The roles of ability, personality, and interests in acquiring current events knowledge: A longitudinal study

Abstract

The purpose of this study was to investigate sources of inter-individual differences in current events knowledge. The study occurred in two sessions. In the initial session, 579 participants completed tests to ability, personality, and interest factors, as well as prior knowledge of current events. Approximately 10 weeks later, participants completed tests to assess new knowledge of current events, acquired since the initial session. Structural equation modeling revealed positive effects of both ability and non-ability factors on prior knowledge, and in turn, a large positive effect of prior knowledge on new knowledge. Results are interpreted in the context of theories of human intelligence that integrate ability and non-ability traits.

Full text

Page 1

The roles of ability, personality, and interests in acquiring current
events knowledge: A longitudinal study☆
David Z. Hambricka,⁎, Jeffrey E. Pinkc, Elizabeth J. Meinzb,
Jonathan C. Pettiboneb, Frederick L. Oswalda
aMichigan State University, United States
bSouthern Illinois University Edwardsville, United States
cUniversity of Virginia, United States
Received 3 July 2006; received in revised form 4 June 2007; accepted 22 June 2007
Available online 1 August 2007
Abstract
The purpose of this study was to investigate sources of inter-individual differences in current events knowledge. The study
occurred in two sessions. In the initial session, 579 participants completed tests to ability, personality, and interest factors, as well as
prior knowledge of current events. Approximately 10 weeks later, participants completed tests to assess new knowledge of current
events, acquired since the initial session. Structural equation modeling revealed positive effects of both ability and non-ability
factors on prior knowledge, and in turn, a large positive effect of prior knowledge on new knowledge. Results are interpreted in the
context of theories of human intelligence that integrate ability and non-ability traits.
© 2007 Elsevier Inc. All rights reserved.
Keywords: higher-level cognition; individual differences; knowledge; abilities; interest; personality
The importance of knowledge has been stressed in a
number of recent definitions of intelligence. Ackerman
(1996) described it as the “central ingredient of adult
intellect” (p. 241), and Schank and Birnbaum (1994)
stated: “The bottom line is that intelligence is a function
of knowledge. One may have the potentiality of intel-
ligence, but without knowledge, nothing will become of
that intelligence” (p. 102). Similarly, Ceci (1996) pro-
posed, “Intelligence is a function of cognitive complex-
ity, which in turn is dependent upon the operation of
cognitive processesonaspecifiableknowledge structure
and, conversely, cognitive processes are dependent upon
the sheer quantity of knowledge a person possesses…”
(p. 27). Why, though, do some people know more than
others? Exposure is an obvious requirement for acquir-
ing knowledge: Toacquire knowledge about some topic,
a person must be exposed to information about that topic
(e.g., Stanovich & Cunningham, 1992, 1993). But why
are some people more likely to seek out knowledge than
others? Or given equivalent exposure, why do some
peopleacquiremoreknowledgethanothers?Simplyput,
what are the factors that contribute, indirectly and
directly, to individual differences in knowledge?
Available online at www.sciencedirect.com
Intelligence 36 (2008) 261–278
☆The authors thank Alison Gillings, Carlee Hawkins, Angella
MacDonald, Kristin Stege, Parth Tikiwala, and Emelia Zerkel for the
help with data collection and entry.
⁎Corresponding author. Department of Psychology, Michigan State
University, East Lansing, MI 48824, United States.
E-mail address: hambric3@msu.edu (D.Z. Hambrick).
0160-2896/$ - see front matter © 2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.intell.2007.06.004

Page 2

1. Perspectives on inter-individual differences in
knowledge
More than a century after Spearman (1904) first
observed it, the question of what psychometric g re-
presents remains unanswered, as does the broader
question of how to define intelligence. Nevertheless,
many theorists have proposed that intelligence com-
prises the ability to learn. For example, in a 1921
symposium (Thorndike, 1921), Woodrow stated, “I
should first say that it is an acquiring capacity” (p. 207),
and Henmon drew a distinction between two factors of
intelligence: “the capacity for knowledge and knowl-
edge possessed” (p. 195). Later, Hebb (1942) distin-
guished between a genetically-based Intelligence A and
an experience-based Intelligence B. Around the same
time, Cattell (1943) proposed a similar view in his
distinction between a general capacity for solving novel
problems–fluid intelligence (Gf)–and the skills ac-
quired through the use of this capacity—crystallized
intelligence (Gc). More recently, in his landmark survey
of factor-analytic studies, Carroll (1993) found evi-
dence for a three-stratum model, with a general factor
(g) at the highest level (Stratum III); eight broad
abilities, including Gf and Gc, at the next level (Stratum
II); and approximately 70 narrow abilities at the lowest
level (Stratum I).
This idea that intelligence and learning are closely
related is supported by a large amount of evidence. In
fact, Jensen (1998) concluded, “there is no general
learning factor…that is independent of psychometric g”
(p. 276). At the level of more specific abilities, Gf and
Gc are typically reported to correlate in the .40 to .60
range (see Horn & Noll, 1998; McGrew, 1997), where
Gf is typically assessed with tests of reasoning and Gc
with tests of vocabulary, comprehension, and general
information. Gc has been observed to correlate espe-
cially highly with knowledge in specific domains. For
example, across four domains, Beier and Ackerman
(2001) found that current events knowledge correlated
.81 with Gc, compared to .45 with Gf. Beier and
Ackerman (2003) observed a similar pattern of results–
Gc correlated much more strongly with knowledge in a
wide range of health-related domains than did Gf–and
Beier and Ackerman (2005) found that Gc was a
stronger correlate of self-directed learning about two
topics (xerography and cardiovascular disease) than was
Gf.
It appears, then, that Gc is predictive of acquiring
knowledge in a given domain, presumably because the
verbal skills and knowledge that Gc comprises have an
impact on how well information is comprehended when
first encountered, how well it is integrated into long-
term memory, or both. However, Gc is certainly not the
only factor that appears to play a role. Another is
knowledge itself. Hambrick (2003), for example, found
that the best predictor of acquiring new knowledge
about college basketball over the course of a season was
prior knowledge of basketball—evidence for a “snow-
ball” effect in knowledge acquisition. There also have
been many reports of prior knowledge of some topic
facilitating retention of new information about that
topic. For example, Spilich, Vesonder, Chiesi, and Voss
(1979) found that prior knowledge of baseball facilitated
retention of information from a passage about a baseball
game (see also Hambrick & Engle, 2002; Recht &
Leslie, 1988; Walker, 1987), and in the study by Beier
and Ackerman (2005) already mentioned, effects of Gc
on learning were mediated through prior topic knowl-
edge (see also Ackerman & Beier, 2006). One way to
think about how prior knowledge might operate directly
to facilitate acquisition of new knowledge stems from
the common view of long-term memory as a network of
interconnected concepts or “nodes” (e.g., Anderson,
1983; McClelland & Rumelhart, 1981): When some
new piece of information is encountered, activation
spreads throughout the network to associatively related
nodes, and the new piece of information becomes part of
the existing knowledge structure through some binding
process.
1.1. The role of non-ability factors
Theory and evidence suggest that non-ability factors
also contribute to knowledge acquisition. In his invest-
ment theory of intelligence, Cattell (1971) conceptual-
ized the growth of Gc in terms of a complex set of
influences, comprising not only Gf, but also personality
and interest traits, which jointly influence how Gf is
“invested” in learning opportunities. Cattell therefore
observed that the correlation between Gf and Gc should
be positive but generally far less than unity. Cattell also
recognized a distinction between a Gc factor comprising
relatively general skills acquired primarily during the
school years and knowledge acquired thereafter: “The
practicing psychologist must realize that crystallized
ability begins after school to extend into Protean forms
and that no single investment such as playing bridge or
skill in dentistry can be used as a manifestation by which
to test all people” (p. 121). In his theory of intelligence-
as-process, personality, interest, and intelligence-as-
knowledge (or PPIK), Ackerman (1996) expanded on
this idea in his distinction between intelligence-as-
process (Gp) and intelligence-as-knowledge (Gk). Like
262D.Z. Hambrick et al. / Intelligence 36 (2008) 261–278

