Are There Gender Differences in The Cognitive Components of Adult

Learning and Individual Differences 32 (2014) 6979
Contents lists available at ScienceDirect
Learning and Individual Differences

journal homepage: www.elsevier.com/locate/lindif
Are there gender differences in the cognitive components of adult

reading comprehension?
Brenda Hannon
Texas A&M Kingsville, Department of Psychology and Sociology, United States
a r t i c l e
i n f o
Article history:
Received 22 May 2013
Received in revised form 21 January 2014
Accepted 21 March 2014
Keywords:
Gender differences
Reading
Cognitive processes
a b s t r a c t
This study examined: (i) whether gender differences exist in the specic cognitive components that are tapped
by measures of adult reading comprehension and (ii) whether gender differences exist in the powers of these
specic cognitive components to predict reading comprehension. The results revealed a small male advantage
for text inferencing and low-knowledge integration, d = .36 and d = .28, but no gender differences in the
remaining six cognitive components. They also revealed that high-knowledge integration, text memory, and
epistemic belief of learning were more predictive of reading comprehension performance for females than
males, whereas word decoding was more predictive of reading comprehension performance for males than
females. Taken as a whole, these results suggest that there are few quantitative gender differences in the specic
cognitive components that are tapped by measures of adult reading comprehension; however, there are important qualitative gender differences in the predictive powers of these specic cognitive components.
2014 Elsevier Inc. All rights reserved.
1. Introduction
Over the past 100 years both psychologists and the general public
have been fascinated with the idea of gender differences in cognitive
abilities (Hyde & McKinley, 1997). The most popular belief is that males
possess greater quantitative and visuospatial abilities than females
while females possess greater verbal ability than males (Maccoby &
Jacklin, 1974). Summaries of meta-analyses however, suggest that
gender differences in cognitive abilities may be much more specic in
adult populations than is popularly believed (Hyde & McKinley, 1997;
Spelke, 2005; see Hyde, 2005 and Halpern et al., 2007 for a discussion
of this point). In the domain of quantitative and visuospatial abilities,
for example, although males tend to perform better than females on
measures of mathematical problem solving, mental rotation, and spatial
perception (e.g., Halpern, 2000; Halpern et al., 2007), few, if any,
gender differences exist on measures of mathematical computation
(i.e., arithmetic), mathematical concepts, and other measures of visuospatial abilities (Halpern et al., 2007; Hyde & McKinley, 1997; Linn &
Petersen, 1985). Similarly, in the domain of verbal abilities, although
females tend to perform better than males on measures of spelling
(Kimura, 1999), word uency (Halpern, 2000; Kimura, 1999), and
language production (Halpern, 2000; Hyde & McKinley, 1997), few
if any gender differences exist on measures of vocabulary (Halpern
et al., 2007). Indeed, although there is a female advantage on measures
of reading comprehension ability when the target population is
children or adolescents (Logan & Johnston, 2010; Lynna & Mikk, 2009;
Mullis, Martin, Gonzalez, & Kennedy, 2003; Mullis, Martin, Kennedy, &
Foy, 2007) few, if any, gender differences exist when the target population is adults (Hyde & Linn, 1988; Hyde & McKinley, 1997).
http://dx.doi.org/10.1016/j.lindif.2014.03.017
1041-6080/ 2014 Elsevier Inc. All rights reserved.
Nevertheless there are at least two reasons why there is a need to

further examine gender differences in reading comprehension, especially when gender differences in the cognitive components of reading
are being considered and the target population is adults. First from a
practical perspective, determining whether there are gender differences
in specic cognitive components such as those cognitive components
tapped by measures of reading comprehension might provide insight
as to why males achieve higher scores than females on standardized
tests designed for admissions to colleges, universities, and graduate programs whereas females achieve higher grades (i.e., GPAs) than males in
school (Halpern et al., 2007; Mau & Lynn, 2001; also see Halpern, 2004).
Second, from a theoretical perspective, researchers argue that gender
differences should be understood in terms of cognitive processes rather
than the classication or type of task (e.g., verbal, quantitative, visuospatial) (see Halpern, 2004 for a discussion of the cognitive-processes
approach for examining gender differences in cognitive abilities).
Therefore, one goal of the present study was to determine whether
gender differences exist in the specic cognitive components that are
typically tapped by measures of adult reading comprehension. For
example, are there gender differences in the specic cognitive components that are used to learn explicit and implicit text-based information
(i.e., text memory, text inferencing)?; are there gender differences in
the specic cognitive components that connect text-based information
with information from prior knowledge (i.e., knowledge integration)?;
or are there gender differences in the specic cognitive components
that decode and identify words (i.e., lower-level word decoding)? The
specic cognitive components examined in the present study were
higher-level processes, which are used for learning and integrating
text (e.g., text memory, text inferencing, knowledge integration, and
70
B. Hannon / Learning and Individual Differences 32 (2014) 6979
knowledge access), lower-level processes, which are used for decoding

words, and epistemic belief of learning, which is knowledge about learning. This taxonomy of high-low cognitive processes has been adopted in
order to delineate the differences between processes that are used to remember, retrieve, and connect ideas of a text (i.e., higher-level processes) and processes that are used to decode words (i.e., lower-level
processes). These cognitive components were selected because of their
importance for constructing an integrated and coherent representation
of a text (e.g., Daneman & Hannon, 2001; Kintsch, 1988, 1998). Additionally, research suggests that measures of higher- and lower-level cognitive processes account for as much as 55% of the variance in reading
comprehension performance in an adult population (Hannon, 2012a;
Hannon & Daneman, 2001a, 2006, 2009).
A second goal of the present study was to determine whether gender
differences exist in the powers of the specic cognitive components to
predict measures of adult reading comprehension ability. That is, does
text memory predict reading comprehension performance to the same
extent for males and females?; does text inferencing predict reading
comprehension performance to the same extent for males and
females?; does knowledge integration predict reading comprehension
performance to the same extent for males and females?; and do
lower-level word decoding processes predict reading comprehension
performance to the same extent for males and females? Below I briey
review developmental research examining gender differences in reading comprehension performance and its cognitive components. Next I
relate this developmental literature to the adult literature that has
examined gender differences in verbal abilities. In the nal section, I
describe the present study.
1.1. Background
National and international assessments consistently observe a
gender difference in reading comprehension ability in children
(Department for Children, Schools and Families (DCSF) (DCSF),
2008a,b,c; Mullis et al., 2003, 2007). Indeed, international studies
examining reading comprehension in 10-year old children have
observed a gender difference favoring girls in 35 of 40 participating
countries (Mullis et al., 2003, 2007). This gender difference exists
regardless of the type of writing system, alphabetic or ideographic
orthography (Ming Chui & McBride-Chang, 2006; Mullis et al.,
2003, 2007), and it extends well into adolescence (e.g., Ming Chui
& McBride-Chang, 2006). Given the importance of reading skill in
academics and employment, this gender gap potentially has a profound impact on males (Clinton et al., 2012; Riordan, 1999; Wood,
2003).
Reading comprehension, however, is a complex construct that is
composed of a number of cognitive component processes (Cain,
Oakhill, & Bryant, 2004; Graesser, Singer, & Trabasso, 1994; Hannon,
2012a; Hannon & Daneman, 2001a, 2006, 2009; Hannon & Frias,
2012; McNamara & Magliano, 2009).1 Indeed, research suggests that
reading comprehension is composed of: (i) lower-level processes that
identify and decode words (Cunningham, Stanovich, & Wilson, 1990),
(ii) higher-level processes that extract explicit information from text
(Hannon, 2012a), connect text-based ideas (i.e., text-based inferences:
Hannon & Daneman, 2001a), establish text coherence by connecting
or bridging text-based ideas with prior knowledge (Singer & Ritchot,
1996), and embellish the text using prior knowledge (e.g., thematic
and predictive inferences: Hannon & Daneman, 1998; Long, Oppy, &
Seely, 1994), as well as (iii) knowledge about learning (i.e., epistemic
belief of learning: Daneman & Hannon, 2001). Consequently, any one
or a combination of these components might be a major source of the
gender difference in reading comprehension ability.
1
For this reason, measures of reading comprehension are often labeled as measures of
global verbal abilities whereas measures of its cognitive components are labeled as measures of spe verbal abilities or specic cognitive components.
Because of this complexity, a few developmental researchers have

