 items (Gershon et al., 2013). Two target pictures which varied along
two dimensions (shape and colour) were presented to participants, from which they were
required to made selections according to the primed dimension (Gershon et al., 2013).
The Picture Sequence Memory Test (Form A) assessed episodic memory, specifically the
acquisition, storage and retrieval of new information (Gershon et al., 2013). There was
one practice sequence containing four items which had to be ordered according to the
sequence presented, followed by two test sequences; the first contained 15 items while
the second had 18 items (Gershon et al., 2013).
Nelson-Denny Reading Test (Form G; Brown et al., 1993). The purpose of this
standardised measure was to provide an assessment of participants’ reading ability in two
areas of academic achievement: reading rate and reading comprehension (at propositional
READING COMPREHENSION AND WORKING MEMORY
63
level). It consisted of seven passages, with a total of 38 questions (Brown et al., 1993).
This measure required approximately 20 minutes to compete, in which the first minute of
the test measured reading rate (i.e., words read per minute). Regarding the Reading Rate
subtest, the Nelson-Denny Test manual reports a moderate reliability coefficient (r) of
.69. Lewandowski, Codding, Kleinmann and Tucker (2003) report an adequate reliability
coefficient (r) of .81 for alternative forms (i.e., Forms G and H). The Nelson-Denny
Reading Test was used in this study because it was considered a common measure of
adults’ reading performance (Lewandowski et al., 2003).
Apparatus
The RIT (adapted from Gouldthorp & Coney, 2011) was programmed for stimulus
presentation and data collection using DirectRT (Empirisoft Corporation, 2011), which
used Windows 7 operating system. A DirectIN High Speed Button-Box v2012 was used
to record participants’ responses. The 9-button response box was at a fixed position
(50cm) from the computer monitor. Laminated rectangular labels were taped under the
left and right-most buttons which specified which button corresponded to ‘word’ or
‘nonword’ (which was counter-balanced across participants).
The Nelson-Denny Reading Test required: the Part II Comprehension Test
booklet, which contained the seven text passages and series of multiple choice questions
(Brown et al., 1993); a pen (black or blue ink) supplied by the investigator; and a timer
(i.e., iPhone 7) which was set to 20 minutes.
Administration of the NIH Cognition Battery required: a baseline measure
instrument which assessed participants’ RTs during the Flanker Inhibitory Control and
Attention Test and the Dimensional Change Card Sort Test; an examiner scoring sheet
which guided the scoring of participants’ responses during the List Sorting WM Task; a
wireless key-pad which the investigator used to record participants’ verbal responses
READING COMPREHENSION AND WORKING MEMORY
64
during the List Sorting WM Task; and an iPad with the NIH Toolbox App pre-installed
(Gershon et al., 2013).
Procedure
This thesis project was approved by the Murdoch University’s Human Research
Ethics Committee. Participants attended one data collection session in a venue at a
university in Perth (Western Australia), which lasted approximately 120 minutes. The
session was divided into two data collection periods (approximately 50 minutes each),
separated by a 20-minute break. Participants read the information sheet (Appendix B) and
voluntarily signed the consent form (Appendix C) at the beginning of the session.
Participants then completed the three tasks, with the order of the tasks rotated across
participants to minimise possible fatigue and order effects (Shaughnessy, Zechmeister, &
Zechmeister, 2006). Refer to Appendix D for a more comprehensive account of the
procedure. Upon completion of all three measures, participants were provided with a
summary of the purpose of the study (Appendix E).
RIT (adapted from Gouldthorp & Coney, 2011). Participants were required to
read each presented sentence at their own pace and progress to the next sentence or the
target word by pressing any key. One, two or three sentences were centrally presented on
the computer screen, followed by a centrally-presented fixation cross and then a target
word or non-word (Figure 2).
READING COMPREHENSION AND WORKING MEMORY
65
NIH Toolbox Cognition Battery (Gershon et al., 2013). To become proficient
in the administration of the NIH Toolbox Cognition Battery, the investigator underwent
extensive training. This included accessing online-training resources, thoroughly working
through the NIH Toolbox manual (NIH & NU, 2016a; Slotkin et al., 2012) and practising
each instrument several times to ensure competency of administration. During the
administration process, the investigator read the instructions presented on the iPad screen
to the participants (refer to Appendix F for example of one test item). Administration of
each instrument remained consistent across participants.
During the Flanker Inhibitory Control and Attention Task, a row of five arrows
arrayed in different directions were displayed on the iPad screen accompanied by two
response options; an arrow pointing left or an arrow pointing towards the right.
Participants were required to select the arrow that matched the direction the centre arrow
Figure 2. Example stimuli of the RIT depicting the different levels of context
for target words (adapted from Gouldthorp & Coney, 2011).
READING COMPREHENSION AND WORKING MEMORY
66
was pointing (Gershon et al., 2013). A baseline measure (refer to Apparatus) was used,
thus participants were instructed to respond as quickly and as accurately as possible.
Participants then completed the List Sorting WM Test, where they were exposed
to a series of verbal and visual stimuli (i.e., animal or food; Gershon et al., 2013).
Participants were asked to verbally recall the objects in size order from smallest to biggest
once the iPad screen went blank. For example, when presented animal objects (e.g., horse
and dog), the correct verbal response was ‘dog, horse’ (NIH & NU, 2016a). Consecutive
trials increased in difficulty; the first trial consisted of two stimuli, while the last possible
trial consisted of seven. The exposure time to stimuli varied as trials increased in
complexity. The task concluded once participants responded incorrectly during two
consecutive test-trials. The second section of this activity employed the same process,
however, participants were instructed to recall food objects first, from smallest to biggest,
then animal objects, again from smallest to biggest (e.g., popcorn then dog). The examiner
scoring sheet and the wireless keyboard (refer to Apparatus) was used to record
participants’ responses (i.e., the investigator pressed ‘1’ for correct response and ‘0’ for
incorrect response on the wireless keyboard; NIH & NU, 2016a).
During the Dimensional Card Change Sorting Test, participants were instructed
to match the target stimuli according to two cognitive dimensions (i.e., colour or shape)
as quickly as possible (Gershon et al., 2013). A series of ‘switch’ trials were presented
where participants were required to match a series of bivalent test pictures (i.e., yellow
ball and blue truck) to the target picture according to either colour (e.g., yellow or blue)
or shape (e.g., truck or ball; NIH & NU, 2016a). The baseline measure was used to assess
participants’ RTs (refer to Apparatus). Lastly, the Picture Sequence Memory Test was
administered where participants were presented with a sequence of images associated
with playing in a park, accompanied by a narration of the sequence (Gershon et al., 2013).
READING COMPREHENSION AND WORKING MEMORY
67
They were then required to order the images in the same sequence originally presented
(NIH & NU, 2016a).
Nelson-Denny Reading Test (Form G; Brown et al., 1993). This test was
administered in accordance with the guidelines specified in the Comprehension Test
manual (Brown et al., 1993). Participants were instructed to read the seven passages at
their own pace and then answer the corresponding multiple-choice questions provided at
the end of each passage. Reading rate was recorded 60 seconds into the test, where
participants were required to circle the number next to the respective sentence they were
reading at that time, upon the investigator’s instruction. According to test procedure,
participants were required to complete the test within the 20-minute time limit.
Results
Reaction time
The RIT data was initially screened using an outlier deletion criterion of ±2.00
standard deviations from participants’ mean response times for each condition. This
resulted in 5.80% of the total observations being excluded. A one-way repeated-measures
analysis of variance (ANOVA) was initially performed on the IV of context (e.g., neutral,
low, medium and high; summarised in Table 2). The results indicated the assumption of
sphericity was violated for the main effect of context. The Huynh-Feldt correction was
used in this instance. The ANOVA revealed a significant effect of context, F(2.53,
126.24) = 10.30, MSe = 19868.27, p < .001. Participants’ RTs decreased as context
increased from neutral (i.e., slowest RTs) to low to medium to high (i.e., quickest RTs).
READING COMPREHENSION AND WORKING MEMORY
68
Table 2
Mean RT (ms) and Mean Accuracy (%) for Target (Words) as a Function of Sentence Context.
Level of context Mean RT (SD) Mean Accuracy (SD)
Neutral 818.44 (323.18) 91.57 (7.58)
Low 810.23 (311.82) 91.57 (6.74)
Medium 714.48 (264.93) 93.14 (7.07)
High 713.24 (248.85) 92.35 (6.51)
Note. N = 51.