Page 3

Cattell, Ackerman posited both ability and non-ability
influences on the growth of knowledge. However, Gk
is much broader than Gc, encompassing specialized
knowledge acquired through vocational and avocational
activities.
Consistent with these proposals, there are a number
of different sources of evidence for the role of non-
ability factors in knowledge acquisition. At least two
major types of interest can be distinguished. Accord-
ing to Hidi and Renninger (2006), situational interest
is a transient state that is activated in the moment by an
environmental stimulus, and promotes knowledge ac-
quisition in activities like reading by directing
attention toward relevant information and away from
irrelevant information (e.g., Hidi, 1990, 1995; McDa-
niel, Waddill, Finstad, & Bourg, 2000). On the other
hand, individual interest can be thought of as a more
stable predisposition to attend to certain objects or
events, and to engage in activities like reading about a
certain topic (e.g., science). Individual interests have
also been implicated as a contributor to knowledge
acquisition. For example, Ackerman et al. observed
correlations between vocational interest themes, which
might be thought of as broad collections of individual
interests, and knowledge of various domains (e.g.,
Ackerman, 2000; Rolfhus & Ackerman, 1996, 1999),
and Reeve and Hakel (2000) found a positive average
intraindividual correlation between interest and knowl-
edge levels across twelve domains, indicating that
people learn more about what they like. There have
also been a number of reports of individual interest
facilitating learning from text (e.g., Alexander, Jetton,
& Kulikowich, 1995; Alexander, Kulikowich, &
Schulze, 1994), and in a meta-analysis, Schiefele,
Krapp, and Winteler (1992) observed an average cor-
relation of .31 between interest and achievement
across a wide range of academic topics (see also
Tobias, 1994, for a review).
Certain personality characteristics may also contrib-
ute to individual differences in knowledge. In an early
study, Cattell (1947) noted a positive correlation be-
tween Gc and a cluster of personality traits that he
labeled “intellectual/wide interests”. Around the same
time, Gough (1953) described a measure of Intellectual
Efficiency (Ie), which included items like “I read at least
ten books a year” and “I like to read about history”.
Gough reported positive correlations between Ie and a
standard assessment of intelligence, which included Gc-
type measures, as did Gough and Weiss (1981). And,
more recently, in a meta-analysis, Ackerman and
Heggestad (1997) found moderate positive correlations
between Gc and two personality constructs: Openness
to Experience and Typical Intellectual Engagement.
Openness to Experience–a factor of Costa and
McCrae's (1992) big-five personality structure–reflects
intellectual and artistic interests and has been alterna-
tively interpreted as a Culture or Intellect factor (e.g.,
Goldberg, 1999), whereas Goff and Ackerman's (1994)
concept of Typical Intellectual Engagement refers to a
person's “interest in a wide variety of things and…pref-
erence for a complete understanding of a complex topic
or problem…” (p. 539). Finally, there have been a few
reports of positive correlations between Need for
Cognition–one's preference or propensity for intellec-
tual engagement (Cacioppo & Petty, 1982)–and various
Gc measures (e.g., Sadowski & Cogburn, 1997;
Salthouse, Berish, & Miles, 2002). At least in part, all
of these personality characteristics might be thought of
as reflecting a general interest in learning, and it seems
reasonable to suggest that they contribute to acquiring
verbal abilities (Gc) through activities like reading,
which then contribute directly to knowledge acquisition.
General interest in learning may also contribute to
individual differences in knowledge independent of Gc
(or other ability factors). In particular, people with a
strong general interest in learning may be more likely to
develop individual interests in specific topics (e.g., art,
science, politics), and hence to seek information about
these topics, than are people with a weaker interest in
learning.
2. The present study
The evidence just reviewed suggests that individual
differences in knowledge arise from both ability
and non-ability factors. The present research extends
a study by Hambrick, Meinz, and Oswald (2007)
concerning the relative contributions of these factors to
individual differences in knowledge of current events—
the sort of knowledge that may be important for
everyday tasks such as deciding how to vote in an
election or how to invest money in the stock market.
Hambrick et al. distinguished between distal and
proximal predictors of knowledge acquisition. Distal
predictors reflect indirect causes originating from
general dimensions of psychological functioning, such
as intelligence and personality, whereas proximal
predictors reflect more direct causes tied to particular
domains, such as individual interests and exposure.
Hambrick et al. found evidence for two predictive
“pathways”. In the ability pathway, Gf predicted Gc,
which predicted current events knowledge. In the non-
ability pathway, Need for Cognition predicted current
events interest, and current events interest predicted
263D.Z. Hambrick et al. / Intelligence 36 (2008) 261–278

Page 4

news exposure. In turn, news exposure predicted cur-
rent events knowledge.
The major goal of this study was to address two
limitations of the Hambrick et al. (2007) study. The first
limitation is that the predictor variables (ability and non-
ability) and current events knowledge were measured at
the same point in time. Hambrick et al. were therefore
limited in their ability to draw causal conclusions about
theroleofthepredictorvariablesinacquiringknowledge.
The present study used a longitudinal design. The major
question of interest was how ability and non-ability var-
iables, along with preexisting current events knowledge,
would influence acquisition of current events knowledge
over an extended period of time. An advantage of tradi-
tional approaches to research on learning, in which ex-
posure to some information is experimentally controlled
(e.g., in a paired-associates task), is that individual dif-
ferences in learning can be attributed to factors other than
exposure. Nevertheless, a limitation of this approach is
thattheconditionsoflearningmaybedifferentfromthose
encountered outside of the laboratory, where learning
tends to occur over an extended period of time and under
self-paced conditions (Neisser, 1978). Hambrick (2003)
conducted a longitudinal study of the acquisition of
knowledge in the domain of college basketball that in-
volved two test sessions. In the first session, participants
completed tests to assess potential predictors of knowl-
edge acquisition, including ability factors and domain-
specific knowledge and interest. Then, approximately
2.5 months later, these same participants completed tests
of basketball acquired over the just-completed season.
The present studyparallels thisapproachinthedomainof
current events.
The second limitation of the Hambrick et al. (2007)
study is that assessment of potentially relevant predictor
constructs was limited. Two broad abilities were con-
sidered in this earlier study—Gf and Gc. We expanded
the ability assessment to include tests of a third, short-
term memory (Gsm). Gsm might contribute to individual
differences in current events knowledge, because
information must presumably be held in immediate
awareness before it can be transferred to long-term
memory. A specific question was whether Gf, Gc, and
Gsm would contribute incrementally to current events
knowledge, beyond any influence of g. As already
discussed, previous research suggests that Gc should be
highlypredictiveofcurrenteventsknowledge(Ackerman
&Beier,2006;Beier&Ackerman,2001;Hambricketal.,
2007). Furthermore, within the Cattell–Horn–Carroll
(CHC) framework, a number of studies have demonstrat-
ed incremental validity of Gc and Gsm for outcomes like
reading achievement (Evans, Floyd, McGrew, & Lefor-
gee, 2001; McGrew, Keith, Flanagan, & Vanderwood,
1997; Vanderwood, McGrew, Flanagan, & Keith, 2002).
Furthermore, Gustafsson and Balke (1993) found that Gc
added over g to the prediction of course grades across a
wide range of topics (see also Thorndike, 1991; Young-
strom, Kogos, & Glutting, 1999). We also expanded
assessment of personality variables. Hambrick et al. ob-
served a non-ability influence on current events knowl-
edge originating from Need for Cognition. However,
Need for Cognition may be a manifestation of broader
dimensions of personality. In fact, in a longitudinal study
described by McCrae (2000), Openness to Experience
and Need for Cognition correlated strongly (r=.55), and
the correlation was even stronger for the Ideas facet of
Openness (r=.68).
2.1. Sex differences in knowledge
An extensive literature documents a link between sex
differences in ability and non-ability characteristics and
sex differences in educational and career trajectory and
achievement (e.g., Lubinski & Benbow, 1992; see also
Spelke, 2005). In the general spirit of this research,
another goal of this study was to investigate sex differ-
ences in current events knowledge. In a recent study,
Lynn and Irwing (2002) observed higher levels of
general knowledge in males than females (see also
Lynn, Irwing, & Cammock, 2002; Wilberg & Lynn,
1999). This sex difference could not be explained by
either Gf or an experience factor assumed to reflect
acculturation. However, the assessment of experience in
this study was very limited, as it was based on only two
indicators (father's occupation and father's profession).
Ackerman, Bowen, Beier, and Kanfer (2001) assessed a
broader range of ability, personality, and interest factors,
and they found evidence that these factors contributed
to, but could not completely explain, sex differences in
knowledge favoring males. In another relevant study,
Hambrick (2003) found that although males tended to
acquire much more knowledge about basketball over the
course of a college season than females, there were no
direct effects of sex on basketball knowledge, in part due
to a mediating effect of basketball interest and exposure.
A specific question of interest here was whether sex
differences in current events knowledge, if observed,
would be accounted for by sex differences in current
events interest.
2.2. Overview
To recap, the major goal of the present study was
to investigate ability and non-ability influences on
264 D.Z. Hambrick et al. / Intelligence 36 (2008) 261–278