begun to examine gender differences in the specic cognitive that are
typically tapped by measures of global reading comprehension ability.
One surprising nding is that, for children, gender differences in the
specic cognitive components are limited to only a small subset of
components rather than all of the cognitive components that are part
and parcel of reading comprehension. For example, using a think
aloud protocol Clinton et al. (2012) examined gender differences in
the frequency that grade 4 children generated four cognitive components of reading comprehension, namely re-instatement inferences
(i.e., re-instatement of a previous fact to explain a current fact), connective inferences (i.e., inferences that connect concurrent sentences),
knowledge-based inferences (i.e., inferences requiring explanations/
predictions based on prior knowledge), and text-based memory statements (i.e., paraphrases or repetition of the text). Although Clinton
et al. observed that females generated more re-instatement inferences
than their male counterparts, they also observed no gender differences
in the generation of connective inferences, knowledge-based inferences, or text-based memory statements.
In a subsequent study, Seipel, Clinton, and Carlson (2012) extended
the ndings of Clinton et al. (2012) by examining whether a gender difference existed in connective inferences that were either semanticallyor episodically-based. Although Seipel et al. observed that females generated more episodic connective inferences than their male counterparts,
there was no gender difference in semantic connective inferences.
Therefore, the combined results of Clinton et al.'s and Seipel et al.'s studies suggest that, for grade 4 children, gender differences in the specic
cognitive components of reading comprehension are limited to reinstatement inferences and episodically-based connective inferences.
With respect to adults, to date no study has examined gender differences in the specic cognitive components that are tapped by measures
of reading comprehension ability. Thus, from this viewpoint, it is unclear whether gender differences in the specic cognitive components
of reading comprehension do or do not exist. Moreover, in the broader
context of verbal abilities, the evidence for gender differences is equivocal. On the one hand research suggests that gender differences in an
adult population are minimal to non-existent on measures of reading
comprehension and vocabulary (e.g., Halpern et al., 2007; Hyde, 2005;
Hyde & McKinley, 1997). For the present study, this lack of a relationship between gender and vocabulary is particularly relevant because
vocabulary knowledge is highly predictive of adult reading comprehension ability (e.g., Daneman, 1991; Sternberg & Powell, 1983). And because the shared variance between vocabulary knowledge and reading
comprehension ability is largely a consequence of specic cognitive
components that are common to both constructs (e.g., Hannon &
Daneman, 2001a, 2006; Sternberg & Powell, 1983), it is possible that
the present study will reveal no gender differences in the specic component processes that are tapped by measures of adult reading comprehension ability.
However other research suggests that there are gender differences in some of the important measures of verbal abilities. For instance, females perform better than males on measures of spelling
(e.g., Kimura, 1999), word uency (e.g., Halpern, 2000; Kimura, 1999),
language production (e.g., Halpern, 2000; Hyde & McKinley, 1997), perceptual speed (e.g., Halpern, 2000), and episodic memory (e.g., Guillem
& Mograss, 2005; Halpern, 2000). On the other hand, males perform
better than females on measures of verbal analogies (e.g., Lim, 1994),
deductive reasoning, and analytic reasoning (e.g., Colom, Contreras,
Arend, Leal, & Santacreu, 2004). These latter ndings are particularly
relevant to the present study because other research suggests that measures of the higher-level cognitive components of text inferencing and
knowledge integration are predictive of performance on measures
of verbal analogies, deductive reasoning, and analytic reasoning
(e.g., Hannon & Daneman, 2001a). Thus, it is possible that the present
study will reveal male advantages on the higher-level cognitive components of text inferencing and knowledge integration.
71
1.2. Summary and present study
2.2. Design
In summary, a few developmental studies have revealed gender differences in re-instatement inferences, a type of inference that is tapped
by measures of children's reading comprehension ability, and episodic
connective inferences, inferences that connect concurrent sentences in
a text. However to date, no study has examined gender differences in
the specic cognitive components that are tapped by measures of
adult reading comprehension ability. Nevertheless, given that there is
some evidence of both female and male advantages on some measures
of verbal abilities, it is quite possible that there are also gender differences on some of the specic cognitive components tapped by measures
of adult reading comprehension ability. Therefore one goal of the
present study was to determine whether gender differences exist in
some of the specic cognitive components that are typically tapped by
measures of reading comprehension in adults. A second, novel goal
was to examine whether gender differences exist in the powers of the
specic cognitive components to predict performance on measures of
adult reading comprehension ability.
In order to achieve these goals, participants completed a number of
measures. To assess higher-level processes, participants completed
Hannon and Daneman's cognitive component task (e.g., Hannon,
2012a,b; Hannon & Daneman, 2001a, 2006, 2009). The cognitive component task provides estimates of a reader's ability to access prior
knowledge from long-term memory and to integrate this prior knowledge with new text-based information, to learn new text-based information, and to make text-based inferences. The cognitive component
task was selected over other measures because its multi-component nature saves time by eliminating the need to administer multiple measures (Hannon & NcNaughton-Cassill, 2011), it has a high degree of
reliability (Hannon, 2012a), and it is highly predictive of reading comprehension ability (Hannon, 2012a; Hannon & Daneman, 2001a, 2006,
2009; Hannon & NcNaughton-Cassill, 2011).
Participants also completed ve measures of lower-level word processes. These measures included Bell and Perfetti's (1994) word and
pseudoword decoding task as well as their phonological and orthographic lexical decision tasks. In addition, participants completed a
measure of epistemic belief of learning, which has been shown to be
predictive of reading comprehension ability (Daneman & Hannon,
2001). Finally, participants completed two measures of quantitative
abilities, namely math word problems and arithmetic, and two measures of visuospatial abilities, namely spatial rotation and spatial construction. These measures were administered for two reasons. First,
because previous research shows a male advantage on measures of
math word problems and spatial rotation (Halpern, 2000; Halpern
et al., 2007) but no gender differences on measures of arithmetic and
spatial construction (Halpern, 2000; Hyde & McKinley, 1997), these
four measures provide evidence of the normality/typicality of the
present study's participants. Second, because the measurers of arithmetic and spatial construction should reveal no gender differences,
these two measures should serve as evidence of discriminant validity
if gender differences in the cognitive components of reading comprehension should be observed.
Participants completed the following tasks in the following order:

(i) a word and pseudoword decoding task (Bell & Perfetti, 1994),
(ii) Baddeley, Logie, Nimmo-Smith, and Brereton's (1985) word-nonword decoding task, (iii) the cognitive component task (Hannon &
Daneman, 2001a), (iv) form E of the NelsonDenny test of reading comprehension ability (Brown, Bennett, & Hanna, 1981), (v) a standardized
measure of math problems (Nelson Education, 1983), (vi) a standardized measure of spatial rotation (Shepard & Metzler, 1971), (vii) an
orthographic lexical decision task (Bell & Perfetti, 1994), (viii) a phonological lexical decision task (Bell & Perfetti, 1994), (ix) form F of the
NelsonDenny test of reading comprehension ability (Brown et al.,
1981), (x) a standardized measure of arithmetic (Nelson Education,
1983), and (xi) a standardized measure of spatial construction (Nelson
Education, 1983). All tasks were completed in one session. The rst
two tasks were completed individually and the remaining tasks were
completed in small groups of one to three. The total session time was
2.5 to 3.0 h.
2. Methods
2.1. Participants
One hundred and forty Introductory to Psychology students participated in this study for course credit. All students were pre-screened to
ensure that they were free of any known learning disability and that
they were monolingual, native English speakers who spoke few words
in another language. Monolingual English speakers were used because
many of the measures used in the present study were designed for
only English speakers. Seventy participants were male and 70 participants were female. The average age was 19.6 years, SD = 1.89.
2.3. Measures
2.3.1. Arithmetic ability
This measure was a subtest of the General Aptitude Test Battery
(i.e., GATB) (e.g., Form A: Nelson Education, 1983). There were 50
multiple-choice items; for example 634 6 = [a] 3794, [b] 3804,
[c] 3812, [d] 3854, and [e] none of these. Students were given
6 min to complete as many items as possible. The dependent measure
was the total number correct.
2.3.2. Math word problems
This measure was a subtest of the GATB (e.g., Form A: Nelson
Education, 1983). There were 25 multiple-choice items; for example,
how many pages are there in 27 booklets with 15 pages each? [a] 12
pages, [b] 42 pages, [c] 395 pages, [d] 405 pages, and [e] none of these.
Students were given 7 min to complete as many items as possible. The
dependent measure was the total number correct.
2.3.3. Spatial rotation ability
This was Shepard and Metzler (1971). Each of its 20 trials consisted
of a target drawing depicting a complex three-dimensional object
composed of small blocks and four additional drawings of threedimensional multi-block objects. Two of these four additional drawings
were the original target drawing except that they were represented at a
different angle of orientation. Students were asked to identify these two
drawings in as many trials as possible in 5 min. The dependent measure
was the total number correct.
2.3.4. Spatial construction ability
This measure was a subtest of the GATB (e.g., Form A: Nelson
Education, 1983). There were 40 multiple-choice trials. Each trial
consisted of a two-dimensional drawing (i.e., target drawing), which
had dotted lines separating its subsections, and four additional drawings, which depicted three-dimensional objects. Students were asked
to identify which of the four three-dimensional objects represented
the target drawing when it was folded on the dotted lines. Students
were given 6 min to complete as many trials as possible. The dependent
variable was the total number correct.
2.3.5. Reading comprehension
Students were administered two standardized tests of global reading comprehension ability called the NelsonDenny (Forms E and F:
Brown et al., 1981). Each reading test consisted of eight short prose
passages and 36 multiple-choice questions. For each test, students
were given 20 min to read the passages and answer as many questions
as possible. Scores on the Nelson Denny correlate well with scores on
72
the SAT-V reading passages, r = .55 to .74 (e.g., Daneman & Hannon,
2001; Hannon & Daneman, 2006).
In order to verify that the two measures of reading comprehension
ability loaded on the same factor, a factor analysis with a promax
rotation (i.e., correlated solution) was completed by entering the two
measures of reading comprehension ability; no other measures were
entered. The results of the factor analysis conrmed that the two measures loaded on the same factor. Indeed the reading comprehension factor had an eigenvalue of 1.73 that accounted for 86.6% of the variance.
Because both measures of reading comprehension loaded on a single
factor, a composite reading comprehension score was calculated by
rst multiplying each reading comprehension measure's factor loading
with a student's score for that measure and then summing the two
products. For example, (reading comprehension measure one factor
loading reading comprehension measure one score) + (reading
comprehension measure two factor loading reading comprehension
measure two score). A higher score indicated higher reading comprehension ability.
2.3.6. Higher-level processes
Text memory, text inferencing, knowledge integration, and knowledge access were assessed using a variant of the component processes
task (Hannon, 2012a; Hannon & Daneman, 2001a, 2001b, 2006, 2009;
Hannon & Frias, 2012). Because the component processes task is described in detail elsewhere (i.e., Hannon, 2012a; Hannon & Daneman,
2001a, 2006, 2009) it is only briey described below.
In the component processes task students learned three-sentence
paragraphs that described relationships among two real and three nonsense terms, for instance: A NORT resembles a JET but is faster and weighs
more. A BERL resembles a CAR but is slower and weighs more. A SAMP
resembles a BERL but is slower and weighs more. After learning these
sentences, students answered truefalse statements that assessed
low-knowledge access (e.g., A JET is faster than a CAR.), high-knowledge
access (e.g., A JET has a pilot whereas a MOTORCYCLE does not.). lowknowledge integration (e.g., A JET is faster than a CAR.), high-knowledge
integration (e.g., Like PLANES, NORTS y in the air), text memory (e.g., A
NORT is faster than a JET.), and text inferencing. For example, for the
above vehicle paragraph a participant can make the text inference that
A SAMP is slower than a CAR. from the paragraph facts that a BERL is slower
than a CAR (i.e., sentence 2 of the paragraph) and a SAMP is slower than a
BERL (i.e., sentence 3 of the paragraph). For each of the components,
accuracy was the dependent measure with higher scores representing
better performance. There were 72 text memory statements, 36 text
inferencing statements, 24 low-knowledge integration statements, 36
high-knowledge integration statements, 36 low-knowledge access statements, and 24 high-knowledge access statements.
2.3.7. Lower-level word processes
Five measures of lower-level word processes were administered.
Two of the measures the orthographic and phonemic lexical decision
tasks were variants of tasks administered by Bell and Perfetti (1994)
and Olson, Kliegl, and Davidson (1983). In the orthographic task,
students decided which of the two letter strings was a real word
(e.g., mare or mair), whereas in the phonemic task students decided
which of the two pseudowords could be pronounced as a real word
(e.g., mair or muit). In each task, students viewed two letter strings
positioned in the middle of a computer screen (e.g., mait muit) and
then pressed a key marked L or R to represent their decision. Students
completed four practice trials and then two blocks of 48 trials each. For
each task, accuracy was the dependent measure. The orthographic task
took 810 min to complete and the phonemic task took 1015 min.
The word and pseudoword decoding task was a variant of a task used
by Bell and Perfetti (1994) in which students vocalized 13 lists of stimuli
as quickly and accurately as possible. Each list consisted of 22 letter
strings varying in length (i.e., 5-letter or 8-letter), number of syllables
(i.e., 1, 2, or 3 syllables) and type of letter strings (i.e., pseudowords,
high-frequency words, or low-frequency words). The frequencies for