Additionally, a one-way repeated measures ANOVA was conducted on the IV of
context facilitation, which was computed by subtracting RTs of the low, medium and
high-context conditions from the neutral condition. These results also indicated the
assumption of sphericity was violated from the main effect of context and the Huynh-
Feldt correction was thus applied. The main effect of context was significant, F(1.77,
88.53) = 12.69, MSe = 14052.24, p < .001, reflecting that increasing context from low (M
= 8.22, SD = 233.80 ms) to medium (M = 103.96, SD = 188.45 ms) to high (M = 105.20,
SD = 188.31 ms) resulted in a large increase in facilitation.
A paired samples t-test with an α of .05 was used to compare mean facilitation
effects of the medium and high-context conditions. On average, there was little difference
in facilitation effects between the high and medium-context conditions (D = 1.23 ms, 95%
CI [-33.29, 35.76]), which was not statistically significant, t(50) = .072, p > .05, and a
small, d = .01. Furthermore, post-hoc pairwise comparisons (using Bonferroni correction)
for mean RTs of each target word context condition (i.e., neutral, low, medium and high)
revealed that the difference in mean RTs between low and medium-context conditions
was statistically significant, D = 95.75 ms, 95% CI [31.82, 159.68], p = .001.
READING COMPREHENSION AND WORKING MEMORY
69
Accuracy
A one-way repeated-measures ANOVA was conducted on the IV accuracy for
each context level for word-target trials (summarised in Table 2). Mauchly’s test indicated
that assumption of sphericity was not violated for the main effect of context. A main
effect of context, F(3, 150) = .578, MSe = 49.76, p > .05 was not statistically significant,
indicating that there was no significant difference in accuracy (%) as context increased
from neutral (M = 91.57; SD = 7.58) to low (M = 91.57; SD = 6.74) to medium (M =
93.14; SD = 7.07) to high (M = 92.35; SD = 6.51).
Speed-accuracy trade-off effect
To assess whether there were consistent strategies of sacrificing speed for
accuracy, or vice versa, the relationship between RT and accuracy for target words was
examined. Specifically, to investigate the size and direction of the linear relationship
between mean accuracy (%) and overall mean RT (ms) for target words, Spearman’s rho
correlation coefficient (𝑟𝑟𝑠𝑠) was calculated, as the assumptions of normality was violated
for the Pearson’s product-moment correlation. The results indicated an absence of a
significant correlation between mean accuracy (M = 92.16, SD = 3.39) and overall mean
RT (M = 764.10, SD = 266.28), 𝑟𝑟𝑠𝑠 = -.03, p > .05, two-tailed, N = 51, evidencing that
participants were not strategically trading off accuracy for speed, or vice versa.
In addition, Spearman’s rho correlation coefficients (𝑟𝑟𝑠𝑠) were performed between
mean RT and accuracy within each condition (i.e., within each context level; neutral, low,
medium and high) to ensure no strategic responding was occurring within a single context
condition. No significant correlations were identified in any of the four context conditions
(lowest p = .457). Thus, no evidence of a speed-accuracy trade-off effect was identified.
READING COMPREHENSION AND WORKING MEMORY
70
Discourse-level reading comprehension and cognitive processes
Majority of the cognitive functions scores fell broadly within the expected range
(i.e., age-corrected standard scores, for which the normative mean is 100 (SD = 15; NIH
& NU, 2016b). The cognitive flexibility measure (M = 111.86, SD = 17.26) was an
exception, in which the mean scores were over one standard deviation above the national
average when compared to like-aged participants (NIH & NU, 2016b). Participants
scored highest for the EF cognitive flexibility, followed by episodic memory (M = 106.12,
SD = 17.58), WM capacity (M = 99.33, SD = 14.94) and lastly, inhibition and attention
(M = 97.82, SD = 13.62).
A standard multiple regression analysis (MRA) was conducted to test whether
components of WM and EFs would account for a significant proportion of variance in
adults’ discourse-level comprehension ability, as indicated by the extent of the facilitation
effect produced in the RIT. Examination of the standardised residuals identified one
possible outlier (1.96% of cases). Cook’s and Mahalanobis distances, average leverage,
and standardised DfBeta values were calculated to determine whether the outlier exerted
any undue influence in the regression model, determining it did not (Field, 2013; refer to
Appendix I for statistical calculations). The covariance ratio calculations identified four
values exceeding the upper limit of an acceptable covariance ratio. However, given
Cook’s distance for each of these participants fell within the acceptable range, the
covariance ratios were not considered to be of concern. Visual inspection of the
scatterplot of the normal probability of standardised residuals and standardised residuals
in relation to standardised predicted values indicated the assumptions of normality,
linearity and homoscedasticity of residuals were met. High tolerances for predictors
evidenced that multicollinearity did not interfere with the ability to interpret the outcome
of the MRA.
READING COMPREHENSION AND WORKING MEMORY
71
WM and EFs accounted for a nonsignificant 10% of the variance in discourse-
level comprehension ability, R² = .099, adjusted R² = -.021, F(4, 46) = 1.27, p = .30.
Unstandarised (B) and standardised errors (SE B) as well as standardised (β) regression
coefficients and squared semi-partial (‘part’) correlations (sr²) for each cognitive process
in the multiple regression model are presented in Table 3, identifying inhibition and
attention as significant predictors of adults’ discourse-level reading comprehension
ability.
Table 3
Unstandarised (B) and standardised errors (SE B) as well as standardised (β) Regression Coefficients and
Part Correlations (sr²) for Predictors in a Multiple Regression Model Predicting Discourse-level Reading
Comprehension.
Cognitive Process B [95% CI] SE B β sr² p
Inhibition and Attention -7.67 [-15.01, -.33]* 3.65 -.55 .09 .041
Cognitive flexibility 4.93 [-.94, 10.81] 2.92 .45 .06 .098
WM capacity -3.26 [-7.91, 1.40] 2.31 -.26 .04 .17
Episodic memory 1.60 [-2.30, 5.51] 1.94 .15 .01 .41
Note. N = 51. CI = confidence interval. * p < .05. ** p < .01
Propositional-level reading comprehension and cognitive processes
A standard MRA was conducted to determine whether WM and its EFs would
account for a significant proportion of variance in adults’ propositional-level reading
comprehension ability, as indicated by performance on the Nelson Denny Reading
Comprehension Test. Participants scored an average of 75.49% (SD = 16.95) on the
reading comprehension measure, which was above the standardised scores (M = 59.07%,
SD = 12.29) specified in the Nelson-Denny Reading Test Administration and Scoring
Manual. Inspection of the scatterplot of the normal probability of standardised residuals
and standardised residuals in relation to standardised predicted values indicated the
READING COMPREHENSION AND WORKING MEMORY
72
assumptions of normality, linearity and homoscedasticity of residuals were met.
Furthermore, Mahalanobis distance did not exceed the critical 𝜒𝜒² for df = 4 (at α = .001)
of 18.47 for the data set, indicating multivariate outliers were not of concern.
Multicollinearity did not require further investigation, as tolerances for predictors in the
regression model were moderately high.
WM and the EFs accounted for a significant 25% of the variance in propositional-
level comprehension ability, R² = .25, adjusted R² = .19, F(4, 46) = 3.90, p = .008.
Unstandarised (B) and standardised errors (SE B) as well as standardised (β) regression
coefficients and squared semi-partial (‘part’) correlations (sr²) for each cognitive process
in the regression model are reported in Table 4. Significant predictors of participants’
propositional-level reading comprehension ability included inhibition and attention, WM
capacity and cognitive flexibility.
Table 4
Unstandarised (B) and standardised errors (SE B) as well as standardised (β) Regression Coefficients and
Part Correlations (sr²) for Predictors in a Multiple Regression Model Predicting Propositional-Level
Reading Comprehension.
Cognitive Process B [95% CI] SE B β sr² P
Inhibition and Attention .66 [.06, 1.26] * .30 .53 .08 .033
Cognitive flexibility -.58 [-1.06, -.10] * .24 -.59 .09 .020
WM capacity .56 [.18, .94] ** .19 .49 .14 .005
Episodic memory .02 [-.28, .36] .16 .04 .00 .82
Note. N = 51. CI = confidence interval. * p < .05. ** p < .01
READING COMPREHENSION AND WORKING MEMORY
73
Propositional-level versus discourse-level reading comprehension
The partial correlation coefficients for each cognitive variable in relation to
discourse-level and propositional-level reading comprehension ability are depicted in
Figure 3. Refer to Appendix H for relevant statistical data output.