Page 5

acquisition of current events knowledge. A specific
question of interest was how the ability and non-
ability factors, along with preexisting knowledge,
would impact acquisition of new knowledge. Another
question was whether sex differences in current
events interest would contribute to sex differences in
acquisition of current events knowledge. To address
these questions, we conducted a study involving
over 500 participants. The two sessions of the study
were separated by approximately 10 weeks between
September, 2004, and December, 2004. In Session 1,
participants completed tests and questionnaires
designed to measure ability, personality, interest,
and exposure factors. Participants also completed a
test to assess current events knowledge from the
previous two years (2002 and 2003). In Session 2,
participants completed additional tests to assess
knowledge of events that occurred in the period
since Session 1. Data were analyzed using structural
equation modeling.
3. Method
3.1. Participants
Participants were recruited from Michigan State
University and Southern Illinois University Edwards-
ville. Five-hundred seventy-nine participants (74%
female) completed Session 1, of which 536 (92.6%)
returned for Session 2. Participants received credit in
an introductory psychology course for volunteering
their time. We assume that the range of cognitive
ability in this sample was restricted, given that
all participants were college students. However, there
was still a relatively wide range of ability in our
sample, as ACT scores ranged from 13 to 33 (M=23.7,
SD=3.3).1Therefore, it appears that our sample
was selective, but not extremely so relative to all
students who apply to college (M=20.8; SD=4.8; see
www.act.org).
3.1.1. Session 1
3.1.1.1. Procedure and materials.
in September, 2004, and lasted approximately 2.5 h;
participants were tested in groups (up to 70). After
signing an informed consent form, participants complet-
ed a background questionnaire that asked for age, sex,
ethnicity, and ACT score. Next, participants completed a
battery of materials in a fixed order of (1) News Exposure
Questionnaire; (2) Need for News; (3) Thinking Disposi-
tions;(4) News Headlines;(5) WordSpan,(6)DigitSpan,
(7) Letter Span, (8) Matrix Reasoning; (9) Reading
Comprehension; (10) Letter Sets; (11) Vocabulary;
(12)SeriesCompletion;(13)LetterComparison;(14)Pat-
tern Comparison; (15) Number Comparison; (16) Current
Events Knowledge.2
3.1.1.1.1. Cognitive ability.
ed tests to measure four ability constructs: fluid
intelligence (Gf), crystallized intelligence (Gc), and
short-term memory (Gsm). Where noted, items were
drawn from the following batteries: Air Force Officer
Qualifying Test (Berger, Gupta, Berger, & Skinner,
1990); Educational Testing Service Kit of Factor-
Referenced Tests (Ekstrom, French, Harman, & Dar-
man, 1976); and the Shipley Institute for Living Scale
(Zachary, 1986).
Session 1 occurred
Participants complet-
Gf—(1) Series Completion (from Zachary, 1986):
Each item consisted of a sequence of letters,
numbers, or both, followed by blank spaces; the
task was to fill in the blank spaces with the logical
continuation of the sequence. Four minutes were
allowed for 20 items, and participants marked
responses on the test form. (2) Letter Sets (from
Ekstrom et al., 1976): Each item consisted offive sets
of letters; the task was to infer the rule that made
these letter sets similar and to identify the letter set
that did not fit this rule. Eight minutes were allowed
for 14 items, and participants marked responses on a
scantron. (3) Matrix Reasoning: Items in this test
were drawn from Raven's Advanced Progressive
Matrices (Raven, 1962). Each item consisted of a
3×3 matrix in which each cell except the one in the
1The ACT is a widely used college entrance exam in the United
States and is designed to assess broad academic achievement. We
were not able to verify the self-reported ACTscores against university
records. However, there is no reason to suspect that there was
significant degree of misreporting, given that ACT had a high positive
correlation with the two other Gc measures, Synonym Vocabulary
(r=.56) and Reading Comprehension (r=.64). Seventeen participants
reported an SAT score instead of an ACT score; we converted these
SAT scores into ACT scores using a table published by the College
Board (see www.act.org).
2Given that the period between Session 1 and Session 2
encompassed the final stages of the 2004 U.S. general election,
participants completed additional tests and questionnaires to assess
politics interest and exposure, and politics items were overrepresented
on the test of new current events knowledge (i.e., 30 items). Results
pertaining to these materials are to be described in a separate report
focusing specifically on individual differences in politics knowledge.
265 D.Z. Hambrick et al. / Intelligence 36 (2008) 261–278