high- and low-frequency words were 100 to 199 per million and 4 to
9 per million respectively (Kucera & Francis, 1967). List one served as
practice. The other lists were randomized and each student completed
the same randomized order. Bell and Perfetti propose that vocalization
response times to real words measure speed of lexical access (and
name production) which is sensitive to the familiarity and length of
words. In contrast, vocalization response times to pseudowords measure phonemic translations of letter strings. Because the two response
times reect different aspects of word decoding, two dependent variables were created; word decoding was the average vocalization time
for lists containing words and pseudoword decoding was the average
vocalization time for lists containing pseudowords.
The fth task was a variant of a word-pseudoword decoding
measure created by Baddeley et al. (1985). In this task, students were
presented with ve lists of 50 letters strings that were either words or
pseudowords. The rst list served as practice. For each list, students
were given 20 s to indicate whether the letter strings were words or
non-words. The dependent variable was the average number of letter
strings correctly identied across the four critical lists.
In order to verify that all of the measures of lower-level word
processes loaded on the same factor, a factor analysis with a promax
rotation (i.e., correlated solution) was completed by entering the ve
measures of word decoding; no other measures were entered into the
factor analysis. The results of the factor analysis conrmed that all the
measures loaded on the same factor. Specically, this word decoding
factor had an eigenvalue of 2.66 that accounted for 53.2% of the variance.
Because each measure loaded on a single factor, a composite word
decoding score was calculated by rst multiplying each measure's factor
loading with a student's score for that measure and then summing the
ve products; for example, (orthographic factor loading orthographic
score) + (phonological factor loading phonological score) + (word
decoding factor loading word score) + (pseudoword decoding factor
loading pseudoword decoding score) + (Baddeley et al.'s task factor
loading Baddeley et al.'s score). A higher score indicated better skill at
word decoding. See Hannon & NcNaughton-Cassill (2011) for a similar
procedure for calculating a composite measure.
2.3.8. Epistemic belief of learning
In this task students responded to 12 items selected from two subsets of Schommer's (1990) epistemology questionnaire (see Hannon &
Daneman, 2001a; Hannon & NcNaughton-Cassill, 2011; Rukavina &
Daneman, 1996 for an identical administration). For each statement
students identied their level of agreement using a ve-point Likert
scale. Lower scores represented mature beliefs about learning whereas
higher scores represented nave beliefs. A sample item from this
measure is The best thing about science courses is that most problems
have only one solution.
3. Results
The Results section includes ve subsections. Section 3.1 reports the
results of the data screening while Section 3.2 reports the validation of
the sample of participants as well as a factor analysis of the two
measures of reading comprehension ability. Section 3.3 reports the
descriptive statistics as well as replications of previous ndings that predict specic relationships among the predictors. Section 3.4 reports the
analysis examining gender differences in global verbal abilities and
specic cognitive components and nally, Section 3.5 reports the results
of the regression analyses that assess gender differences in the powers
of the specic components (e.g., higher- and lower-level processes
and epistemic belief of learning) to predict reading comprehension
ability.
As in Hannon and Frias (2012) a correlation greater than .50 was
deemed to be large, a correlation between .36 and .49 was deemed to
be medium, and a correlation of .35 or less was deemed to be small.