Discussion
The current study investigated the relationship between WM, its central EFs and
adults’ reading comprehension performance, particularly the contributions to discourse-
level comprehension ability. The study aimed to investigate whether specific functions of
WM, in addition to episodic memory, attributed to individual differences in reading
comprehension and situation model construction. Participants completed a series of
cognitive tasks, a standardised reading comprehension task and a measure of discourse-
level reading comprehension. Three hypotheses were postulated for this study, of which
only the first was supported.
First, the high-context condition of the RIT task (i.e., the presentation of the three
context sentences; adapted from Gouldthorp & Coney, 2011) was proposed to produce
greater facilitation for target words than the low and medium-context conditions. The
Figure 3. Partial Correlations for Predictors in a Multiple Regression Model predicting Discourse-
Level and Propositional-Level Reading Comprehension Ability. Note * p < .05; ** p < .01.
READING COMPREHENSION AND WORKING MEMORY
74
findings supported this hypothesis; it demonstrated that facilitation for target words
increased as more contextual information became available to participants. Second, it was
hypothesised that performance on cognitive tasks from the NIH measure would account
for a significant proportion of variance in discourse-level reading comprehension ability.
This premise was not supported as the results indicated that WM, its central EFs and
episodic memory did not significantly predict discourse-level comprehension
performance. Third, the hypothesis that the EF cognitive flexibility will predict
performance on both propositional-level and discourse-level of processing was not
supported. The findings indicated that inhibition and attention had a greater influence on
propositional-level and discourse-level comprehension in contrast to the other cognitive
processes assessed.
Facilitation effect
To test the first hypothesis, participants were presented centrally-presented targets
primed by sentential contexts and neutral sentences (Gouldthorp & Coney, 2011). The
results supported this hypothesis; target word RTs decreased as the level of context
increased from neutral to high. This demonstrated that, by integrating target concepts (i.e.,
words) with preceding text, context did influence reading comprehension performance
(i.e., faster RTs as context increased), which was consistent with Gouldthorp and Coney
(2011). Thus, stronger discourse-level comprehension skills were involved in the
construction of mental representations described in text (Davoudi & Moghadam, 2015;
Zwaan et al., 2002). This suggested that, although each sentence individually provided
minimal contextual information, cumulatively they contributed to greater context
(Gouldthorp & Coney, 2011). The aforementioned finding substantiates that individual
differences in integrating information across sentences with world-knowledge influenced
the production of discourse-level meaning.
READING COMPREHENSION AND WORKING MEMORY
75
Previous research identified that the presentation of three sentences specifically
facilitated greater recognition of target words than situations where one or two sentences
preceded a target word (Gouldthorp & Coney, 2011). In contrast to the findings of
Gouldthorp and Coney (2011), the present study found no significant difference in
facilitation effects between the high-context condition (i.e., three context sentences) and
the medium-context condition (i.e., two context sentences). This inconsistency in findings
possibly indicated that participants utilised contextual information from two related
sentences to accurately predict the forthcoming word; however, they did not substantially
benefit from the presentation of a third context sentence. This may have been influenced
by the cognitive strategies utilised by participants to complete the task. For example,
while completing the RIT, participants may have relied on the first two context sentences
to construct mental representations of the situation, suggesting they ignored information
provided from a third context sentence. This allowed a quicker response to the target
word, supporting the argument that the second context sentence primed the target word
as opposed to the third context sentence.
The results indicated that there was no significant difference in accuracy as
context increased from neutral to high. This was inconsistent with the findings presented
in Gouldthorp and Coney (2011); their results mirrored the pattern found for their
facilitation data (i.e., there was an increase in accuracy with more contextual support). In
the present study, the RIT (adapted from Gouldthorp & Coney, 2011) demonstrated high
levels of accuracy across all conditions, which potentially highlights the presence of
ceiling effects for this study (Ho & Yu, 2015). As a result, no meaningful variation in
participants’ performance was evidenced, thus, individual differences in reading
comprehension ability could not be identified within this study.
READING COMPREHENSION AND WORKING MEMORY
76
Relationship between cognitive processes and discourse-level reading ability
The results did not support the second hypothesis which proposed that
performance on EF tasks would account for a significant proportion of variance in
discourse-level reading comprehension ability. However, the results did indicate a
significant negative relationship between inhibition and attention and discourse-level
comprehension ability. This suggested the ability to construct mental representations of a
situation described in a text was negatively influenced by participants’ ability to ignore
irrelevant information and direct attentional resources to task-relevant information. This
was contrary to what was expected as past findings claimed that inhibition and attention
increased reading comprehension performance (Cain, 2006; Cantin et al., 2016; Diamond,
2013; Engel de Abreu et al., 2014; Nouwens et al., 2016).
This could be explained by the RIT being predicated on the situation model and
participants’ ability to create mental representations of a situation described in text
(Davoudi & Moghadam, 2015; Zwaan et al., 2002). Inhibition in this case should have
represented a participant’s ability to ‘de-activate’ or rule out irrelevant mental
representations as context from the sentences increased from neutral to high (i.e., lexical
inhibition; Davis & Lupker, 2006). However, the Flanker task employed in this study was
more appropriate for measuring response inhibition (e.g., the suppression of automatic
response tendencies which may interfere with goal achievement; NIH & NU, 2016a).
Hence, the measure may not have accurately portrayed lexical inhibition in this study. As
previously discussed in relation to context and facilitation effects, participants may have
utilised the second context sentence to conceptualise the target word and ignored the third
context sentence. Consequently, participants possibly engaged in inhibition by not
attending to the third context sentence.
READING COMPREHENSION AND WORKING MEMORY
77
Although cognitive flexibility did not significantly predict performance in
discourse-level reading comprehension ability, the results depicted a moderate positive
relationship between cognitive flexibility and discourse-level reading comprehension
performance. As such, it is possible that the study design did not have sufficient power
(based on a priori power analysis) to detect the significance of this relationship, which is
attributed to a smaller effect size obtained from the limited sample than was expected. A
larger sample size may have potentially detected this dependence.
The positive correlation between cognitive flexibility and discourse-level reading
comprehension can be explained by participants’ ability to adapt WM representations of
a situation presented in the text (Diamond, 2013; Engel de Abreu et al., 2014). As context
increased, participants were able to make amendments to the meanings they associated
with the words in the context sentence, thereby constructing a more accurate situation
model. This would have facilitated discourse-level reading comprehension ability by
allowing participants to respond more rapidly to the congruent target words. A crucial
part of this process was the activation and manipulation of different language codes (e.g.,
phonological, orthoraphic and semantic), which enabled readers to adapt mental
representations from textual information (Colé, Duncan & Blaye, 2014).
Individual differences in language processing can explain the relationship
between inhibition, attention and cognitive flexibility and discourse-level reading
comprehension. Individual differences, such as the cognitive strategies employed by
participants, may have influenced participants’ ability to construct accurate mental
representations of the textual information described in the RIT task. Moreover, it may
have masked the true effect of inhibition and cognitive flexibility, in discourse-level
reading comprehension performance. For example, the meanings participants associated
with words presented in text cannot be controlled with this experimental design. These
READING COMPREHENSION AND WORKING MEMORY
78
meanings are often multimodal (e.g., lexical or visual) and rely on the preference of the
reader (Davoudi & Moghadam, 2015; Zwaan et al., 2002). In addition, these meanings
can become irrelevant to the situation and thus were inhibited or forgotten as more context
became available to participants (Baddeley, 2012; Davoudi & Moghadam, 2015; Engel
de Abreu et al., 2014; García-Madruga et al., 2014; Nouwens et al., 2016).
Conversely, participants may have focused on only one meaning for each sentence
rather than multiple meanings. This single meaning was either replaced or affirmed as
context increased. This strategy relied on attentional control where participants allocated
attentional resources to a specific meaning (García-Madruga et al., 2014) and disregarded
alternative meanings which were relevant to the target word. This may have hindered
reading comprehension ability as the meanings associated with the context sentences
would be inconsistent with the target words, thereby delaying facilitation effects. The
delay in facilitation would not be strongly influenced by this strategy if the participants
were able to attribute meaning to the sentences that were closely related to the target word
(e.g., JAM and HONEY vs. KITE and MOON; Faust, Bar-Lev, & Chiarello, 2003).
Furthermore, participants may have utilised cognitive flexibility by switching between
the abovementioned strategies to determine which was most effective for obtaining
accurate responses on the adapted RIT measure (Nouwens et al., 2016). For example,
participants may have utilised one strategy by focussing on one meaning for each sentence
at the start and, as trials progressed, they may have changed their processing strategy to
include multiple meanings.