Page 6

lower right-hand corner contained a pattern; the task
was to choose from among eight alternatives a
pattern that made logical sense for the missing ninth
cell. Eight minutes were allowed for 14 items, and
participants marked responses on a scantron. For
each test, a participant's score was the proportion
correct.
Gc—(1) Synonym Vocabulary(from Zachary,1986):
Each item in this test consisted a target word, printed
in all capitals, followed by five lettered alternatives;
the task was to choose the alternative that was similar
in meaning to the target word. Five minutes were
allowed for 15 items, and participants marked
responses on a scantron. (2) Reading Comprehension
(from Berger et al., 1990): Items in this test consisted
of a short paragraph; the task was to select the
alternative(fromamongfour)thatcompletedthefinal
sentence. Six minutes were allowed for 10 items, and
participants responded on a scantron form. For each
test, a participant's score was the proportion correct.
Gsm—(1) Word Span (from Kane et al., 2004):
Participants attempted to recall sequences of com-
mon nouns, presented in all capital letters. Set sizes
ranged from four to nine words, with one trial for
each set size. (2) Digit Span (from Kane et al.):
Participants attempted to recall sequences of digits
(0–9). Set sizes ranged from five to nine digits, with
one trial for each set size; digits repeated across but
not within trials. (3) Letter Span (from Kane et al.):
Participants attempted to recall letters. Set sizes
ranged from five to nine letters, with one trial for
each set size; letters repeated across but not within
sets. Nine letters were used (B, F, H, J, L, M, Q, R,
X). For each test, the stimuli were presented on a
projection screen for 1 s each, with a 1 s blank
screen between items. After presentation of the final
item in a set, a prompt (RECALL) appeared in the
center of the screen, and participants attempted to
recall the items (words, digits, or letters) in the order
presented. For each test, a participant's score was
the number of items recalled in the correct serial
order.
3.1.1.1.2. Intellectual openness.
gree of intellectual openness with a scale (Thinking
Dispositions) that included items drawn from two
sources. One source (18 items) was the Need for
Cognition scale (Cacioppo, Petty, & Kao, 1984). The
other source (10 items) was the Intellect facet scale from
the International Personality Item Pool (2001), which
appears to measure the same construct as the Ideas facet
scale of the NEO-PI-R (Goldberg, 1999). Each item was
We assessed de-
a statement describing an attitude toward or propensity
for intellectual engagement (e.g., I prefer complex to
simple problems). Using a five-point scale with anchors
of 1 (Completely Inaccurate) and 5 (Completely
Accurate), participants were to assign a value to each
item reflecting the degree to which they believed the
statement was an accurate description of them. Half of
the items were positively worded, and half are nega-
tively worded. There was no time limit; most partici-
pants finished within 10 min. Participants marked
responses on a scantron form. For each scale, a par-
ticipant's score was the average rating.
3.1.1.1.3. News exposure.
to news media with a paper-and-pencil questionnaire
(News Exposure Questionnaire from Hambrick et al.,
2007). Participants were to estimate for a typical week
the number of times they engaged in each of following
five activities and how long they spent on each activity:
(1) reading the newspaper, (2) reading news magazines,
(3) watching news programs on television, (4) listening
to news programs on the radio, and (5) reading the news
on the Internet. For each activity, the exposure estimate
was the frequency estimate multiplied by the time
estimate (i.e., minutes per week). Participants were also
asked to list the newspapers, news magazines, news
programs, and news websites they rely upon most.
As an additional assessment of news exposure,
participants completed a newly developed scale (Need
for News, from Hambrick et al., 2007) in which each of
the 12 items was a statement describing an attitude
toward or propensity for seeking news (e.g., It bothers
metofeellikeI'mbehindonthenews).Usingafive-point
rating scale with endpoints of 1 (Completely Inaccurate)
and 5 (Completely Accurate), participants were to rate
the degree to which each statement described them. A
participant's score was the average rating.
3.1.1.1.4. Current events interest.
interest in current events using an inventory (News
Headlines) in which each item was a news headline
from one of the following seven categories: (1) Arts/
Entertainment, (2) Business/Economy, (3) Crimes/
Accidents/Disasters, (4) U.S. Politics/Government,
(5) World Politics/Government, (6) Science/Medicine,
and (7) Sports/Recreation. There were nine items for
each category, for a total of 63 items. For each item,
using a five-point scale with endpoints of 1 (Not at all
Interested) and 5 (Very Interested), participants were to
indicate how interested they would be in reading or
hearing about the news story. The items were selected
from a larger pool of items administered by Hambrick
et al. (2007). For each category, a participant's score was
the average rating.
We assessed exposure
We assessed
266 D.Z. Hambrick et al. / Intelligence 36 (2008) 261–278

Page 7

3.1.1.1.5. Current events knowledge.
prior (pre-study) knowledge of current events with a
multiple-choice test consisting of questions about news
stories from 2002 and 2003. For each year, there were
three questions for each of the seven current events
categories listed above (for a total of 42 items). The
items were selected from a larger pool of items ad-
ministered by Hambrick et al. (2007). For each category,
a participant's score was the proportion correct of items
attempted. Participants were encouraged to answer all of
the questions, even if they had to guess. There was no
time limit; most participants finished within 10 min.
We assessed
3.1.2. Session 2
3.2.1.1. Procedure and materials.
in December, 2004, and lasted approximately 45 min;
participants were tested in groups (up to 50). First,
participants completed a new version of the News
Exposure Questionnaire, in which they were asked to
estimate time spent engaging in the news-seeking
activities described previously during the period be-
tween Sessions 1 and 2. Next, participants completed a
multiple-choice test to assess newly acquired current
events knowledge; that is, knowledge acquired since
Session 1. This test consisted of 90 questions about
news stories from the 10-week period between the day
after the final participation date for Session 1 (Septem-
ber 19, 2004) and the day prior to the first participation
date for Session 2 (November 29, 2004). Participants
were encouraged to answer all of the questions, even if
they had to guess. For each category, a participant's
score was the proportion correct of items attempted;
missing values were assigned to participants who re-
sponded to fewer than 80% of items. There was no time
limit; most participants finished within 35 min.
We wrote the current events questions to minimize
the possibility that participants could answer the ques-
tions correctly without having been exposed to the
information between Sessions 1 and 2 and without
guessing. Consider the following item:
In ___, a Laotian man shot and killed six deer
hunters; the shooter alleges that the hunters threatened
him and made racist remarks.
Session 2 occurred
a) Illinois
b) Minnesota
c) North Dakota
d) Wisconsin
The answer is Wisconsin, but each of the other
alternatives is plausible, because deer are legally hunted
in Illinois, Minnesota, and North Dakota. In short, we
cannot completely rule out the possibility that prior
knowledge enabled correct guessing on some of the
items, but it seems reasonable to assume that the
majority of the questions assessed newly acquired
knowledge (see Hambrick et al., 2007, and Hambrick,
2003, for further discussion of this issue). Additional
sample items appear in Table 1.
3.2.1.1. Item selection.
istered knowledge items to a pilot sample, and then
administered the items with the best psychometric
characteristics (e.g., item-total correlations) to a test
sample. However, a constraint of this study is that we
wantedtomaximizetheamountoftimebetweenSessions
1and2foracquisitionofcurrenteventsknowledge,while
leavingenoughtimeattheendofthesemesterforSession
2. Consequently, there was not enough time for a pilot
Ideally, we would have admin-
Table 1
Sample items from Current Events Knowledge test
CategoryItem Alternatives
Arts/
Entertainment
Television host _ was
accused of sexual
harassment by an associate
producer working on his
show.
Which American retail giant
recently announced a merger
with K-mart?
a) Geraldo Rivera
b) Bill O'Reilly⁎
c) Jerry Springer
d) Chris Matthews
Business/
Economy
a) Target
b) Wal-Mart
c) Sears⁎
d) JC Penney
a) Nicole
b) George
c) Lucy
d) Ivan⁎
Crimes/
Accidents/
Disasters
In September, this hurricane
battered the U.S. Gulf Coast,
causing billions of dollars in
damages In Florida and
Alabama:
Following the election, who
did President Bush appoint
Secretary of State?
U.S. Politics/
Government
a) Colin Powell
b) Arlen Specter
c) Condoleeza Rice⁎
d) Orrin Hatch
a) Muamarr Qadaffi
b) Fidel Castro⁎
c) Yassir Arafat
d) Jacques Chirac
a) Vioxx⁎
b) Viagra
c) Zoloft
d) Claritin
World Politics/
Government
Aging world leader _ tripped
and fell as he stepped off a
stage after making a public:
appearance.
This medication was taken
off drugstore shelves by
pharmaceutical company
Merck due to safety concerns
of increased risk of heart
attack and stroke:
A brawl involving both
players and fans erupted in
a pro basketball game
between the Detroit Pistons
and ___.
Science/
Medicine
Sports/
Recreation
a) Indiana Pacers⁎
b) L.A. Lakers
c) Houston Rockets
d) Philadelphia 76ers
Note.⁎=correct alternative.
267D.Z. Hambrick et al. / Intelligence 36 (2008) 261–278