Table 1
Descriptive statistics for all of the measures for the group data (n = 140).
Mean (std.)
Quantitative abilities
Math problems
Arithmetic
Visuospatial abilities
Spatial rotation
Spatial construction
Global verbal abilities
Reading comp-1
Reading comp-2
Composite reading comp
Specic component processes
Text memory
Text inferencing
Low-knowledge integration
High-knowledge integration
Low-knowledge access
High-knowledge access
Composite word decoding
Knowledge about learning
Epistemic belief
Range
Skewness
Kurtosis
11.30 (2.36)
23.59 (3.52)
6.0020.00
13.0033.00
0.43
0.15
0.21
0.09
16.14 (6.48)
2157 (4.72)
5.0039.00
12.0034.00
0.91
0.16
0.91
0.23
23.70 (5.40)
27.12 (4.64)
47.31 (8.71)
13.0035.00
14.0036.00
27.0065.17
0.07
0.36
0.01
0.86
0.31
0.74
64.60 (23.09)
24.72 (5.10)
20.33 (2.64)
26.46 (5.49)
33.59 (1.85)
22.53 (1.64)
120.05 (11.69)
36.0084.00
10.0034.00
13.0024.00
16.0036.00
27.0036.00
17.0024.00
79.48143.77
0.40
0.63
0.43
0.11
0.85
1.38
0.71
0.73
0.11
0.82
1.13
0.80
1.56
0.32
34.47 (4.99)
23.0045.00
0.09
0.32
Note. For the composite measure of word decoding, higher scores are considered to be
better scores.
For all t-tests, signicance was p .05 and effect sizes (i.e., Cohen's d)
accompanied each t-test. According to Howell (1997), a d 0.80 is a
large effect that is rare in the behavioral sciences, a d between 0.50
and 0.79 is a medium effect, a d between 0.20 and 0.50 is a small effect,
and a d 0.20 is trivial. See Runyon, Coleman, and Pittenger (2000)
for a discussion of effect sizes and Hannon (2012b) and Hannon
and Daneman (2001b, 2007) for calculations of effect sizes using
eta-squared.
3.1. Data screening
SAS and PRELIS were used to screen the data for: (i) skewness
(i.e., values 3.00), (ii) kurtosis (i.e., values 3.00), (iii) outliers
(univariate statistics: studentized residual, DFITTS, DFBETAS, Cook's
D; Tabachnick & Fidell, 2007); (iv) values that exerted excessive leverage (leverage statistic, h, also known as the hat value); (v) linearity
(bivariate scatterplots); (vi) normality (univariate: normality probability plots; multivariate: Mardia's statistic; West, Finch, & Curran, 1995);
and (vii) multicollinearity (tolerance test via regression analysis).
As Table 1 shows, skewness and kurtosis statistics for all of the measures were well within acceptable limits. The remaining data screening
revealed that no outlier exerted considerable leverage (i.e., studentized
residual, DFITTS, DFBETAS, and the leverage statistic, h, all within acceptable limits). Most notably, Mardia's statistic was 1.09, which was
well below the critical 1.96 limit (i.e., there was multivariate normality)
and tolerance values were all above the 0.20 limit, which indicated very
little multicollinearity.
3.2. Validation of sample
As Table 1 shows, the pattern of correlations among the measures of
quantitative abilities, visuospatial abilities, and global verbal abilities
was consistent with the ndings of previous research (e.g., Daneman
& Tardif, 1987). That is, the two measures of quantitative ability,
math problems and arithmetic, were highly correlated with one another,
r = 0.64, but were less correlated with (i) the two measures of visuospatial ability, spatial construction and spatial rotation, range of r =
0.05 to r = 0.30, as well as (ii) the two measures of global verbal abilities, reading comprehension, range of r = 0.27 and 0.39. Additionally,
the two measures of visuospatial ability were highly correlated with
73
one another, r = 0.50, but were, at best, weakly correlated with the
two measures of global verbal abilities, range of r = 0.14 to r = 0.28. Finally, the two measures of global verbal abilities were highly correlated
with each other, r = 0.73.
In addition, as Tables 1 and 2 show the typical gender differences
in quantitative and visuospatial abilities were also observed. That is,
there was a signicant correlation between gender and math problem
ability (i.e., r = 0.30), such that males performed better than females
(i.e., 12.01 versus 10.59 respectively), t (138) = 3.74, p b .05, d =
0.63 but there was no signicant correlation between gender and
arithmetic ability, r = 0.05 (i.e., males: 23.75 versus females:
23.43), t b 1.0, d = 0.09. There was also a signicant correlation
between gender and spatial rotation ability (i.e., r = 0.38), such
that males performed better than females (i.e., 18.60 versus 13.69
respectively), t (138) = 4.83, p b .05, d = 0.82 but there was no
signicant correlation between gender and spatial construction ability, r = 0.07 (i.e., males: 21.91 versus females: 21.23), t b 1.0, d =
0.15. Because all of these results are consistent with previous research
(e.g., Halpern, 2000; Halpern et al., 2007), it can be concluded that the
group of participants used in the present study is a typical one.
3.3. Descriptive statistics and pattern of correlations
Descriptive statistics and correlations are reported in Tables 1 and 2.
As Table 1 shows, all of the measures captured a wide range of abilities.
Further as Table 2 shows, the pattern of correlations among the higherlevel processes was consistent with previous research (e.g., Hannon,
2012a; Hannon & Daneman, 2001a, 2006, 2009; Hannon & Frias,
2012). That is, the two text-based processes, text memory and text
inferencing, were highly correlated with one another, r = 0.81, but
were, at best, weakly correlated with the two measures that assessed
access to prior knowledge, low- and high-knowledge access, average
r = 0.22, range of r = 0.11 to r = 0.31. On the other hand, the two
measures of knowledge integration, low- and high-knowledge integration, were strongly correlated with the two measures of textbased processes, average r = 0.62, and also correlated with the two
measures of knowledge access, average r = 0.32. Finally, there was
a small correlation between the two measures of knowledge access,
r = 0.33; see Hannon and Daneman (2001a) for a similar sized
correlation.
In addition, the composite measure of word processing was, at best,
weakly correlated with the six measures of higher-level processes,
average r = 0.27 and range of r = 0.19 to 0.34, a nding that is consistent with previous research that suggests lower-level word and
higher-level processes are separate constructs (Hannon, 2012a). As
well, the correlations between the rst reading comprehension measure (i.e., reading comprehension-1) and the specic cognitive components and the correlations between the second reading comprehension
measure (i.e., reading comprehension-2) and the specic cognitive
components are highly similar, range of r = 0.35 versus average r =
0.36 respectively. This nding suggests that the two reading measures
are equivalent.
3.4. Are there gender differences in measures of global verbal abilities and
specic cognitive components?
This section had two goals. The rst goal determined whether a
gender difference existed in global verbal abilities, namely the measures
of reading comprehension ability. The second goal determined whether
gender differences existed in specic cognitive components, namely the
measures of higher-level processes (e.g., text memory, text inferencing,
knowledge integration, and knowledge access), the composite measure
of word decoding, and the measure of epistemic belief of learning. As
mentioned earlier, the term global verbal abilities is used to describe
reading comprehension, which is composed of multiple component
74
Table 2
Correlations among gender, measures of quantitative abilities, visuospatial abilities, global verbal abilities and specic cognitive components for the group data (n = 140).
1. Gender
2. Math problems
3. Arithmetic
4. Spatial rotation
5. Spatial construction
6. Reading comprehension-1
7. Reading comprehension-2
8. Text memory
9. Text inferencing
10. Low-knowledge integration
11. High-knowledge integration
12. Low-knowledge access
13. High-knowledge access
14. Epistemic belief
15. Composite word decoding
16. Composite reading comprehension
10
11
12
13
14
15
16
.30
.05
.64
.38
.23
.01
.39
.35
.18
.27
.05
.03
.37
.27
.03
.20
.07
.30
.21
.50
.04
.20
.27
.07
.13
.36
.37
.73
.65
.66
.02
.12
.24
.26
.31
.25
.41
.41
.11
.10
.10
.06
.07
.35
.36
.18
.02
.40
.33
.04
.17
.32
.35
.62
.50
.06
.11
.19
.07
.16
.21
.27
.20
.07
.47
.51
.15
.28
.73
.11
.03
.02
.36
.36
.81
.07
.22
.17
.14
.23
.13
.03
.09
.45
.41
.14
.21
.19
.11
.20
.29
.33
.14
.17
.21
.22
.12
.03
.22
.53
.46
.27
.25
.30
.34
.19
.29
.15
.27
.94
.92
.44
.39
.36
.41
.27
.25
.07
.38
.53
Denotes p .05.
processes, whereas the term specic cognitive components is used to

describe the components of reading comprehension.
Multiple types of statistical analyses were completed: (i) t-tests
were used to assess gender differences between means, (ii) ranges
and upper and lower tails of the distributions (i.e., upper and lower
10%) were used to assess differences in variabilities, and (iii) Cohen's
d's (Cohen, 1988) were used to report the effect sizes.
3.4.1. Global verbal abilities
Table 3 shows the means, ranges, upper and lower tails of the
distributions, and Cohen's d's for the measures of global verbal abilities
(i.e., reading comprehension) as a function of gender. Consistent with
previous research (e.g., Hyde & McKinley, 1997), there was no evidence
of gender differences in global verbal abilities. Specically, as Table 3
shows there were no gender differences in performance on the two individual measures of reading comprehension (e.g., 23.87 versus 23.53
and 27.16 versus 27.14 respectively), t b 1.0 and t b 1.0 respectively,
or the composite measure of reading comprehension (e.g., 47.45 versus
47.17), t b 1.00 and all of the Cohen's d's were trivial (i.e., below 0.20).
As well, there were no gender differences in the ranges of scores for
these three measures of global verbal abilities and the ratios of males
to females in the upper and lower tails of the distributions for these
three measures were nearly equivalent.
3.4.2. Specic cognitive components
Table 3 shows the means, ranges, upper and lower tails, and Cohen's
d's for the measures of specic cognitive components as a function of
gender (i.e., the higher-level processes: text memory, text inferencing,
knowledge integration, and knowledge access, as well as word
decoding, and epistemic belief of learning). Only two small consistent
gender differences were observed. One difference was that males
performed better than females on text inferencing, 25.63 versus 23.81
respectively; t (138) = 2.13, p b .04, and d = 0.36. Subsequent variability analysis partially supported this male advantage by showing
that there were more females than males in the bottom tail of the distribution, 79.00% versus 22.00% respectively. However, there were no
Table 3
Means, standard deviations, ranges, proportions of males and females in the upper/lower 10% (i.e., tails) of the distributions, and Cohen's d for the measures of quantitative abilities,
visuospatial abilities, global verbal abilities, and specic cognitive components as a function of gender.
Mean (std.)
Females
Range
Males
Females
Upper (lower) tails

Males
Females
Cohen's d
Males
Quantitative abilities
Math problems
Arithmetic
10.59 (2.26)
23.43 (3.84)
12.01 (2.26)
23.74 (3.18)
6.0015.00
13.0033.00
8.0020.00
16.0031.00
.43 (.86)
.64 (.57)
.57 (.14)
.36 (.43)
.63
.09
Visuospatial abilities
Spatial rotation
Spatial construction
13.69 (4.81)
21.23 (4.41)
18.60 (7.02)
21.94 (5.03)
5.0025.00
12.0029.00
8.0039.00
12.0034.00
.00 (.93)
.29 (.59)
1.00 (.07)
.71 (.43)
.82
.15
General verbal abilities