The results indicated a weak negative correlation between WM capacity and
discourse-level reading ability, however this was not significant. Consequently, WM
capacity did not differentiate between good and poor comprehenders, which concurred
with Daneman and Carpenter’s (1980) and Georgiou and Das’s (2016) findings. The
READING COMPREHENSION AND WORKING MEMORY
79
results from this study suggested that integration of information did not rely on
participants having to recall the spatial and/or temporal order in which context sentences
were provided to respond to target words. Moreover, the greater amount of information
participants stored in WM for cognitive processing may potentially obstruct their ability
to integrate information into coherent situation models. Accordingly, the ability to store
and manipulate information in WM was not a crucial component for discourse-level
processing (Dutke & von Hecker, 2011; Meinz & Hambrick, 2010; Turner & Engle,
1989).
There was a weak positive correlation between episodic memory and discourse-
reading comprehension, however not as significant as postulated by Mumper (2013). This
supported the claim that retrieving contextual information previously stored in LTM
slightly facilitated reading comprehension ability. Thus, participants could generate
inferences by activating additional information from long-term memory during the
reading process (Ababneh & Ramadan, 2013; Davoudi & Moghadam, 2015; Yeari & van
den Broek, 2015; Zwaan et al., 2002). This enabled participants to create visual
experiences of the textual information, thereby aiding in the construction of situation
models (Yeari & van den Broek, 2015; Zwaan et al., 2002).
Individual differences in discourse-level reading comprehension may not have
been apparent in terms of the underlying cognitive processes (i.e., inhibition and attention,
cognitive flexibility, WM capacity and episodic memory) due to the nature of discourse-
level reading comprehension measure (Georgiou and Das, 2016). This was illustrated by
the inability of the RIT to differentiate between good and poor reading comprehenders,
which potentially masked the true interactions between WM, its central EFs, episodic
memory and discourse-level comprehension. Due to the multi-dimensional nature of
discourse-level comprehension (Ababneh & Ramadan, 2013; Davoudi & Moghadam,
READING COMPREHENSION AND WORKING MEMORY
80
2015; Yeari & van den Broek, 2015; Zwaan et al., 2002), there was no apparent consensus
in the literature for situation model construction. Moreover, there is no direct measure of
overall discourse-level reading comprehension. The RIT measure employed in this study
is thus a preliminary measure which assessed the integration of world-knowledge across
sentences, which is one component of situation model construction. Consequently, it has
not effectively tapped into overall discourse-level reading comprehension processes.
Relationship between cognitive processes and propositional-level reading ability
The results indicated that WM, its central EFs and episodic memory significantly
accounted for individual differences in propositional-level reading comprehension
ability, which was consistent with several past studies (e.g., Cain, 2006; Cantin et al.,
2016; Diamond, 2013; Engel de Abreu et al., 2014; Nouwens et al., 2016). This study
revealed a significant positive relationship between inhibition and attention and
propositional-level comprehension performance. This corresponded with Arrington et al
(2014) and Gerst et al (2017), who suggested inhibition enabled readers to focus on words
and sentences relevant to the main topic while simultaneously ignoring irrelevant
information as it was presented. In addition, response inhibition facilitated propositional-
level reading ability as it enabled readers to suppress automatic and dominant responses
to the text (García-Madruga et al., 2016). Similar to García-Madruga et al (2014) and
Yogev‐Seligmann et al (2008), a ttention facilitated propositional-level comprehension
performance by enabling participants to successfully direct their attentional resources
towards task-relevant information to complete the reading comprehension task.
Contrary to previous research (e.g., Cantin et al., 2016; Diamond, 2013; Engel de
Abreu et al., 2014; Gerst et al., 2017; Nouwens et al., 2016), cognitive flexibility
negatively contributed to propositional-level reading comprehension ability. The findings
indicated that, as cognitive flexibility increased, comprehension performance on the
READING COMPREHENSION AND WORKING MEMORY
81
Nelson-Denny Comprehension Test decreased. This may be attributed to participants
relying on multiple reading strategies whilst completing the standardised comprehension
measure. By relying on multiple strategies, participants were potentially unable to
successfully adjust their cognitive strategies to meet the reading objective (i.e., selecting
the appropriate response to the corresponding reading passage; Colé et al., 2014;
Nouwens et al., 2016), thus impeding their propositional-level comprehension
performance.
There was a moderate positive correlation between WM capacity and
propositional-level reading comprehension, which significantly accounted for individual
differences in participants’ reading comprehension performance. This was consistent with
previous research which claimed that participants have a limited WM capacity (Cowan,
2010) and that the ability to store and manipulate task-relevant information temporarily
in WM accounted for individual differences in propositional-level reading
comprehension performance (Daneman & Carpenter, 1980; Dutke & von Hecker, 2011;
Meinz & Hambrick, 2010; Turner & Engle, 1989). Furthermore, this was indicative of
participants with higher scores on the comprehension measure having greater WM
capacity than those who obtained lower scores (Dutke & von Hecker, 2011; Turner &
Engle, 1989). This suggested that, in addition to the ability to store and manipulate
information in WM, skilled comprehenders could rely on fewer cognitive processes in
comparison to less skilled comprehenders (Daneman and Carpenter, 1980; Dutke & von
Hecker, 2011; Turner & Engle, 1989).
Lastly, there was a very weak positive correlation between episodic memory and
propositional-level reading comprehension ability, which was not significant. The
correlation indicated that participants did not rely on background knowledge stored in
LTM to complete the Nelson-Denny Comprehension Test. Consequently, the results
READING COMPREHENSION AND WORKING MEMORY
82
demonstrated that individual differences in adults’ propositional-level reading
comprehension ability was attributed to WM and its central EFs inhibition, attention and
cognitive flexibility rather than episodic memory.
Propositional-level versus discourse-level reading ability
The third hypothesis postulated that the EF cognitive flexibility would predict
performance on both propositional-level and discourse-level reading comprehension
ability more strongly in comparison to inhibition, WM capacity, attention and episodic
memory. The results contested this hypothesis; inhibition and attention appeared to more
strongly predict performance on both reading comprehension measures. Specifically,
inhibition and attention positively correlated with propositional-level reading
comprehension and negatively correlated with discourse-level reading comprehension.
The findings suggested the ability to inhibit dominant or automatic responses while
attending to task-relevant information aided propositional-level reading comprehension
ability, however hindered discourse-level reading comprehension ability. An explanation
for this difference in relationship can be attributed to discourse-level requiring a deeper
level of cognitive processing (i.e., construction of situation models; Fincher-Kiefer, 1993;
Zwaan & Radvansky, 1998; Zwaan et al., 2002). Furthermore, the impact of inhibition
on discourse-level reading performance may have been masked by participants utilising
various cognitive strategies whilst completing the RIT task, as previously discussed.
Cognitive flexibility significantly predicted performance on propositional-level
reading comprehension ability, however, it was not a significant predictor of discourse-
level reading comprehension performance. Cognitive flexibility positively correlated
with discourse-level processing, though the current study obtained a smaller effect size
than anticipated with the limited sample size (based on the priori power analysis).
Additionally, cognitive flexibility negatively correlated with propositional-level reading
READING COMPREHENSION AND WORKING MEMORY
83
comprehension. This difference between propositional-level and discourse-level
processing can be attributed to participants’ use of cognitive flexibility. Specifically, the
ability to adapt mental representations to the contextual information provided during the
RIT facilitated reading comprehension (Diamond, 2013; Engel de Abreu et al., 2014)
whereas switching between multiple reading strategies appeared to prevent participants
from achieving desired reading goals (Colé et al., 2014; Nouwens et al., 2016).
WM capacity did significantly contribute to propositional-level but not to
discourse-level reading comprehension ability, implying that the ability to hold
information in WM was not necessary for gaining a deeper meaning of a text. Moreover,
WM capacity positively correlated with propositional-level reading comprehension and
negatively correlated with discourse-level reading comprehension performance. Thus,
greater WM capacity appeared to assist participants to perform better on the
propositional-level comprehension measure than those with less WM capacity.
Alternatively, greater WM capacity could have hindered participants’ ability to construct
coherent situation models of textual information in relation to the discourse-level reading
comprehension.
Lastly, episodic memory was not significantly correlated with either
propositional-level or discourse-level reading comprehension performance, suggesting
that retrieving knowledge from LTM was not required to construct basic or deeper
meanings of information depicted in a text. Neither reading processes specifically relied
on the retrieval of temporal and spatial information from LTM (as assessed by the Picture
Sequence Memory Test; NIH &NU, 2016a). However, episodic memory was slightly
more positively correlated with discourse-level reading comprehension. Thus, gaining a
deeper meaning of textual information potentially relied on drawing information from
background information and generating inferences from world knowledge (Ababneh &
READING COMPREHENSION AND WORKING MEMORY
84
Ramadan, 2013; Davoudi & Moghadam, 2015; Yeari & van den Broek, 2015; Zwaan et
al., 2002).