Page 8

study. Therefore, as a compromise, participants complet-
ed all of the items that we generated, but we anticipated
using only a subset of these items in the final analyses
based on psychometric characteristics of the items. Items
were selected in two steps. First, we selected items that
could be clearly classified into one of the seven current
events categories. The classification procedure was as
follows: Using short descriptions of the categories, 10
raters were asked to give each item a primary classifica-
tion, reflecting the most relevant category, and secondary
classificationsifanyothercategorieswererelevant.There
was good agreement (70% or greater) on the primary
classification for 79 of 90 items, leaving at least eight
itemspercategory,exceptBusiness/Economy(sixitems).
Second, for each category, we selected the six items
having the highest correlations with total correct for the
category (i.e., item-total correlations). (We used all
Business/Economy items.) Thus, as for prior knowledge,
there were six items per category, for a total of 42 new
knowledge items.
4. Results
There were statistically significant differences
(pb.01) between participants who returned for Session
2 (n=536) and those who did not (n=51) for only 2 of
39 variables in the data set (i.e., Letter Sets, Science/
Medicine Interest). This small rate of statistical signif-
icance suggests that attrition from Session 1 to Session 2
was random. All subsequent analyses are based on
listwise deletion, including only those participants who
completed both sessions.
4.1. Data preparation
Three steps were involved in preparing the data for
analysis. First, for each scantron-based test, excluding
the (timed) reasoning tests, we assigned missing values
to participants who responded to fewer than 80% of
the items. Second, we discarded participants who were
missing values on two or more variables. This resulted
Table 2
Descriptive statistics for news exposure variables
Session 1 Session 2
M SDSk. Ku.dsex
−.28⁎⁎
−.11
−.41⁎⁎
−.04
−.13
M SD Sk.Ku.dsex
−.37⁎⁎
−.28⁎⁎
−.42⁎⁎
−.13
−.03
Newspaper min/week
Television min/week
Internet min/week
Magazines min/week
Radio min/week
36.5
84.3
33.8
7.9
21.8
47.3
103.6
60.0
21.3
45.3
2.03
2.60
2.56
3.08
2.98
4.44
9.72
6.62
9.02
9.55
44.0
57.7
30.1
3.9
11.4
47.1
80.0
51.6
12.9
30.9
1.57
2.12
2.41
3.78
3.77
2.35
4.36
5.57
13.90
15.08
Note. Because of high skewness and kurtosis values, data for these values were square-root transformed in subsequent analyses. Cohen's d statistics
reflect sex differences; negative values indicate lower averages in females (n=374) than in males (n=144).
⁎⁎pb.01.
Table 3
Correlations among news exposure variables
123456789 10
Session 1
1. Newspaper min/week
2. Television min/week
3. Internet min/week
4. Magazines min/week
5. Radio min/week
−
.29
.24
.26
.10
−
.11
.24
.21
−
.10
.13
−
.12
−
Session 2
6. Newspaper min/week
7. Television min/week
8. Internet min/week
9. Magazines min/week
10. Radio min/week
.59
.24
.29
.16
.08
.16
.53
.13
.24
.18
.21
.10
.65
.08
.01
.20
.20
.13
.46
.07
.10
.22
.10
.18
.44
−
.15
.30
.19
.08
−
.16
.27
.24
−
.10
.05
−
.10
−
Note. All variables are square-root transformed. Correlations with an absolute magnitude greater than .09 are statistically significant (pb.05).
268D.Z. Hambrick et al. / Intelligence 36 (2008) 261–278

Page 9

in elimination of 18 participants. For 39 participants,
each of whom had only one missing value, we regressed
each variable onto other variables of the same type
(ability, interest, etc.) and replaced missing values with
predicted values. Finally, we screened the data for
outliers. We screened the data for univariate outliers on a
variable-by-variable basis, replacing any value greater
than 3.5 SD units from the variable mean (b.5% of the
data) with a less extreme value of 3.5 SD units from the
mean. We then inspected Mahalanobis D2values to
screen the data for multivariate outliers, but there were
no extreme values. The final sample consisted of 518
participants.
4.2. Descriptive statistics
As shown in Table 2, participants reported a wide
range of news exposure. For example, estimates for the
10-week period between Sessions 1 and 2 ranged from 0
to 220 min/week for newspaper reading (M=44.0;
SD=47.1) and from 0 to 369 min/week for television
watching (M=57.7; SD=80.0); on average, males
reported higher levels of news exposure than females
(avg. d=−.22). There also was evidence that that esti-
mates were moderately reliable. Across sessions, cor-
relations between the two estimates for each activity
ranged from .44 to .65, and each variable correlated
positively with Need for News, an independent index of
news exposure (see Table 3): Session 1 (avg. r=.33) and
Session 2 (avg. r=.29). Many exposure variables were
non-normal (skewness valuesN2 and kurtosis val-
uesN10). For this reason, all exposure variables were
square-root transformed for subsequent analyses. Skew-
ness and kurtosis values were in the acceptable range
following transformation.
Descriptive statistics for the ability, non-ability, and
knowledge variables are displayed in Tables 4–6, and a
correlation matrix appears in Table 7. It can be seen in
Table 4 that the reliability estimates were somewhat,
though not extremely, low for some ability variables
Table 4
Descriptive statistics for ability variables
M SD Sk. Ku.αdsex
Gf
Series completion
Letter sets
Matrix reasoning
.75
.79
.59
.10
.20
.20
−.65
−1.12
−.93
1.01
.93
.81
.62
.76
.74
−.18⁎
.07
−.35⁎⁎
Gc
Synonym vocabulary
Reading comprehension
ACT
.59
.53
.17
.24
3.3
.00
−.63
−.77
−.36
.60
.66
−
−.44⁎⁎
−.31⁎⁎
−.46⁎⁎
−.05
−.02 23.7
Gsm
Word Span
Digit Span
Letter Span
.65
.79
.70
.14
.14
.15
−.17
−.48
−.06
−.21
−.30
−.48
.78
.82
.85
−.19⁎
−.15
−.28⁎⁎
Note. Means reflect proportion correct. Cohen's d statistics reflect sex
differences, with negative values indicating lower averages in females
(n=372) than in males (n=144).
⁎pb.05,⁎⁎pb.01.
Table 5
Descriptive statistics for personality and interest variables
M SD Sk. Ku.αdsex
Intellectual openness
Need for cognition
Intellect
3.3
3.5
.6
.6
−.15
−.09
−.01
.00
.87
.78
.17
−.24⁎⁎
Current events interest
Arts/Entertainment
Business/Economy
Crimes/Accidents/Disasters
U.S. Politics/Government
World Politics/Government
Science/Medicine
Sports/Recreation
3.4
2.7
4.6
3.1
2.5
4.4
3.1
1.4
1.1
1.2
1.3
1.1
1.1
1.5
.28
.63
−.74
.31
−.18
−.41
.43
.03
−.47
.93
.89
.92
.92
.92
.86
.93
.56⁎⁎
−.38⁎⁎
.52⁎⁎
−.08
−.38⁎⁎
−.05
−.62⁎⁎
−.36
.39
.87
−.24
.57
Note. Values reflect average ratings. Cohen's d statistics reflect sex
differences, with negative values indicating lower averages in females
(n=372) than in males (n=144).
⁎⁎pb.01.
Table 6
Descriptive statistics for Current Events Knowledge variables
M SD Sk. Ku.αdsex
Prior CE Knowledge
Arts/Entertainment
Business/Economy
Crimes/Accidents/Disasters
U.S. Politics/Government
World Politics/Government
Science/Medicine
Sports/Recreation
Total
.73
.47
.42
.50
.47
.45
.39
.49
.22
.23
.24
.27
.26
.24
.27
.17
−.68
.24
.35
.17
.21
.18
.72
.47
−.07
−.37
−.48
−.66
−.57
−.62
−.16
−.30
.46
.32
.39
.57
.49
.42
.59
.83
−.52⁎⁎
−.57⁎⁎
−.45⁎⁎
−.51⁎⁎
−.68⁎⁎
−.33⁎⁎
−1.31⁎⁎
−.97⁎⁎
New CE Knowledge
Arts/Entertainment
Business/Economy
Crimes/Accidents/Disasters
U.S. Politics/Government
World Politics/Government
Science/Medicine
Sports/Recreation
Total
.57
.35
.54
.65
.42
.50
.44
.50
.24
.18
.23
.25
.24
.19
.26
.14
−.01
.29
−.18
−.41
.34
.11
.48
.33
−.57
−.14
−.31
−.44
−.33
−.16
−.22
−.16
.43
.01
.34
.50
.47
.13
.62
.75
−.43⁎⁎
−.02
−.24⁎⁎
−.33⁎⁎
−.61⁎⁎
−.17
−1.13⁎⁎
−.75⁎⁎
Note. CE=Current Events. Prior CE Knowledge was assessed in
Session 1; New CE Knowledge was assessed in Session 2. Values
reflect proportion correct. Cohen's d statistics reflect sex differences,
with negative values indicating lower averages in females (n=374)
than in males (n=144).
⁎⁎pb.01.
269D.Z. Hambrick et al. / Intelligence 36 (2008) 261–278