Reading comp-1
Reading comp-2
Composite reading comp
23.87 (4.91)
27.16 (4.91)
47.45 (8.67)
23.53 (5.88)
27.14 (4.38)
47.17 (8.80)
14.0035.00
14.0035.00
27.0065.16
13.0034.00
16.0036.00
27.9364.24
.50 (.57)
.50 (.43)
.06
.01
.03
63.97 (12.75)
23.81 (5.43)
19.96 (2.82)
26.07 (5.64)
33.66 (1.93)
22.63 (1.61)
119.19 (12.4)
64.91 (11.48)
25.63 (4.60)
20.70 (2.40)
26.67 (5.36)
33.51 (1.78)
22.43 (1.68)
121.90 (10.93)
36.0084.00
10.0034.00
13.0024.00
16.0036.00
28.0036.00
17.0024.00
79.48140 .29
38.0084.00
11.0033.00
15.0024.00
16.0035.00
27.0036.00
18.0024.00
93.25143.77
.57 (.57)
.50 (.79)
.57 (.71)
.50 (.57)
.57 (.64)
.64 (.43)
.64 (.57)
.33 (.33)
.50 (.22)
.43 (.29)
.50 (.43)
.43 (.36)
.36 (.57)
.36 (.43)
.08
.36
.28
.11
.08
.12
.17
33.94 (4.38)
35.0 (5.51)
24.0043.00
23.0045.00
.14 (.50)
.86 (.50)
.19
Specic cognitive components

Text memory
Text inferencing
Low-know. integration
High-know. integration
Low-know. access
High-know. access
Composite word decoding
Knowledge about learning
Epistemic belief
Note. For the composite measure of word decoding, higher scores are considered better scores. There were 70 males and 70 females. A positive Cohen's d indicates an effect size favoring
females, whereas a negative Cohen's d indicates an effect size favoring males.

Table 4
Correlations between reading comprehension and specic cognitive components as a
function of gender.
Text memory
Text inferencing
Low-knowledge integration
High-knowledge integration
Low-knowledge access
High-knowledge access
Lower-level word decoding
Epistemic belief of learning
Reading comprehension
Females
Males
.53
.47
.46
.52
.24
.35
.33
.26
.30
.30
.35
.61
.33
.15
.48
.45
Note. There were 70 females and 70 males.

Indicates p b .05.
gender differences in the ranges of scores or in the top tail of the distribution for text inferencing.
The second difference was a marginally signicant gender difference
in low-knowledge integration, t (138) = 1.68, p = .09, and d = 0.28
(for 1-tailed test, t (138) = 1.68, p b .05), such that males performed
better than females, 19.96 versus 20.70 respectively. Again, subsequent
variability analysis revealed that there were more females than males in
the bottom tail of the distributions for low-knowledge integration,
71.00% versus 29.00% respectively. However, there were no gender
differences in the ranges of scores or in the top tail of the distribution.
For the remaining four higher-level processes (i.e., text memory,
high-knowledge integration, low-knowledge access, and highknowledge access) there were no gender differences in the means, all
t's b 1.00, all d's b 0.12; all of the ranges were equivalent; and, at best,
there were only a few, very small gender differences in the upper and
lower tails of the distributions. There was also no gender difference in
the means on the composite measure of word decoding (i.e., 119.19
versus 121.00 respectively), t b 1.00, d = 0.17. Nor were there gender
differences in the ranges of scores or in the ratio of females to males
in the lower tail of the distribution for word decoding, 57.00% and
43.00% respectively. However, there were slightly more females than
males in the upper tail of the distribution, 64.00% versus 36.00%
respectively.
Finally, there were no gender differences in the means for epistemic
belief of learning, t (138) = 1.26, p N .20 [d = 0.19] nor were there
gender differences in the ranges of scores for females versus males.
However, whereas there were an equal number of males and females
in the lower tail of the distribution (i.e., 50.00% versus 50.00% respectively), there were fewer females than males in the upper tail of the
distribution (i.e., 14.00% versus 86.00% respectively). Because higher
scores represent beliefs about learning that are more nave, this nding
suggests that males tend to have more nave beliefs about learning than
do females.
3.5. Predictive powers of the specic cognitive components
Tables 4 and 5 report statistics assessing gender differences in the
powers of the specic cognitive components (e.g., higher- and lowerlevel processes and epistemic belief of learning) to predict reading
comprehension performance. Table 4 reports the correlations between
reading comprehension and the specic cognitive components as a
function of gender and Table 5 reports the regression analyses that
use the specic components to predict reading comprehension. Below
the correlational and regression analyses are reported separately.
3.5.1. Correlations
As Table 4 shows, a number of correlations varied as a function of
gender; for example, (i) many of the measures of the higher-level
processes as well as the measure of epistemic belief of learning were
more predictive of reading comprehension ability for females than for
75
Table 5
Regression analyses of reading comprehension as a function of gender.
Variable
R2
R2
(i) Females (n = 70)

2. Word decoding
3. Epistemic belief of learning
4. Text memory
.52
.62
.71
.74
.27
.38
.50
.54
.27
.11
.12
.04
25.01
12.25
15.33
5.79
(ii) Males (n = 70)

1. Word decoding
.61
.66
.37
.43
.37
.07
39.56
7.65
(iii) Males predictors entered in same order as for females in panel (i) (n = 70)
.30
.09
.09
6.53
2. Word decoding
.61
.37
.29
30.59
.66
.43
.06
6.94
4. Text memory
.67
.45
.02
2.39
(iv) Females predictors entered in same order as for males in panel (ii) (n = 70)
1. Word decoding
.48
.23
.23
20.09
.65
.42
.19
22.35
p b .05.
males whereas (ii) the composite measure of word decoding was

more predictive of reading comprehension ability for males than for
females.
Specically, with the exception of the higher-level processes of lowand high-knowledge access all of the correlations between the measures of higher-level processes (i.e., text memory, text inferencing,
low-knowledge integration, and high-knowledge integration) and
reading comprehension were greater for females than for males,
average r = 0.50 versus average r = 0.31 respectively. In addition, the
correlation between the measures of epistemic belief of learning and
reading comprehension was higher for females than males, r = 0.45
versus r = 0.33 respectively. Again, these differences in magnitudes
suggest that at least four of the measures of higher-level processes as
well as the measure of epistemic belief of learning are more predictive
of reading comprehension performance for females than males. Additionally, these gender differences in magnitudes exist despite the lack
of quantiable female advantages in the means and the variabilities
for these ve cognitive processes/components. In fact, as the earlier
analysis revealed, the opposite occurred. That is, males performed
better than females on the measures of text inferencing and lowknowledge integration.
In contrast, as Table 4 shows the correlation between word decoding
and reading comprehension performance was higher for males than for
females, r = 0.61 versus r = 0.48 respectively. This difference in magnitudes suggests that the composite measure of word decoding is more
predictive of reading comprehension performance for males than it is
for females. These gender differences in magnitudes exist despite the
lack of a quantiable male advantage in the means and variabilities of
word decoding.
3.5.2. Regression analyses
Table 5, panels (i) and (ii) report the results of two stepwise regressions, one regression for females (i.e., panel (i)) and the other one for
males (i.e., panel (ii)). In both of these analyses the best predictor
entered the model rst, the second best predictor entered the model
second, and so on.
As these two panels show, there were a number of differences
between the two regression analyses. First, whereas high-knowledge
integration was the best predictor of reading comprehension performance for females (i.e., 27.00%), word decoding was the best predictor
for males (i.e., 37.00%). Second, the combined inuences of the
higher-level processes of high-knowledge integration and text memory
accounted for 31.00% of the variance in reading comprehension
76
performance for females (i.e., panel i), whereas for males, higher-level
processes failed to account for any unique variance in reading comprehension performance once the variance attributed to the rst two
predictors (i.e., word decoding and epistemic belief of learning) was
partialed out, see panel (ii). In fact, as panel (iii) shows, even when
the regression analysis for males was repeated using the same predictors in the same order as they were for females (i.e., panel i), higherlevel processes accounted for only 11.00% of the variance in reading
comprehension performance. This 11.00% of the variance for males is
signicantly lower than the 31.00% variance that was accounted for
when the target gender was females, z = 1.91, p b .02. Third, as panel
(i) shows, for females word decoding accounted for 11.00% of the
variance in reading comprehension performance, whereas as panel
(ii) shows for males it accounted for 37.00%. Indeed, as panel (iv)
shows, even when the regression analyses for females was repeated
using the same predictors in the same order as they were for males
(i.e., panel ii), word decoding accounted for only 23.00% of the variance
in reading comprehension performance; a percentage that is signicantly lower than the 37.00% observed for the males, z = 1.66, p b .05.
Finally, whereas epistemic belief of learning accounted for 12.00% of
the variance in reading comprehension performance for females, for
males it accounted for only 7.00% of the variance, see panel (i) versus
panel (ii) respectively.
4. Discussion
In the domain of verbal abilities, studies suggest a female advantage
on measures of spelling, word uency, and language production
(Halpern, 2000) and a male advantage on measures of verbal reasoning
(e.g., Colom et al., 2004). Few if any gender differences, however, exist
on measures of vocabulary and adult reading comprehension ability
(Halpern et al., 2007). Furthermore, little is known about gender differences in the specic cognitive components that predict adult reading
comprehension ability. That is, are there gender differences in the specic cognitive components of text memory, text inferencing, knowledge
integration, knowledge access, and/or lower-level word decoding? The
present study begins to address this issue by: (i) determining whether
gender differences exist in the specic cognitive components typically
tapped by measures of adult reading comprehension and by (ii) examining whether gender differences exist in the powers of these cognitive
components to predict adult reading comprehension ability.
The results revealed both quantitative and qualitative gender differences. Specically, there were two small quantitative gender differences
in the measures of text inferencing and low-knowledge integration.
However, there were few, if any, gender differences in the remaining
six cognitive components (i.e., text memory, high-knowledge integration, low-knowledge access, high-knowledge access, word decoding,
and epistemic belief of learning). In addition, the results of the regression analyses revealed that the measures of high-knowledge integration,
text memory, and epistemic belief of learning were more predictive of
reading comprehension performance for females than males, whereas
the composite measure of word decoding was more predictive of
reading comprehension performance for males than females. Below
is a detailed discussion of these results, their implications, and their
limitations.
4.1. Are there gender differences in the measures of specic cognitive
components?
The rst goal of the present study was to determine whether gender
differences existed in the specic cognitive components typically
tapped by measures of adult reading comprehension ability. In order
to address this goal, gender differences in each of the cognitive components (i.e., text memory, text inferencing, low-knowledge integration,
high-knowledge integration, low-knowledge access, high-knowledge
access, word decoding, and epistemic belief of learning) were assessed
using t-tests, ranges, upper/lower tails of the distributions, and Cohen's