Limitations
First, the nature of the discourse-level reading comprehension measure may have
affected the results of the study (Georgiou and Das, 2016). The RIT is a preliminary
measure of discourse-level reading comprehension and only assessed one aspect of
situation model construction – the participants’ ability to integrate information with
world-knowledge (Gouldthorp & Coney, 2011). Hence, this measure did not measure
overall situation model construction. However, an alternative measure which assesses the
multiple components of discourse-level reading comprehension (e.g., inference
generation, metaphor comprehension and integration of information) is yet to be
developed (Gouldthorp & Coney, 2011). The difficulty in constructing such a measure
can be attributed to the multi-dimensional nature of discourse-level processing (Davoudi
& Moghadam, 2015; Zwaan et al., 2002).
Second, the design of the study may have unduly influenced the results obtained.
For example, the limited sample size of the study reduced the effect size and power to
detect the relationship between cognitive flexibility and discourse-level reading
comprehension. Furthermore, the sample was restricted to participants with post-
secondary qualifications. Restricting the sample to individuals who are more
academically inclined may have skewed the results and produced higher performance on
the RIT task. Consequently, the results indicated less performance variation in the sample
thereby reducing the ability to detect individual differences in participants’ discourse-
level reading comprehension.
Third, the inability to control or monitor participants’ strategies when completing
the RIT task and the nature of the inhibition and attention measure (i.e., Flanker task) may
READING COMPREHENSION AND WORKING MEMORY
85
have affected the facilitation effect attained on the RIT. The study was unable to
determine whether participants utilised the desired WM and cognitive processes (e.g.,
inhibition and attention). Furthermore, the Flanker task may not have been appropriate in
measuring inhibition in relation to situation model construction. A measure of lexical
inhibition is possibly more relevant in terms of measuring participants’ ability to rule out
irrelevant mental representations constructed by textual information (Davis & Lupker,
2006). Nevertheless, this study provided an insight of how general cognitive abilities,
such as inhibition and attention, are drawn upon for language-based tasks (e.g., reading).
Implications
The current study’s findings are of theoretical importance as they provide an
insight into the central mechanisms of Baddeley’s (2000) WM model commonly depicted
in the reading comprehension literature. Moreover, it is a preliminary investigation into
how these central cognitive processes influence discourse-level reading comprehension.
The results of this study attempted to elucidate why inconsistencies exist in relation to
WM and discourse-level reading comprehension ability. These inconsistencies were
attributed to the validity and reliability of the reading comprehension and EF measures
used in conjunction with the relative sample and effect size of the study’s design. This
can be related to previous studies, which demonstrated similar inconsistencies in findings
(Arrington et al., 2014, Cantin et al., 2016; Christopher et al., 2012, Engel de Abreu et
al., 2014; Gerst et al., 2017; Gouldthorp et al., 2016). Furthermore, this study has
important implications in interpreting existing literature, as previous research has mainly
focussed on propositional-level comprehension. Hence, inconsistencies in the literature
can also be explained by the type of comprehension what was investigated (i.e.,
propositional-level versus discourse-level), as demonstrated in this study.
READING COMPREHENSION AND WORKING MEMORY
86
Regarding the practical implications, these results partially demonstrate that
individual differences in discourse-level and propositional-level reading comprehension
ability depend on multiple EFs. Consequently, the ability to identify, conceptualise and
evaluate these EFs is advantageous for allied health professionals as well as educators
and researchers. Comprehensively understanding the relationship between WM and
discourse-level reading comprehension can lead to the development of theoretical
models, with the focus on the importance of EFs underpinning discourse-level reading
comprehension (Sesma et al., 2009). The benefit of the abovementioned model is its
potential to facilitate the development and implementation of training and intervention
programs (e.g., García-Madruga et al., 2016), with aims of improving executive skills and
academic performance across various age ranges. In addition, the variation in the findings
between proposition and discourse-level comprehension has important implications for
developing assessment and intervention programs which can accommodate reading
impairment at both levels of comprehension.
Future research
Future studies could examine the link between WM, EFs and discourse-level
reading comprehension. Specifically, studies should develop discourse-level reading
comprehension measures and evaluate their psychometric properties to establish reliable
and valid assessment measures. Additional studies should also be conducted to gain
clearer understanding of the relationship between EFs and the different levels of reading
comprehension to clarify the unique contributions of EFs to comprehension abilities. The
research would include modelling analyses to investigate the effect of cognitive
processes, such as updating, planning and temporal awareness, in mediating the
relationship between WM, its central EFs and discourse-level reading comprehension
READING COMPREHENSION AND WORKING MEMORY
87
(García-Madruga et al., 2014; Gerst et al., 2017; Nouwens et al., 2016; Sesma et al.,
2009).
It is also recommended that this study be replicated with a larger sample size.
Participants should be recruited with a variety of education backgrounds to explicate
whether education had a significant influence on discourse-level reading comprehension
performance. That is, a sample which includes individuals with poorer reading
comprehension ability and/or lower EF and WM would enable researchers to detect more
distinct relationships between reading comprehension and WM. Alternative measures of
WM and EFs should be explored, such as the Colour-word inference test for inhibition
and the Trail-Making test for cognitive flexibility, to establish concurrent validity of the
measures (Cicchetti, 1994). Alternative measures of discourse-level reading
comprehension include those employed in studies by Fincher-Kiefer (1993; 2001) on
inferencing and Zwaan and colleagues (1998; 2002) on situation model construction.
These may assist in evaluating reading comprehension at discourse-level, however their
ability to identify individual differences in reading comprehension is yet to be assessed.
Moreover, a minor improvement can be made to the RIT (adapted from
Gouldthorp & Coney, 2011). Target words which are incongruent with the centrally-
presented sentences (e.g., replacing the target word HONEY with CAKE) should be
included. This will allow investigators to assess whether an inhibition effect for
incongruent conditions would occur (Faust et al., 2003). Thus, it may provide further
evidence of a facilitation effect where congruent words are strengthened by the context
sentences. Additionally, it would demonstrate that participants are integrating
information across sentences to develop a deeper, discourse-level understanding of the
text.
READING COMPREHENSION AND WORKING MEMORY
88
Although the relationship between WM and children’s propositional-level reading
comprehension performance has been studied extensively, this relationship with
discourse-level reading comprehension has been less empirically explored.
Approximately 10% of school-aged children demonstrate age-appropriate word decoding
skills, yet have difficulty understanding the meaning of what they read (Cain & Oakhill,
2006). Thus, it is of interest to replicate the study with children (e.g., between the ages of
7 and 14 years). Moreover, assessing reading comprehension is vital for monitoring
students’ progress, detecting comprehension difficulties and testing theories of
comprehension (Cain & Oakhill, 2006; Carlson, Seipel, & McMaster, 2011). Hence, it is
recommended that a study be conducted to examine the underlying cognitive processes
and abilities that allow primary-school aged children to produce meaning (i.e.,
comprehension) during reading. Future research in this area may identify individual
differences in children’s level of reading comprehension as well as how WM and its
central EFs account for these differences.
Conclusion
This study aimed to investigate whether individual differences in reading
comprehension and situation model construction were partly explained by differences in
functioning of specific EFs central to WM. The study’s results only substantiated the first
hypothesis (i.e., facilitation of target words increased as context increased from neutral to
high). The remaining two hypotheses were not substantiated; WM and its EF did not
significantly influence participants’ discourse-level comprehension ability; and cognitive
flexibility did not predict reading comprehension more strongly than inhibition and
attention, WM capacity and episodic memory.
Understanding the relationship between WM and its central EFs is crucial for
developing training programs aimed at improving individuals’ executive skills and
READING COMPREHENSION AND WORKING MEMORY
89
academic performance. Considering the limitations and recommendations for future
studies, this study had important theoretical and practical implications. It provided an
insight into the underlying mechanisms of Baddeley’s (2000) WM model as well as the
potential for the development of a theoretical, reading comprehension model based on the
central EFs of WM. This study also has important applications for future research,
particularly in identifying individual differences in adults’ levels of reading
comprehension as a function of these cognitive processes. Furthermore, it is
recommended future studies examine how these cognitive processes in children affect
their ability to produce meaning.
READING COMPREHENSION AND WORKING MEMORY
90
References
Ababneh, T., & Ramadan, S. (2013). Inferences in the Comprehension of Language.