Page 10

(αsb.70), perhaps because the tests were administered
in large groups or contained a relatively small number of
items. Nevertheless, the three variables for each ability
construct correlated positively and moderately with each
other (see Table 7). For this reason, we retained all of the
variables for subsequent analyses. Males tended to
outperform females on the tests of cognitive ability (avg.
d=−.25), but this advantage was consistently signifi-
cant only for Gc (avg. d=−.43). As can be seen in
Table 5, the two measures of intellectual openness had
acceptable reliability (αsN.70) and correlated highly
with each other (r=.77); averages for these variables
were higher for males than for females (ds=−.17 and
−.24). The interest variables also had acceptable
Table 7
Correlation matrix
123456789 1011 12 131415 1617 18
Ability
1. Series completion
2. Letter sets
3. Matrix reasoning
4. Synonym vocabulary
5. Reading comprehension
6. ACT
7. Word Span
8. Digit Span
9. Letter Span
−
.48 −
.43
.27
.36
.45
.25
.18
.24
.40 −
.16
.26
.30
.24
.21
.23
.25 −
.35
.46
.25
.14
.18
.55 −
.56
.30
.10
.23
.64 −
.30
.14
.17
.34 −
.21
.28
.34 −
.43 .45 −
Personality and CE Interest
10. Need for cognition
11. Intellect
12. Arts/Entertainment
13. Business/Economy
14. Crimes/Acc's/Dis's
15. U.S. Politics/Gov't
16. World Politics/Gov't
17. Science/Medicine
18. Sports/Recreation
.15
.15
.04
.09
.06
.07
.15
.16
.01 −.18 −.08 −.07 −.01
.03 .02
.05 −.02 −.09 −.06 −.07 −.05 −.04 −.04
−.03 −.06 −.03 .11.11
−.01 −.11 −.04.18 .13
.05 .06 .10.13.20
.12 .06.11 −.02.02
.30
.40
.27
.34
.26
.38
.11
.23
.02
.05
.06
.03 −
.10
.01 −.21 −.19 −
.24
.01
.32
.31
.33
.14 −.08 −.07
.77 −
−.02 −.08 −.05
−.01
.04 −.07 −.01 −.03.20
.01
.32 −.03 .69
.30 −.09 .73
.33 −.05 .32
.32 .27
.06 −
.40 .34 −
.08
.11 −.01 −.03 −.03
.20.02 −.03 −.06
.08 .08
.00 −.02 −.01.29 −
.30 .74 −
.38 .34 .38 −
.31 .15 .13 .07.02 −
News Exposure
19. Newspaper min/week
20. Internet min/week
21. Magazines min/week
22. Television min/week
23. Radio min/week
.04
.02 −.05
−.10 −.02
−.11 −.10 −.13 −.03 −.03 −.05 −.01 −.03
−.03 −.09 −.15 .00 −.03 −.02 −.04 −.03 −.02
.04.09
.02
.00 −.02
.15
.04
.14
.05
.03 −.01 −.09 −.06 −.05
.11
.06 −.03
.05.03
.04
.08
.01
.18
.06
.11
.08
.05
.19 −.08 .23
.09 −.01 .24
.10 −.06 .20
.09 −.04 .31
.05 −.07 .07
.10 .30 .33
.00 .26 .29
.03 .20 .16
.10 .31 .30
.01 .09 .11
.13
.07
.09
.05
.06 −.05
.21
.09
.10
.14 .01
Prior CE Knowledge
24. Arts/Entertainment
25. Business/Economy
26. Crimes/Acc's/Dis's
27. U.S. Politics/Gov't
28. World Politics/Gov't
29. Science/Medicine
30. Sports/Recreation
.20
.10 −.02
.02 −.02
.10 −.03
.10
.18
.06 −.02
.01.15
.09
.06
.12
.14
.17
.06
.32
.32
.30
.39
.42
.33
.20
.29
.28
.26
.30
.38
.33
.12
.29
.29
.25
.31
.36
.32
.15
.13
.11
.12
.14
.13
.12
.04
.00
.01
.05
.00
.01
.06
.10
.10
.04
.08
.09
.08
.11
.14
.11
.24
.14
.20
.22
.18
.04
.22 −.11 .12 −.04 .17 .22
.24 −.23 .20 −.06 .23 .31
.18 −.17 .14 −.03 .20 .31
.24 −.17 .21 −.06 .31 .35
.31 −.16 .21 −.05 .29 .37
.20 −.16 .11 −.05 .17 .18
.06 −.15 .16 −.03 .10 .15
.11
.15
.10
.13
.15
.16
.02
.10
.09
.07
.10
.06
.01
.46
.00
.11
New CE Knowledge
31. Arts/Entertainment
32. Business/Economy
33. Crimes/Acc's /Dis's
34. U.S. Politics/Gov't
35. World Politics/Gov't
36. Science/Medicine
37. Sports/Recreation
.13.05 .10.32
.02
.22
.32
.30
.28
.21
.32
.02
.23
.30
.29
.21
.18
.32
.06
.19
.32
.26
.23
.22
.14 −.03
.03 −.04
.10 −.04
.16 −.03
.10 −.03
.08 −.05
.14
.08
.03
.03
.08
.02
.06
.21
.08
.05
.05
.19
.16
.03
.05
.16 −.13 .10
.04.04 .09
.12 −.06 .10
.25 −.17 .11 −.04 .23 .24
.24 −.15 .21 −.03 .24 .33
.14 −.13 .05 −.05 .10 .14
.11 −.09 .17 −.04 .10 .16
.01 .16 .22
.02 .09 .08 −.03 −.02
.07 .13 .15
.10 .11
−.04 −.05 −.01
.07.08
.10.03
.02.04
.04.01
.11.02
.02
.15
.09
.04
.17
.11
.13
.10
.09
.02
.07
.00
.08
.02
.48.15
19
Note. N=518. CE=Current Events. Correlations with an absolute magnitude greater than.09 are statistically significant (pb.01). All news exposure
estimates are square-root transformed and reflect news exposure between Session 1 and Session 2.
2021222324252627 2829 3031 32 3334353637
270D.Z. Hambrick et al. / Intelligence 36 (2008) 261–278