d's. The results revealed small male advantages in the means of text
inferencing and low-knowledge integration, d's = 0.36 and 0.28
respectively. There were also more females than males in the bottom
tails of the distributions for both of these components, 79.00% versus
22.00% and 71.00% versus 29.00% respectively. However, no quantiable
gender differences were observed in the remaining six components:
text memory, high-knowledge integration, low-knowledge access,
high-knowledge access, word decoding, and epistemic belief of learning. Rather, the only other gender difference that existed was one of variability inasmuch as there were more males than females (i.e., 86.00%
versus 14.00%) in the upper distribution of epistemic belief of learning.
Because no published study has examined whether gender differences exist in the cognitive components that are typically tapped by
measures of adult reading comprehension ability, the nding that
there are male advantages for text inferencing and low-knowledge integration is a novel one. Furthermore, in the context of theories of reading
comprehension, text inferencing and low-knowledge integration are
particularly important because they are necessary for establishing
coherence in a text (Kintsch, 1988, 1998). Nevertheless the nding of
male advantages on text inferencing and low-knowledge integration
is consistent with two previous ndings, namely the nding that
measures of text inferencing and knowledge integration are predictive
of measures of deductive and analytic reasoning (e.g., Hannon &
Daneman, 2001a) and the nding that there are male advantages on
measures of deductive and analytic reasoning (e.g., Colom et al., 2004).
On the other hand, the nding of a male advantage for text
inferencing is inconsistent with Clinton et al. (2012) and Seipel et al.
(2012) who showed a female advantage for re-instatement and
episodic-based connective inferences using a group of grade four
children. Although the exact source of this discrepancy is unknown,
task/population differences do suggest some possibilities. For example,
Clinton et al.'s and Seipel et al.'s task used talk aloud protocols, whereas
the present task used truefalse statements. However, because talk
aloud protocols rely heavily on word uency and language production,
two cognitive components that have female advantages (e.g., Halpern,
2000), quite possibly Clinton et al.'s and Seipel et al.'s results are a
consequence of gender differences in the cognitive components used
to perform talk aloud protocols rather than gender differences in inferences. In addition, Clinton et al.'s and Seipel et al.'s tasks assessed
episodic memory, whereas the present inferential task assessed a mix
of episodic and semantic memory. However, because research suggests
a female advantage for episodic memory (e.g., Halpern, 2000) quite possibly Clinton et al.'s and Seipel et al.'s results are partially a consequence
of gender differences in the type of memory they assessed (i.e., episodic
memory) rather than gender differences in connective inferences.
Finally, differences in reasoning abilities between grade four children
and adults (i.e., population differences), might explain the discrepancy
between Clinton et al.'s and Seipel et al.'s studies and the present
one. Currently, there is no direct evidence for developmental trends/
gender trends in reasoning abilities. However, research examining
developmental/gender trends in quantitative abilities provide some
indirect evidence given that these trends are frequently explained in
terms of reasoning abilities (Halpern et al., 2007). Specically, whereas
there is no evidence of gender differences in quantitative skills during
primary- and early secondary-school years, a male advantage for mathematical concepts emerges in higher secondary-school grades (Halpern
et al., 2007). In terms of reasoning abilities these ndings suggest that
although there is no evidence for gender difference in reasoning abilities
prior to adulthood, a male advantage emerges in early adulthood; a
nding that is consistent with the present ndings of a male advantage
in text inferencing. Of course, whereas this population difference
explains the results of the present study, it does not explain the female
advantage for grade four children observed by Clinton et al. and Seipel
et al. Nevertheless, the two earlier task-based explanations do. Thus,
perhaps a combination of explanations accounts the discrepancy
between Clinton et al.'s and Seipel et al.'s results and those of the
present study. Regardless of the explanation, it is certainly clear that
there is a need to examine developmental trends in gender differences
in the components of reading comprehension ability.
In addition, the nding that there were more males than females
(i.e., 86.00% versus 14.00%) in the upper distribution of epistemic belief
of learning is also a novel one. Because higher scores on the measure of
epistemic belief of learning indicate more naive beliefs, this difference
suggests that more males have more naive beliefs about learning than
females do. Although the exact reason for this difference is unclear
the present nding is consistent with research that suggests gender differences in GPA are eliminated when the variance associated with learning strategies and metacognitive skills is partialed out (e.g., Grifn,
MacKewn, Mose, & Van Vuren, 2012).
In the context of gender differences, the present ndings are related
to a recent study by Colom et al. (2004). In their study, Colom et al. observed a male advantage on both a measure of dynamic visuospatial
ability (perceiving and extrapolating real motion such as predicting
trajectories of moving objects) and a measure of verbal reasoning
(i.e., combining the relations among terms, A N B: B N C, and then inferring the relationship between A and C). More interesting though, they
observed that the male advantage in verbal reasoning was eliminated
when the gender difference in dynamic visuospatial performance was
statistically removed. Given this nding and the fact that measures of
verbal reasoning are predicted by measures of text inferencing and
knowledge integration (e.g., Hannon & Daneman, 2001a), it is possible
that the male advantages for text inferencing and knowledge integration that were observed in the present study were also a consequence
of gender differences in dynamic visuospatial ability. Future research
might wish to investigate this possibility.
Finally, the male advantages for text inferencing and low-knowledge
integration and the lack of gender-specic advantages for the other six
cognitive components support the gender similarity hypothesis. According to this hypothesis, males and females are similar on most, but not
all, psychological variables (Hyde, 2005, p. 581). Consistent with this
view, although small quantiable gender differences were observed in
the means and the tails of the distributions for two cognitive components (e.g., text inferencing and low-knowledge integration), few if
any gender differences were observed for the other six cognitive
components (e.g., text memory, high-knowledge integration, lowknowledge access, high-knowledge access, word decoding, and epistemic belief of learning). In other words, consistent with the gender
similarity hypothesis, six of the eight cognitive components of reading
comprehension that were assessed in the present study revealed little
evidence of gender differences. See also Clinton et al. (2012) for a similar
nding when the target population is children.
4.2. Do the predictive powers of the specic cognitive components vary as a
function of gender?
A second goal of the present study was to determine whether gender
differences exist in the predictive powers of the specic cognitive components typically tapped by measures of adult reading comprehension
ability. In order to address this goal, a number of regression analyses
predicting reading comprehension performance were completed. The
results revealed signicant gender differences in the combined predictive powers of two measures of higher-level processes (i.e., text memory and high-knowledge integration) as well as the composite measure
of word decoding. Specically, whereas for females the measures of
text memory and high-knowledge integration accounted for 31.1% of
the variance in performance on the composite measure of reading
comprehension ability, for males these same two measures accounted
for only 10.8% of variance. This gender difference in the predictive
powers of the higher-level processes was not a consequence of different
orders of the predictors in the regression models for females versus
males. Indeed as noted in the Results section, the order of the predictors
77
was identical for females and males in this comparison. Furthermore,