Research on Humanities and Social Sciences, 3(4), 48-51. Retrieved April 4,
2017, from http://www.iiste.org/Journals/index.php/RHSS/article/viewFile/4945/
5028.
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(5), 325-346. doi:
10.1080/10888438.2014.902461.
Baddeley, A. (2012). Working memory: Theories, models, and controversies. Annual
Review of Psychology, 63(1), 1-29. doi:10.1146/annurev-psych-120710-100422.
Beeman, M., Friedman, R. B., Grafman, J., Perez, E., Diamond, S., & Lindsay, M. B.
(1994). Summation priming and coarse semantic coding in the right hemisphere.
Journal of Cognitive Neuroscience, 6(1), 26-45. doi:10.1162/jocn.1994.6.1.26.
Brown, J. I., Fishco, V. V., & Hanna, G. (1993). The Nelson-Denny reading test: Form
G. Itasca, IL: The Riverside Publishing Company.
Cain, K. (2006). Individual differences in children's memory and reading comprehension:
An investigation of semantic and inhibitory deficits. Memory, 14(5), 553-569. doi:
10.1080/09658210600624481.
Cain, K., Oakhill, J., & Bryant, P. (2004). Children's Reading Comprehension Ability:
Concurrent Prediction by Working Memory, Verbal Ability, and Component
Skills. Journal of Educational Psychology, 96(1), 31-42. doi: 10.1037/0022-
0663.96.1.31.
Cantin, R. H., Gnaedinger, E. K., Gallaway, K. C., Hesson-McInnis, M. S., & Hund, A.
M. (2016). Executive functioning predicts reading, mathematics, and theory of
READING COMPREHENSION AND WORKING MEMORY
91
mind during the elementary years. Journal of Experimental Child Psychology,
146, 66-78. doi: 10.1016/j.jecp.2016.01.014.
Carlson, S., Seipel, B., & McMaster, K. (2011). Understanding a new reading
comprehension assessment to measure discourse representationations and identify
types of comprehenders. Society for Research on Education Effectiveness.
Retrieved September 28, 2017, from https://eric.ed.gov/?id=ED528950.
Chein, J. M., & Fiez, J. A. (2010). Evaluating models of working memory through the
effects of concurrent irrelevant information. Journal of experimental psychology,
139(1), 117-137. doi: 10.1037/a0018200.
Christopher, M. E., Miyake, A., Keenan, J. M., Pennington, B., DeFries, J. C.,
Wadsworth, S. J., . . . Filosofiska, F. (2012). Predicting word reading and
comprehension with executive function and speed measures across development:
a latent variable analysis. Journal of experimental psychology, 141(3), 470.
Retrieved March 26, 2017 from https://www.ncbi.nlm.nih.gov/pmc/articles
/PMC3360115/.
Chrysochoou, E., Bablekou, Z., & Tsigilis, N. (2011). Working Memory Contributions
to Reading Comprehension Components in Middle Childhood Children. The
American Journal of Psychology, 124(3), 275-289. doi: 10.5406/amerjpsyc.
124.3.0275.
Cicchetti, D. V. (1994). Guidelines, criteria, and rules of thumb for evaluating normed
and standardized assessment instruments in psychology. Psychological
Assessment, 6(4), 284-290. doi:10.1037/1040-3590.6.4.284.
Colé, P., Duncan, L. G., & Blaye, A. (2014). Cognitive flexibility predicts early reading
skills. Frontiers in Psychology, 5, 565. doi:10.3389/fpsyg.2014.00565.
READING COMPREHENSION AND WORKING MEMORY
92
Collette, F., & Van der Linden, M. (2002). Brain imaging of the central executive
component of working memory. United States: Elsevier Ltd.
Cook, A. E., & Myers, J. L. (2004). Processing discourse roles in scripted narratives: The
influences of context and world knowledge. Journal of Memory and Language,
50(3), 268-288. doi:10.1016/j.jml.2003.11.003.
Cowan, N. (2010). The Magical Mystery Four: How is Working Memory Capacity
Limited, and Why? Current Directions in Psychological Science, 19(1), 51–57.
doi: 10.1177/0963721409359277.
Daneman, M., & Carpenter, P. A. (1980). Individual differences in working memory and
reading. Journal of Verbal Learning and Verbal Behavior, 19(4), 450-466. doi:
10.1016/S0022-5371(80)90312-6.
Davis, C. J., & Lupker, S. J. (2006). Masked inhibitory priming in english: Evidence for
lexical inhibition. Journal of Experimental Psychology: Human Perception and
Performance, 32(3), 668-687. doi:10.1037/0096-1523.32.3.668.
Davoudi, M., & Moghadam, H. R. H. (2015). Critical review of the model of the model
of reading comprehension with a focus on situation models. International Journal
of Linguistics, 7(5), 173-187. doi:10.5296/ijl.v7i5.8357.
Diamond, A. (2013). Executive functions. Annual review of psychology, 64(1), 135-168.
doi: 10.1146/annurev-psych-113011-143750.
Dutke, S., & von Hecker, U. (2011). Comprehending ambiguous texts: A high reading
span helps to constrain the situation model. Journal of Cognitive Psychology,
23(2), 227-242. doi:10.1080/20445911.2011.485127.
Empirisoft Corporation. (2011). User's Guide and Reference DirectRT v2010. New York.
NY: Empirisoft Corporation.
READING COMPREHENSION AND WORKING MEMORY
93
Engel de Abreu, P. M. J., Abreu, N., Nikaedo, C. C., Puglisi, M. L., Tourinho, C. J.,
Miranda, M. C., . . . Martin, R. (2014). Executive functioning and reading
achievement in school: a study of Brazilian children assessed by their teachers as
"poor readers". Frontiers in psychology, 5, 550. doi: 10.3389/fpsyg.2014.00550.
Faust, M., Bar-lev, A., & Chiarell, C. (2003). Sentence priming effects in the two cerebral
hemispheres: Influences of lexical relatedness, word order, and sentence anomaly.
Neuropsychologia, 41(4), 480-492. doi:10.1016/S0028-3932(02)00138-0.
Federmeier, K. D. (2007). Thinking ahead: The role and roots of prediction in language
comprehension. Psychophysiology, 44(4), 491-505. doi:10.1111/j.1469-8986.
2007.00531.x.
Field, A. (2013). Discovering statistics using IBM SPSS statistics. California, Thousand
Oaks: SAGE Publications.
Fincher-Kiefer, R. (1993). The role of predictive inferences in situation model
construction. Discourse Processes, 16(1-2), 99-124. doi: 10.1080/0163853
9309544831.
Fincher-Kiefer, R. (2001). Perceptual components of situation models. Memory &
Cognition, 29(2), 336-343. doi:10.3758/BF03194928.
Follmer, D. J. (2017). Executive Function and Reading Comprehension: A Meta-Analytic
Review. Educational Psychologist, 1-19. doi: 10.1080/00461520.2017.1309295.
García-Madruga, J. A., Gómez-Veiga, I., & Vila, J. Ó. (2016). Executive Functions and
the Improvement of Thinking Abilities: The Intervention in Reading
Comprehension. Frontiers in psychology, 7, 58. doi: 10.3389/fpsyg.2016.00058.
READING COMPREHENSION AND WORKING MEMORY
94
García-Madruga, J. A., Vila, J. O., Gómez-Veiga, I., Duque, G., & Elosúa, M. R. (2014).
Executive processes, reading comprehension and academic achievement in 3th
grade primary students. Learning and Individual Differences, 35, 41-48. doi:
10.1016/j.lindif.2014.07.013.
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. doi: 10.1016/j.ridd.2015. 12.001.
Gershon, R. C., Wagster, M. V., Hendrie, H. C., Fox, N. A., Cook, K. F., & Nowinski, C.
J. (2013). NIH toolbox for assessment of neurological and behavioral function.
Neurology, 80(11 Suppl 3), S2. Retrieved August 9, 2017 from https://www.ncbi.
nlm.nih.gov /pmc/articles/ PMC3662335/.
Gerst, E. H., Cirino, P. T., Fletcher, J. M., & Yoshida, H. (2017). Cognitive and
behavioral rating measures of executive function as predictors of academic
outcomes in children. Child Neuropsychology, 23(4), 381-407. doi:
10.1080/09297049.2015.1120860.
Gnaedinger, E. K., Hund, A. M., & Hesson-McInnis, M. S. (2016). Reading-specific
flexibility moderates the relation between reading strategy use and reading
comprehension during the elementary years: Reading-specific flexibility. Mind,
Brain, and Education, 10(4), 233-246. doi:10.1111/mbe.12125.