Page 11

reliability (αs=.86 −.93), and indicative of a general
interest factor, there were moderate-to-strong correla-
tions (avg. r=.45) among the variables for five of seven
domains: Business/Economy, Crimes/Accidents/Disas-
ters, U.S. Politics/Government, World Politics/Govern-
ment, and Science/Medicine (see Table 7). We found
that Arts/Entertainment and Sports/Recreation interest
correlated more weakly with these interest variables.
Averages were significantly (pb.01) higher for females
in the Arts/Entertainment (d=.56) and Crimes/Acci-
dents/Disasters (d=.52) domains, but were higher for
males in the Business/Economy (d=−.38), World
Politics/Government (d=−.38), and Sports/Recreation
(d=−.62) domains.
Table 7 (continued)
19 20 2122 23 2425 26 27 2829 3031 3233 34 353637
−
.30
.19
.15
.08
−
.10
.16
.05
−
.27
.10
−
.24
−
.22
.17
.16
.24
.22
.13
.31
.19
.08
.20
.17
.22
.13
.16
.08
.08
.15
.16
.06
.07
.11
.13
.14
.22
.17
.08
.13
.15
.05
.04
.02
.05
.06
.08
.03
−
.34
.31
.48
.31
.34
.30
−
.37
.50
.40
.32
.30
−
.45
.43
.32
.26
−
.49
.40
.36
−
.35
.30
−
.21
−
.26
.06
.18
.25
.25
.06
.27
.09
.06
.10
.13
.18
.09
.16
.15
.00
.10
.00
.11
.07
.06
.14
.10
.12
.14
.25
.10
.10
.12
.02
.03
.01
.07
.03
.42
.15
.27
.40
.34
.21
.28
.31
.06
.25
.35
.36
.22
.32
.29
.10
.31
.32
.39
.25
.28
.45
.13
.30
.44
.41
.29
.32
.36
.11
.25
.42
.42
.25
.28
.31
.04
.20
.31
.24
.20
.14
.31
.01
.16
.23
.26
.17
.60
−
.12
.32
.39
.37
.34
.28
−
.06
.18
.15
.03
.05
−
.30
.35
.27
.17
−
.38
.28
.25
−
.25
.28
−
.17
−.03
−
271 D.Z. Hambrick et al. / Intelligence 36 (2008) 261–278

Page 12

Finally, an inspection of Table 6 reveals that there
was a wide range of scores, both for prior knowledge
(M=.49; SD=.17; Range=.14 to 1.00) and new knowl-
edge (M=.50; SD=.14; Range=.14 to.93). Reliability
estimates were quite low for some of the individual
knowledge scales. Nevertheless, the overall scales had
acceptable reliability (αs=.83 and .75), and as shown in
Table 7, the individual knowledge variables tended to
correlate positively with each other: Prior CE Knowl-
edge (avg. r=.36) and New CE Knowledge (avg.
r=.24).3Therefore, we retained all knowledge variables
for subsequent analyses. Means were generally higher
for males than for females (avg. d=−.52), and for all but
one comparison, sex differences were statistically
significant. Finally, the prior and new knowledge totals
correlated very highly (r=.71).
4.2.1. Structural equation modeling
The remainder of this section reports structural
equation modeling (SEM) aimed at explaining variance
in newly acquired current events knowledge. Through-
out this section, we report a number of fit statistics. The
χ2test indicates whether there was a significant differ-
ence between the reproduced and observed covariance
matrixes. Thus, non-significant χ2values reflect a fit of
the model to the data. However, when moderate to large
sample sizes are used, slight differences between
reproduced and observed covariance matrices can result
in significant χ2values. The comparative fit index (CFI)
and non-normed fit index (NNFI) are less sensitive to
sample size and reflect improvement in the fit of a model
compared with a baseline model in which population
covariances among observed variables are assumed to
be zero. The root mean squared error of approximation
(RMSEA) reflects the average difference between the
observed and reproduced covariances. CFI and NNFI
values of greater than .90, and RMSEAvalues less than
.08, indicate acceptable fit (e.g., Browne & Cudeck,
1993; Kline, 2005).
4.2.1.1. Confirmatory factor analyses.
were involved in the SEM. The first step was to perform
confirmatory factor analyses to determine whether we
were successful in measuring the intended constructs.
We used the Holzinger bifactor approach to model the
ability variables (see Jensen & Weng, 1994), first spe-
cifying a model in which each indicator loaded onto a
general factor (g). Factor loadings were generally high,
but model fit was poor: χ2(27)=319.83 (pb.01),
CFI=.78, NFI=.76, RMSEA=.15. We added factors
representing three cognitive abilities (Gf, Gc, Gsm).
Each variable had a positive loading on the general
factor (.27–.77) and one specific factor: Gf (.24–.64),
Gc (.35–.51), and Gsm (.40–.65). Model fit was ex-
cellent: χ2(18)=34.35 (pb.05), CFI=.99, NFI=.97,
RMSEA=.04.
For the non-ability variables, we tested a model
consisting of three factors: Intellectual Openness,
Two steps
3Note also that alowcoefficientalpha for aknowledge scale indicates
only that the items for that scale did not correlate highly with each other
(i.e.,werenotinternallyconsistent).Theitemsmaystillrepresentdiverse
and relevant content, and a low alpha does not preclude total scores for
thatscalefromcorrelatingwithtotalscoresforotherscales.Forexample,
although the coefficient alpha for Science/Medicineknowledge (Session
2) was low (.13), it correlated significantly with 11 of the 12 other
knowledge variables (avg. r=.23; see Table 9).
Fig.1.Measurementmodelforcurrentevents(CE)knowledgevariables.B/E=Business/Economy,C/A/D=Crimes/Accidents/Disasters,USP/G=United
States Politics/Government, W P/G=World Politics/Government, S/M=Science/Medicine. Values are standardized regression weights; value on the
bidirectional arrow is a correlation.
272D.Z. Hambrick et al. / Intelligence 36 (2008) 261–278

Page 13

Current Events Interest, and News Exposure. Fit of an
initial model was unacceptable: χ2(74)=471.90
(pb.01), CFI=.82, NFI=.79, RMSEA=.10. Inspection
of the parameter estimates suggested that this was
because two variables had very low loadings on the
interest factor: Entertainment (−.02) and Sports (.21).
And, in fact, model fit was acceptable with these
variables omitted, χ2(51)=196.83 (pb.01), CFI=.92,
NFI=.90, RMSEA=.07. The remaining variables had
positive loadings on their respective factors, Intellectual
Openness (.85–.90), Current Events Interest (.36–.88),
and News Exposure (.23–.49).
For the current events (CE) knowledge variables, we
tested a model with separate factors for Session 1 (Prior
Table 8
Standardized direct, total indirect, and total effect estimates from structural equation model
Endogenous variable
CE Interest News Exposure Prior CE KnowledgeNew CE Knowledge
Exogenous variableβ C.R.β C.R.β C.R.β C.R.
g
Direct effect
Total indirect effect
Total effect
Gc
Direct effect
Total indirect effect
Total effect
Gf
Direct effect
Total indirect effect
Total effect
Gsm
Direct effect
Total indirect effect
Total effect
Intellectual openness
Direct effect
Total indirect effect
Total effect
CE Interest
Direct effect
Total indirect effect
Total effect
News Exposure
Direct effect
Total indirect effect
Total effect
Prior CE Know.
Direct effect
Total indirect effect
Total effect
–
–
–
–
–
–
–– .44 4.52⁎⁎
.10
.39
.49
1.05
3.21⁎⁎
2.45⁎
.24
.24
1.91
1.91
––
.44 4.52⁎⁎
–
–
–
–
–
–
–– .52 3.78⁎⁎
.09
.46
.55
.64
3.13⁎⁎
2.21⁎
.28
.28
1.82
1.82
––
.52 3.78⁎⁎
–
–
–
–
–
–
–
−.09
−.09
–
−1.58
−1.58
−.17
–
−.17
−2.33⁎
–
−2.33⁎
.03 .30
−.15
−.13
−2.00⁎
.83
–
–
–
–
–
–
–
−.01
−.01
–
−.02
–
−.02
−.28
–
−.28
−.10
−.02
−.12
−1.76
−.22
−1.20
−.24
−.24
.366.90⁎⁎
−.10
.18
.08
−1.42
2.39⁎
.66
−.10
.12
.02
−1.68
4.00⁎⁎
.29
−.01
−.02
−.03
−.16
−.21
−.21
––
.36 6.90⁎⁎
–
–
–
–
–
–
.46
.19
.65
4.85⁎⁎
2.46⁎
5.83⁎⁎
.35 6.67⁎⁎
−.17
.45
.28
−1.98⁎
3.37⁎⁎
2.86⁎⁎
––
.35⁎⁎
6.67⁎⁎
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
.30 2.28⁎
––
.302.28⁎
–
–
–
–
–
–
.54 3.55⁎⁎
–
–
–
–
–
–
.734.78⁎⁎
––––
.543.55⁎⁎
.73 4.78⁎⁎
g Gc GfGsm
CorrelationsβC.R.
5.31⁎⁎
−1.14
−2.46⁎
β C.R.
3.36⁎⁎
.74
−1.69
βC.R.β C.R.
Intellectual Openness
Current Events Interest
News Exposure
.38.34
.07
−.10
−.14
.02
−1.43
−1.83
.16
−.03
−.07
.09
−.51
−1.11
.94
−.07
−.27
−.29
Note. CE=Current Events. Direct effects reflect unmediated associations, total indirect effects reflect the sum of all possible mediated associations,
and total effects reflect the sum of direct and total indirect effects. β=standardized regression weight. C.R.=critical ratio. Bootstrapped standard
errors were used to compute C.R.s for total indirect and total effects (when not the same as direct effects).
⁎pb.05;⁎⁎pb.01.
273D.Z. Hambrick et al. / Intelligence 36 (2008) 261–278