this difference in predictive powers was not a consequence of gender
differences in the means or the variabilities of the two higher-level processes for females versus males; as the analysis of the gender differences
in the measures of the cognitive components revealed, no such gender
differences existed. Rather, it seems that this difference in predictive
powers was a consequence of qualitative differences as to how females
versus males use higher-level processes as they complete measures of
adult reading comprehension ability.
In addition, for males the composite measure of word decoding
accounted for 37.00% of the variance in performance on the composite
measure of reading comprehension ability, whereas for females this
same word decoding measure accounted for only 23.00% of the variance. Again, this difference in the relative predictive power of word
decoding was not a consequence of different orders of the predictors
in the regression models because as noted in the Results section the
order of the predictors was identical for females and males in this comparison nor was this difference in predictive power a consequence of
gender differences in the means or variabilities of word decoding for
males versus females. Indeed, the analysis of gender differences in the
measures of the cognitive components revealed no such differences.
Rather, it seems that this difference in predictive power was a consequence of qualitative differences as to how males versus females use
word decoding processes as they complete measures of adult reading
comprehension ability.
The fact that the present study observed qualitative differences as to
how males versus females complete measures of adult reading comprehension is an uncommon nding in the context of gender differences
but it is not a unique one. In a series of studies, Gallagher and colleagues
(Gallagher & DeLisi, 1994; Gallagher, Levin, & Cahalan, 2002) examined
the cognitive patterns of males and females on GRE math problems. As
expected, they observed the typical male advantage on GRE math problems. However more interestingly, they observed a male advantage on
math problems only when math problems required a spatially-based
solution strategy; when the math problems required a more verballybased solution no gender differences existed. Thus the studies of
Gallagher and colleagues and the present one both show qualitative
differences as to how males versus females use of cognitive abilities/
processes to complete a task. The two major differences between their
studies and the present one are: (i) the use of different methodologies
and (ii) whereas Gallagher and colleagues reported qualitative gender
differences in strategies for solving math problems, the present study
observed qualitative gender differences in the use of specic cognitive
components while reading a measure of global reading comprehension.
Finally, the ndings that there are gender differences in the predictive powers of the specic cognitive components typically tapped
by measures of adult reading comprehension ability contrasts with
the ndings of Hannon (2012b). In Hannon's study, test anxiety and
performance-avoidance goals accounted for gender differences in the
SAT-V, rather than cognitive measures, such as higher-level processes
or word decoding processes. Future research might explore the possible
explanations for these differences in measures of adult verbal ability.
4.3. Other ndings of interest

There are also other ndings that are not the primary focus of the
present study but still warrant mention. For example, the present
study observed no gender differences in either the separate or composite measures of adult reading comprehension ability. These ndings are
consistent with previous ndings that also suggest a lack of gender difference in adult reading comprehension performance (e.g., Hyde & Linn,
1988). The present study also replicated previous ndings concerning
both the existence of and non-existence of gender differences in quantitative and visuospatial abilities (e.g., Halpern, 2000; Halpern et al.,
2007; Hyde & McKinley, 1997).
78
In addition, the present study supports the cognitive-processes

approach for examining gender differences (see Halpern, 2004). According to this approach, rather than understand gender differences in
cognitive abilities in terms of the classication of a task (e.g., verbal,
quantitative, visuospatial), gender differences in cognitive abilities are
understood in terms of cognitive processes (Halpern, 2004). Consistent
with this approach, the present study examined gender differences in
the cognitive components of reading comprehension. Moreover, the
results of the present study also suggest that measures of complex constructs that are composed of multiple cognitive processes, like reading
comprehension ability, may not reveal gender differences even though
measures of its cognitive components do. Given the success of the
present study, future research might examine gender differences in
the cognitive components of other tasks, such as intelligence, math
problems, and dynamic visuospatial ability.
Finally, the present study informs future gender research in at least
two important ways. First, the present study revealed that even if a
variable has a quantiable gender difference that is observed in the
means, variabilities, or tails of the distributions, this same variable
may not reveal gender differences in its power to predict another
variable. For example, although the present study revealed that males
performed signicantly better than females on text inferencing, text
inferencing was not more predictive of adult reading comprehension
ability for males than females. Rather, the opposite was true. Text
inferencing was more predictive of reading comprehension performance for females than for males. Second, the present study revealed
that even if a variable reveals no observable quantiable gender difference in the means, variabilities, or tails of the distributions, this same
variable may reveal gender differences in its power to predict another
variable. For example, although the present study revealed no gender
difference in the means, variabilities, or tails of the distribution for
word decoding, word decoding was more predictive of reading comprehension ability for males than for females.
4.4. Limitations
The present study also had a number of limitations that should be
mentioned. First, the present ndings were based on data for one
group of students at one university. Thus it is possible that the observed
gender differences might be larger or smaller if other populations were
targeted. The present ndings are also limited by the measures that
were selected. There are other measures that assess the higher- and
lower-level processes that were targeted in the present study. Similarly,
there are other cognitive components of reading comprehension that
could have been assessed; for example learning strategies, bridging
inferences, thematic inferences, or predictive inferences. Each of these
limitations should be investigated in future research.
4.5. Conclusions
In conclusion, the present study examined: (i) whether gender
differences exist in the specic cognitive components that are typically
tapped by measures of adult reading comprehension as well as
(ii) whether gender differences exist in the powers of these cognitive
components to predict performance on measures of adult reading
comprehension ability. The results revealed male advantages for the
cognitive components of text inferencing and low-knowledge integration. The results also revealed a qualitative difference in the powers of
some of the cognitive components to predict adult reading comprehension ability. Specically, whereas high-knowledge integration and text
memory were more predictive or reading comprehension performance
for females than males, the composite measure of word decoding was
more predictive of reading comprehension performance for males
than females.
References
Baddeley, A., Logie, R., Nimmo-Smith, I., & Brereton, N. (1985). Components of uent
reading. Journal of Memory and Language, 24, 119131.
Bell, L. C., & Perfetti, C. A. (1994). Reading skill: Some adult comparisons. Journal of
Educational Psychology, 86, 244255.
Brown, J. I., Bennett, J. M., & Hanna, G. (1981). NelsonDenny reading test. Chicago:
Riverside.
Cain, K., Oakhill, J., & Bryant, P. (2004). Children's reading comprehension ability: Concurrent
prediction by working memory, verbal ability and component skill. Journal of
Clinton, V., Seipel, B., van den Broek, P., McMaster, K., Kendeou, P., Carlson, S., et al. (2012).
Gender differences in inference generation by fourth grade students. Journal of
Research in Reading, 00, 118, http://dx.doi.org/10.1111/j.1467-9817.2012.01531.x.
Cohen, J. (1988). Statistical power analysis for the behavioural sciences (2nd ed.). New
York: Academic Press.
Colom, R., Contreras, M. A., Arend, I., Leal, O. G., & Santacreu, J. (2004). Sex differences in
verbal reasoning are mediated by sex differences in spatial ability. The Psychological
Record, 54, 365372.
Cunningham, A. E., Stanovich, K. E., & Wilson, M. R. (1990). Cognitive variation in adult
college students differing in reading ability. In T. H. Carr, & B. A. Levy (Eds.), Reading
and its development: Component skills approaches (pp. 129159). San Diego, CA:
Academic Press (xii).
Daneman, M. (1991). Individual differences in reading skill. In R. Barr, M. L. Kamil, P.
Mosenthal, & P. D. Pearson (Eds.), Handbook of reading research, Vol. 2. (pp.
512538). New York: Longman.
Daneman, M., & Hannon, B. (2001). Using working memory theory to investigate the
construct validity of multiple-choice reading comprehension tests such as the SAT.
Journal of Experimental Psychology. General, 130, 208223.
Daneman, M., & Tardif, T. (1987). Working memory and reading skill re-examined. In M.
Coltheart (Ed.), Attention and performance XII: The psychology of reading (pp.
491508). Hove: Erlbaum.
Department for Children, Schools and Families (DCSF) (2008a). National curriculum
assessments at key stage 1 in England, 2008. http://www.dcsf.gov.uk/rsgateway/
DB/SFR/s000806/SFR_21_2008.pdf (accessed January 15, 2013)
Department for Children, Schools and Families (DCSF) (2008b). National curriculum
assessments at key stage 2 in England, 2008 (provisional). http://www.dcsf.gov.uk/
rsgateway/DB/SFR/s000805/DCSFSFR202008_PDF.pdf (accessed January 15, 2013)
Department for Children, Schools and Families (DCSF) (2008c). National curriculum
assessments at key stage 2 in England, 2008 (provisional). http://www.dcsf.gov.uk/
rsgateway/DB/SFR/s000804/DCSFSFR192008_PDF.pdf (accessed January 15, 2013)
Gallagher, A., & DeLisi, R. (1994). Gender differences in Scholastic Aptitude Test
Mathematics problem solving among high-ability students. Journal of Educational
Psychology, 86, 204211.
Gallagher, A., Levin, J., & Cahalan, C. (2002). GRE research: Cognitive patterns of gender differences in mathematics admissions test. ETS Report 0219. Princeton, NJ: Educational
Testing Service.
Graesser, A. C., Singer, M., & Trabasso, T. (1994). Constructing inferences during narrative
text comprehension. Psychological Review, 101, 371395.
Grifn, R., MacKewn, A., Mose, E., & Van Vuren, K. W. (2012). Do learning and study skills
affect academic performance? An empirical investigation. Contemporary Issues in
Education Psychology, 5, 109116.
Guillem, F., & Mograss, M. (2005). Gender differences in memory processing: Evidence
from event-related potentials to faces. Brain and Cognition, 57, 8492.
Halpern, D. F. (2000). Sex differences in cognitive abilities. Mahwah, NJ: Lawrence Erlbaum
Associates.
Halpern, D. F. (2004). A cognitive-process taxonomy for sex differences in cognitive abilities.
Current Directions in Psychological Science, 13, 135139, http://dx.doi.org/10.1111/j.
0963-7214.2004.00292.x.
Halpern, D. F., Benbow, C. P., Geary, D. C., Gur, R. C., Shibley Hyde, J., & Gensbacher, M. A.
(2007). The science of sex differences in science and mathematics. Psychological
Science in the Public Interest, 8, 150.
Hannon, B. (2012a). Understanding the relative contributions of lower-level word
processes, higher-level processes, and working memory to reading comprehension performance in procient adult readers. Reading Research Quarterly, 47,
125152.
Hannon, B. (2012b). Test anxiety and performance-avoidance goals explain gender differences in SAT-V, SAT-M, and overall SAT scores. Personality and Individual Differences,
53, 816820.
Hannon, B., & Daneman, M. (1998). Facilitating knowledge-based inferences in lessskilled readers. Contemporary Educational Psychology, 23, 149172.
Hannon, B., & Daneman, M. (2001a). A new tool for measuring and understanding individual differences in the component processes of reading comprehension. Journal of
Hannon, B., & Daneman, M. (2001b). Susceptibility to semantic illusions: An individualdifferences perspective. Memory and Cognition, 29, 449461.
Hannon, B., & Daneman, M. (2006). What do tests of reading comprehension ability such
as the VSAT really measure?: A componential analysis. In A. V. Mittel (Ed.), Focus on
educational psychology (pp. 105146). New York, NY: Nova Science Publishers.
Hannon, B., & Daneman, M. (2007). Prospective memory: The relative effects of encoding,
retrieval, and the match between encoding and retrieval. Memory, 15, 572604.
Hannon, B., & Daneman, M. (2009). Age-related changes in reading comprehension: An
individual-difference perspective. Experimental Aging Research, 35, 432456.
Hannon, B., & Frias, S. (2012). A new measure for assessing the contributions of higherlevel processes to language comprehension in preschoolers. Journal of Educational