Gouldthorp, B., & Coney, J. (2011). Integration and coarse coding: Right hemisphere
processing of message-level contextual information. Laterality: Asymmetries of
Body, Brain and Cognition, 16(1), 1-23. doi:10.1080/13576500903130751.
Gouldthorp, B., Katsipis., L., & Mueller, C. (2017). An Investigation of the Role of
Sequencing in Children’s Reading Comprehension. Reading Research Quarterly,
1-16. doi:10.1002/rrq.186.
READING COMPREHENSION AND WORKING MEMORY
95
Heaton, R. K., Akshoomoff, N., Tulsky, D., Mungas, D., Weintraub, S., Dikmen, S., . . .
Gershon, R. (2014). Reliability and validity of composite scores from the NIH
toolbox cognition battery in adults. Journal of the International
Neuropsychological Society : JINS, 20(6), 588. doi:10.1017/S1355617714
000241.
Ho, A. D., & Yu, C. C. (2015). Descriptive statistics for modern test score distributions:
Skewness, kurtosis, discreteness, and ceiling effects. Educational and
Psychological Measurement, 75(3), 365-388. doi:10.1177/0013164414548576.
Iglesias-Sarmiento, V., Carriedo-López, N., & Rodríguez-Rodríguez, J. L. (2015).
Updating executive function and performance in reading comprehension and
problem solving. Anales de Psicología, 31(1), 298-309. doi: 10.6018/analesps
.31.1.158111.
Kiss, I., Pisio, C., Francois, A., & Schopflocher, D. (1998). Central executive function in
working memory: event-related brain potential studies. Cognitive Brain Research,
6(4), 235-247. doi: 10.1016/S0926-6410(97)00035-9.
Lewandowski, L. J., Codding, R. S., Kleinmann, A. E., & Tucker, K. L. (2003).
Assessment of reading rate in postsecondary students. Journal of
Psychoeducational Assessment, 21(2), 134-144. doi:10.1177/073428290302100
202.
McCabe, D. P., Roediger, H. L., McDaniel, M. A., Balota, D. A., & Hambrick, D. Z.
(2010). The Relationship Between Working Memory Capacity and Executive
Functioning: Evidence for a Common Executive Attention Construct.
Neuropsychology, 24(2), 222-243. doi: 10.1037/a0017619.
READING COMPREHENSION AND WORKING MEMORY
96
Meinz, E. J., & Hambrick, D. Z. (2010). Deliberate practice is necessary but not sufficient
to explain individual differences in piano sight-reading skill: The role of working
memory capacity. Psychological Science, 21(7), 914-919. doi:10.1177/0956
797610373933.
Mumper, M. L. (2013). The Role of World Knowledge and Episodic Memory in Scripted
Narratives. Psychology Honors Projects, Paper 32, 1-40. Retrieved August 27,
2017 from http://digitalcommons. macalester.edu /psychology_honors/32.
National Institute of Health and Northwestern University. (2016a). Administration
Manual. NIH Toolbox for Assessment of Neurological and Behavioral Function.
Retrieved June 20, 2017 from https://nihtoolbox.desk.com/customer/en/portal/
articles/2191534-administrator-s-manual-and-elearning-course
National Institute of Health and Northwestern University. (2016b). Scoring and
Interpretation Guide for the iPad. NIH Toolbox. Retrieved June 20, 2017 from
https://nihtoolbox.desk.com/customer/en/portal/articles/2437205-nih-toolbox-
scoring-and-interpretation-guide.
Nouwens, S., Groen, M. A., & Verhoeven, L. (2016). How storage and executive
functions contribute to children's reading comprehension. Learning and
Individual Differences, 47, 96-102. doi: 10.1016/j.lindif.2015.12.008.
Nouwens, S., Groen, M. A., & Verhoeven, L. (2017). How working memory relates to
children's reading comprehension: the importance of domain-specificity in
storage and processing. Reading and Writing, 30(1), 105. doi: 10.1007/s11145-
016-9665-5.
Oakhill, J. V., Cain, K., & Bryant, P. E. (2003). The dissociation of word reading and text
comprehension: Evidence from component skills. Language and Cognitive
Processes, 18(4), 443-468. doi: 10.1080/01690960344000008.
READING COMPREHENSION AND WORKING MEMORY
97
Schwanenflugel, P. J., & Knapp, N. F. (2016). The Psychology of Reading: Theory and
Application. New York. NY: Guilford Press.
Sesma, H. W., Mahone, E. M., Levine, T., Eason, S. H., & Cutting, L. E. (2009). The
Contribution of Executive Skills to Reading Comprehension. Child
Neuropsychology, 15(3), 232-246. doi: 10.1080/09297040802220029.
Shaughnessy, J. J., Zechmeister, E. B., & Zechmeister, J. S. (2006). Research methods in
psychology. New York: McGraw-Hill.
Slotkin, J., Kallan, M., Griffith, J., Magasi, S., Salsman, J., Nowinski, C., & Gershon, R.
(2012). NIH toolbox technical manual. Retrieved June 20, 2017 from
http://www.nihtoolbox.org/HowDoI/TechnicalManual/CognitionTechnicalManu
als/Pages/default.aspx.
Turner, M. L., & Engle, R. W. (1989). Is working memory capacity task dependent?
Journal of Memory and Language, 28(2), 127-154. doi:10.1016/0749-
596X(89)90040-5.
Weintraub, S., Dikmen, S. S., Heaton, R. K., Tulsky, D. S., Zelazo, P. D., , J., . . .
Gershon, R. (2014). The cognition battery of the NIH toolbox for assessment of
neurological and behavioral function: Validation in an adult sample. Journal of
the International Neuropsychological Society : JINS, 20(6), 567. doi:10.1017/S13
55617714000320.
Yeari, M., & van den Broek, P. (2015). The role of textual semantic constraints in
knowledge-based inference generation during reading comprehension: A
computational approach. Memory, 23(8), 1193-22. doi:10.1080/09658211.2014.
968169.
READING COMPREHENSION AND WORKING MEMORY
98
Yogev‐Seligmann, G., Hausdorff, J. M ., & Giladi, N. (2008). The role of executive
function and attention in gait. Movement Disorders, 23(3), 329-342. doi:
10.1002/mds.21720.
Zelazo, P. D., Müller, U., Frye, D., Marcovitch, S., Argitis, G., Boseovski, J., . . .
Sutherland, A. (2003). The development of executive function in early childhood.
Monographs of the Society for Research in Child Development, 68(3), vii-137.
doi: 10.1111/j.1540-5834.2003.06803001.x.
Zwaan, R. A., & Radvansky, G. A. (1998). Situation models in language comprehension
and memory. Psychological Bulletin, 123(2), 162-185. doi:10.1037/0033-
2909.123.2.162.
Zwaan, R. A., Stanfield, R. A., & Yaxley, R. H. (2002). Language comprehenders
mentally represent the shapes of objects. Psychological Science, 13(2), 168-171.
doi:10.1111/1467-9280.00430.
READING COMPREHENSION AND WORKING MEMORY
99
Appendix A
Ethics Approval Letter
READING COMPREHENSION AND WORKING MEMORY
100
Appendix B
Information Letter
(continued)
READING COMPREHENSION AND WORKING MEMORY
101
Appendix B Continued
(continued)
READING COMPREHENSION AND WORKING MEMORY
102
Appendix B Continued
READING COMPREHENSION AND WORKING MEMORY
103
Appendix C
Consent Form
READING COMPREHENSION AND WORKING MEMORY
104
Appendix D
Detailed Procedure
The Relationship between Working Memory and Higher-Level Reading Comprehension
Ability in Adults
Procedure
Sessions will be conducted in small groups, with a maximum of three participants at a time. Test sessions will last approximately 120 minutes* (which includes rest breaks), each participant will be asked to:
1. Complete the Nelson-Denny Comprehension Test (Form G) which will take approximately 20 minutes
a. Read the instructions on the test and ask the researcher any questions if needed
b. Read first passage. The first minute will be used to assess your reading rate. Upon instruction from the investigator, participants will circle the number corresponding to the respective sentence they were reading at that time. Participants are then to continue reading the remainder of the passage.
c. Answer the questions about the passage d. Continue until participants have read all seven passages and completed all
questions, or until the researcher asks them to stop.
2. Complete the NIH Toolbox Cognition Battery This task will be completed using an iPad and, for some of the subtests, a wireless keyboard. It takes approximately 25 minutes in total and involves the following tasks
a. Picture Sequence Memory Test, assessing processes involved in the acquisition, storage and retrieval of new information. 7 minutes.
b. Flanker Task, assessing inhibitory control and attention. 3 minutes. c. Listing Sorting Working Memory Test, assessing the ability to store
information until the amount of information exceeds one’s capacity to hold that information. 7 minutes.
d. Dimensional Change Card Sort Test, assessing the capacity to plan, organise and monitor the execution of behaviours that are strategically directed in a goal-orientated manner. 4 minutes.
In each of these component tests, participants are presented with automated instructions on the iPad, followed by practice items before completing the task itself. Copies of instructions and examples of items are attached here.
(continued)
READING COMPREHENSION AND WORKING MEMORY
105
Appendix D Continued
3. Complete the Reading Integration Task
This task will be completed on a computer. One, two or three sentences will be presented to vary the degree of constraint on the possibilities for the upcoming target word. The sentences will be presented in the centre of the computer screen and participants will read them at their own pace, progressing to the next sentence or target word by pressing a key. A central fixation cross will then be presented, followed by a target word or nonword. The participant will then press one of two buttons to indicate whether the target is a word or nonword. Reaction time and accuracy will be recorded. There will be 10 trials per condition, resulting in 80 trials in total. There will be 20 trials for each of: 3 sentences (high constraint), 2 sentences (medium constraint), 1 sentence (low constraint) and a neutral sentence (no constraint). Half of the trials in each of these conditions will involve a word-target and half will involve a nonword-target. Rest breaks will be offered after every 10 trials, to produce 8 blocks in total. The order of conditions presented within each block will be randomised across participants. Example stimuli are attached here.
*In both sessions, the order of the tasks listed above will be rotated across participants. Additionally, the order of the sessions will be rotated across participants.
READING COMPREHENSION AND WORKING MEMORY
106
Appendix E
Debriefing sheet
(continued)
READING COMPREHENSION AND WORKING MEMORY
107
Appendix E Continued
READING COMPREHENSION AND WORKING MEMORY
108
Appendix F
Instructions for NIH Toolbox Cognition Battery Example
C.2.2.4 NIH Toolbox Flanker Inhibitory Control and Attention Test Instructions Ages
12+
This table outlines the item content read by as well as the actions for the examiner.
(continued)
READING COMPREHENSION AND WORKING MEMORY
109
Appendix F Continued
READING COMPREHENSION AND WORKING MEMORY
110
Appendix G
Example stimuli from Reading Integration Task In the examples below, the capatalised work is the target word. The first three lines show the three sentences that could be presented for that target word: only the first sentence will be presented for the low contraint condition; the first two sentences will be presented for the medium constraint condition; the first three sentences will be presented for the high constraint condition. The fourth sentence woll be presented individually for the neutral condition. HORSE I am fast. People bet on me. I can be ridden. This is a neutral sentence. WEED I am considered unattractive. People pull me out. I am a plant. This is a neutral sentence. GRAPE I have a skin. I come in bunches. I am used to make wine. This is a neutral sentence. KITE I can fly. I need wind. I have a string attached to me. This is a neutral sentence. PLANE I am folded. I am made of paper. I can fly. This is a neutral sentence.
READING COMPREHENSION AND WORKING MEMORY
111
Appendix H
Statistical Output
Reaction Time: Repeated-Measures Analysis of Variance (ANOVA)
READING COMPREHENSION AND WORKING MEMORY
112
Facilitation Effect: ANOVA
READING COMPREHENSION AND WORKING MEMORY
113
READING COMPREHENSION AND WORKING MEMORY
114
Accuracy: ANOVA
READING COMPREHENSION AND WORKING MEMORY
115
Speed-accuracy trade-off effect
READING COMPREHENSION AND WORKING MEMORY
116
READING COMPREHENSION AND WORKING MEMORY
117
Discourse reading comprehension ability and cognitive processes
READING COMPREHENSION AND WORKING MEMORY
118
READING COMPREHENSION AND WORKING MEMORY
119
Propositional reading comprehension and cognitive processes
READING COMPREHENSION AND WORKING MEMORY
120
READING COMPREHENSION AND WORKING MEMORY
1
Appendix I
Statistical Calculations Standardised Residuals
1 case with standardised residuals with absolute value > 3
= 151
× 100
= 1.96% of cases*
2 cases with standardised residuals > 2
= 251
× 100
= 3.92% of cases* Note. * % below that specified by Field (2013).
Cook’s and Mahalanobis Distance
Max Cook's distance = .62*; max Mahal. distance was 10.96**
Note. * value below 1.00; ** value below 18.47 at df = 4
Average Leverage
𝐴𝐴𝐴𝐴𝐴𝐴𝑟𝑟𝐴𝐴𝐴𝐴𝐴𝐴 𝐿𝐿𝐴𝐴𝐴𝐴𝐴𝐴𝑟𝑟𝐴𝐴𝐴𝐴𝐴𝐴 = 3 �𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝐴𝐴𝑟𝑟 𝑜𝑜𝑜𝑜 𝑝𝑝𝑟𝑟𝐴𝐴𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑜𝑜𝑟𝑟𝑝𝑝 + 1
𝑝𝑝𝐴𝐴𝑛𝑛𝑝𝑝𝑠𝑠𝐴𝐴 𝑝𝑝𝑝𝑝𝑖𝑖𝐴𝐴� = 3 �
4 + 151
�
𝐴𝐴𝐴𝐴𝐴𝐴𝑟𝑟𝐴𝐴𝐴𝐴𝐴𝐴 𝐿𝐿𝐴𝐴𝐴𝐴𝐴𝐴𝑟𝑟𝐴𝐴𝐴𝐴𝐴𝐴 = .098*
0 cases > 3 times average leverage (i.e., .29) Note. * Max centered levearage < .29
Standardised DFBeta
0 case > standardised DFBeta value of 1.00
Covariance Ratio
Upper limit: 1 + (3 × average leverage) = 1.29
Lower Limit: 1- (3 × average leverage) = -1.29 4 cases > 1.29
Note. Average leverage is .098; * Max covariance value = 1.44
READING COMPREHENSION AND WORKING MEMORY
2
Appendix J
Project Summary
Research has repeatedly demonstrated that the ability to temporary hold incoming
information in the mind whilst carrying out other mental processes contribute to
individual differences in adults’ reading comprehension. However, less is known how
these related to an individuals’ ability to build visual imagery of information in a text.
The cognitive processes investigated within this study included individuals’ ability to:
ignore information that is not relevant to complete the task (i.e., inhibition); switch
between various mental states (i.e., cognitive flexibility); pay attention to task-relevant
information (i.e., attention); and retrieve information stored in long-term memory (i.e.,
episodic memory).
Existing literature has different stances regarding the effect of cognitive processes
on reading comprehension ability. To explain these inconsistencies, this study
investigated to what extent the ability to manipulate, integrate, and update information
quickly and efficiently was related to the depth of meaning adults gain when reading
information. Fifty-one participants, aged 18 to 56 years, completed three tasks: two
measured reading comprehension ability at lower and higher levels, and one which
measured cognitive processes.
Three hypothesises were tested in this study. First, it was proposed that presenting
three sentences (e.g., “I am sweet”, “I am spread on toast” and “I am made by insects”)
will produce greater predictability of word targets, in this case, HONEY. Second, it was
hypothesised that performance on the cognitive task will strongly predict performance on
the higher-level comprehension measure. Third, it was theorised that cognitive flexibility
will predict lower and higher-level comprehension ability more strongly in comparison
to inhibition, WM capacity, attention, and episodic memory.
READING COMPREHENSION AND WORKING MEMORY
3
The results only supported the first hypothesis; presenting three sentences enabled
participants to more accurately predict the target word at a faster rate than when the target
word was presented after a neutral sentence, one sentence or two sentences. The
remaining two hypotheses were not supported; WM and its cognitive processes did not
significantly influence participants’ high-level comprehension ability; and cognitive
flexibility did not predict reading comprehension more strongly than inhibition and
attention, WM capacity and episodic memory.
Clearly understanding the relationship between WM and its cognitive processes
is crucial for the development of training and educational programs which try to improve
children’s and adults’ decision-making skills and academic performance. Although this
study did have some limitations (e.g., small sample size and inability to directly measure
overall higher-level reading comprehension), it did have important implications. It
provided an insight into the cognitive processes underpinning reading comprehension.
Moreover, this study demonstrated that the different research stances can be can be
explained by the depth of comprehension ability that was investigated (i.e., lower versus
higher-level reading comprehension ability). This study also has important applications
for future research, particularly in identifying individual differences in cognitive
processes and how these relate to the depth of adults’ reading comprehension.
Furthermore, it is also recommended future studies examine these how these cognitive
processes in children affect their ability to produce meaning when reading.