Hannon, B., & NcNaughton-Cassill, M. (2011). SAT performance: Understanding the
contributions of cognitive/learning and social/personality factors. Applied Cognitive
Howell, D. C. (1997). Statistical methods for psychology (4th ed.). New York: Duxbury
Press.
Hyde, J. S. (2005). The gender similarities hypothesis. American Psychologist, 60, 581592.
Hyde, J. S., & Linn, M. C. (1988). Gender differences in verbal ability: A meta-analysis.
Psychological Bulletin, 104, 5369.
Hyde, J. S., & McKinley, N. M. (1997). Gender differences in cognition: Results from
meta-analyses. In P. J. Caplan, M. Crawford, J. S. Hyde, & J. R. E. Richardson (Eds.),
Gender differences in human cognition (pp. 3051). New York: Oxford University Press.
Kimura, D. (1999). Sex and cognition. Cambridge, MA: MIT Press.
Kintsch, W. (1988). The role of knowledge in discourse comprehension: A constructionintegration model. Psychological Review, 95, 163182.
Kintsch, W. (1998). Comprehension: A paradigm for cognition. New York, NY: Cambridge
University Press.
Kucera, H., & Francis, W. N. (1967). Computational analysis of present-day American English.
Providence, RI: Brown University Press.
Lim, T. K. (1994). Gender-related differences in intelligence: Application of conrmatory
factor analysis. Intelligence, 19, 179192.
Linn, M. C., & Petersen, A. C. (1985). Emergence and characterization of sex differences in
spatial ability: A meta-analysis. Child Development, 56(6), 14791498.
Logan, S., & Johnston, R. (2010). Investigating gender differences in reading. Educational
Review, 62, 175187.
Long, D. L., Oppy, B. J., & Seely, M. R. (1994). Individual differences in the time course of
inferential processing. Journal of Experimental Psychology. Learning, Memory, and
Cognition, 20, 14561470.
Lynna, R., & Mikk, J. (2009). National IQs predict educational attainment in math, reading
and science across 56 nations. Intelligence, 37, 305310.
Maccoby, E. E., & Jacklin, C. N. (1974). The psychology of sex differences. Stanford, CA:
Stanford University Press.
Mau, W., & Lynn, R. (2001). Gender differences on the Scholastic Aptitude Test, the
American College Test, and college grades. Educational Psychology, 21, 133136.
McNamara, D. S., & Magliano, J. P. (2009). Towards a comprehensive model of comprehension. In B. Ross (Ed.), The psychology of learning and motivation, Vol. 51. (pp.
297384). New York: Elsevier Science.
Ming Chui, M., & McBride-Chang, C. (2006). Gender, context, and reading: A comparison
of students in 43 countries. Scientic Studies of Reading, 10, 331362.
79
Mullis, I. V. S., Martin, M. O., Gonzalez, E. J., & Kennedy, A. M. (2003). PIRLS 2001 international
report: IEA's study of reading literacy achievement in primary schools. Chestnut Hill, MA:
Boston College (Retrieved January 11, 2013, from http://timss.bc.edu/pirls2001.html).
Mullis, I. V. S., Martin, M. O., Kennedy, A. M., & Foy, P. (2007). IEA's progress in international
reading literacy study in primary school in 40 countries. Chestnut Hill, MA: TIMSS &
PIRLS International Study Center, Boston College (Retrieved January 21, 2013, from
http://pirls.bc.edu/pirls2006/intl_rpt.html).
Nelson Education (1983). General aptitude test. Ontario: Scarborough.
Olson, R. K., Kliegl, R., & Davidson, B. J. (1983). Dyslexia and normal readers' eye movements. Journal of Experimental Psychology: Human Perception and Performance, 9,
816825.
Riordan, C. (1999). The silent gender gap: Reading, writing, and other problems for boys.
Education Week, 19(12), 4649.
Rukavina, I., & Daneman, M. (1996). Integration and its effect on acquiring knowledge
about competing scientic theories from text. Journal of Educational Psychology, 88,
272287.
Runyon, R. P., Coleman, K. A., & Pittenger, D. J. (2000). Fundamentals of behavioral statistics
(9th ed.). Boston: McGraw-Hill Higher Education.
Schommer, M. (1990). Effects of beliefs about the nature of knowledge on comprehension.
Journal of Educational Psychology, 82, 498504.
Seipel, B., Clinton, V. E., & Carlson, S. E. (2012). Gender difference in episodic memory
reected in text-connecting inferences poster presented at APS Annual Convention, Chicago,
USA.
Shepard, R. N., & Metzler, J. (1971, February). Mental rotation of threedimensional
objects. Science, 171, 701703.
Singer, M., & Ritchot, K. F. M. (1996). The role of working memory capacity and knowledge access in text inference processing. Memory and Cognition, 24, 733743.
Spelke, E. S. (2005). Sex differences in intrinsic aptitude for mathematics and science?: A
critical review. American Psychologist, 60, 950958.
Sternberg, R. J., & Powell, J. S. (1983). Comprehending verbal comprehension. American
Psychologist, 38, 878893.
Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5th ed.). New York,
NY: Allyn and Bacon.
West, S. G., Finch, J. F., & Curran, P. J. (1995). Structural equation models with non-normal
variables: Problems and remedies. In R. H. Hoyle (Ed.), Structural equation modeling:
Concepts, issues and applications. Thousand Oaks, CA: Sage Publications.
Wood, E. (2003). The power of pupil perspectives in evidence-based practice: The case of
gender and underachievement. Research Papers in Education, 18(4), 365383.

Are There Gender Differences in The Cognitive Components of Adult

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Are There Gender Differences in The Cognitive Components of Adult

Transféré par

Droits d'auteur :

Formats disponibles

Learning and Individual Differences 32 (2014) 6979

Contents lists available at ScienceDirect

Learning and Individual Differences

Are there gender differences in the cognitive components of adult

Nevertheless there are at least two reasons why there is a need to

B. Hannon / Learning and Individual Differences 32 (2014) 6979

knowledge access), lower-level processes, which are used for decoding

Because of this complexity, a few developmental researchers have

B. Hannon / Learning and Individual Differences 32 (2014) 6979

1.2. Summary and present study

Participants completed the following tasks in the following order:

B. Hannon / Learning and Individual Differences 32 (2014) 6979

high-frequency words, or low-frequency words). The frequencies for

B. Hannon / Learning and Individual Differences 32 (2014) 6979

B. Hannon / Learning and Individual Differences 32 (2014) 6979

processes, whereas the term specic cognitive components is used to

Upper (lower) tails

General verbal abilities

Specic cognitive components

B. Hannon / Learning and Individual Differences 32 (2014) 6979

Note. There were 70 females and 70 males.

(i) Females (n = 70)

(ii) Males (n = 70)

males whereas (ii) the composite measure of word decoding was

B. Hannon / Learning and Individual Differences 32 (2014) 6979

using t-tests, ranges, upper/lower tails of the distributions, and Cohen's

B. Hannon / Learning and Individual Differences 32 (2014) 6979

was identical for females and males in this comparison. Furthermore,

4.3. Other ndings of interest

B. Hannon / Learning and Individual Differences 32 (2014) 6979

In addition, the present study supports the cognitive-processes

B. Hannon / Learning and Individual Differences 32 (2014) 6979

Vous aimerez peut-être aussi