Page 14

CE Knowledge) and Session 2 (New CE Knowledge).
Model fit was acceptable, χ2(34)=37.37 (ns), CFI=1.00,
NFI=.97, RMSEA=.01. As shown in Fig. 1, the factors
correlated highly (r=.89). However, when the correlation
was constrained to unity, model fit was reduced, χ2(35)=
50.95 (pb.05), CFI=.99, NFI=.96, RMSEA=.03, and
the difference in the model fits was statistically signifi-
cant: ▵χ2(1)=13.58, pb.01. We were therefore justified
in treating the knowledge factors as distinct.
4.2.1.2. Structural models.
for effects of the predictor variables on newly acquired
current events knowledge (New CE Knowledge). Of
particular interest were the two pathways discussed
previously. The ability pathway included direct effects of
the general (g) and specific ability (Gf, Gc, Gsm) factors
on Prior CE Knowledge and New CE Knowledge, and a
positive effect of Prior CE Knowledge on New CE
Knowledge. We predicted a positive effect of g on Prior
CE Knowledge, and a positive effect of Prior CE
Knowledge on New CE Knowledge. We also predicted
a positive effect of Gc on Prior CE Knowledge, beyond
the influence of g. For the non-ability pathway, we pre-
The next step was to test
dicted a positive effect of Intellectual Openness on Cur-
rent Events Interest, a positive effect of Current Events
Interest on News Exposure, and a positive effect of News
Exposure on New CE Knowledge. We made no specific
predictions about relations between the ability and non-
ability factors (e.g., Gf and Current Events Interest), and
hence paths between these variables were bidirectional.
Direct, indirect, and total effects are summarized in
Table 8, while parameters from the model that reached
statistical significance (pb.05) are displayed in Fig. 2.
There was a direct effect of g on Prior CE Knowledge
(.44), and of Prior CE Knowledge on New CE
Knowledge (.73). However, this was not the whole
story with respect to the ability factors, as there was a
direct effect of Gc on Prior CE Knowledge (.52),
beyond the influence of g. We also found evidence for a
non-ability pathway. That is, there were direct effects of
Intellectual Openness on Current Events Interest (.35),
Current Events Interest on News Exposure (.46), and
News Exposure on Prior CE Knowledge (.30), all as
predicted. Two unexpected findings were direct nega-
tive effects of Gf on Prior CE Knowledge (−.17) and of
CE Interest on New CE Knowledge (−.17). The predictor
Fig. 2. Structural equation model with ability and non-ability factors predicting current events (CE) knowledge. B/E=Business/Economy,
C/A/D=Crimes/Accidents/Disasters,US P/G=UnitedStatesPolitics/Government, W P/G=World Politics/Government, S/M=Science/Medicine.
Values on single-headed arrows are standardized regression weights; values on double-headed arrows are correlations.
274 D.Z. Hambrick et al. / Intelligence 36 (2008) 261–278

Page 15

variables accounted for 83.9% of the variance in New CE
Knowledge. Model fit was acceptable: χ2(395)=690.69
(pb.01), CFI=.94, NFI=.87, RMSEA=.04.4
4.2.1.2.1. Sports/recreation and arts/entertain-
ment. We ran another set of models to test for predictors
of knowledge in the Arts/Entertainment (A/E) and Sports/
Recreation (S/R) domains. For each domain, the models
were configured the same as in the preceding analysis,
exceptthattherewasasingleindicatorforbothinterest,and
for prior knowledge and new knowledge. There were pos-
itive effects of g and Gc on Prior A/E Knowledge (.27 and
.25,respectively),andofGc(.34)onPriorS/RKnowledge.
In turn,there were positive effects of Prior S/RKnowledge
onNewS/RKnowledge(.43),andofPriorA/EKnowledge
on New A/E Knowledge (.20). There was also some
evidence for non-ability contributions: a positive effect of
News Exposure on New A/E Knowledge (.29), and pos-
itive effects of S/R Interest on both Prior S/R Knowledge
(.52) and New S/R Knowledge (.22). (All psb.01).
4.2.1.3. Sex differences.
to investigate sex differences in current events knowledge.
The specific question of interest was whether these sex
differences would remain statistically significant after
controllingforsexdifferencesintheabilityandnon-ability
predictor variables. To answer this question, we added sex
as a predictor variable to the structural equation model
described previously. For statistical reasons, we decided
not to run a model with paths leading from sex to all of the
ability factors in the manner that they were previously
modeled. Instead, we modeled g and Gc only for this
analysis.5Focusing on Gc was a logical choice, because
We performed a final analysis
Gcwasfoundtobeasignificantpredictorofcurrentevents
knowledge in the previous analyses, and because there
were consistent, sizeable sex differences in Gc but not the
other abilities (cf. Table 4).
This analysis revealed that there were direct negative
effects of sex on the ability variables–g (−.16) and Gc
(−.20), psb.01–and the non-ability variables—Intellec-
tualOpenness(−.12),Current EventsInterest(−.11),and
News Exposure (−.18), psb.05. (Mean levels were
higher for males than for females for all variables.)
Critically,however, the negativeeffectofsexonPrior CE
Knowledge (−.20) remained statistically significant
(pb.01) even after taking sex differences in these other
factors into account. (The effect of sex on New CE
Knowledge was non-significant.) This was also true for
theeffectofsexonPriorA/EKnowledge(−.16)andPrior
S/R Knowledge (−.42). Overall, then, the factors con-
sidered here contributed to, but could not completely
explain, sex differences in current events knowledge
favoring males.
5. Discussion
Why do some people know more than others? For
example, in an election, why do some people possess
extensive knowledge of the issues (the economy, the
environment, etc.), whereas others know little more than
the candidates' party affiliations? More generally, what
accounts for individual differences in knowledge? The
purpose of the present study was to test for effects of
ability, interest, personality, and experience variables,
along with prior knowledge, on knowledge acquired
under naturalistic learning conditions. Previous research
hasconsideredsome ofthese factors. Forexample,Beier
and Ackerman (2001) looked at effects of a number of
different predictors on current events knowledge,
including ability and personality variables, but they did
not attempt to measure and model specific interests and
experience. Beier and Ackerman (2005) added topic-
relevant experience as a predictor of knowledge ac-
quisition, but they did not model personality and interest
factors as antecedents of this experience. Thus, toward
the goal of gaining a more complete understanding of
what underlies inter-individual differences in knowl-
edge,thecontributionofthisstudytotheextantliterature
is to consider, within a single model, a broader range of
influences than has been previously examined.
5.1. Major findings
There was evidence for an ability influence on acqui-
sition of current events knowledge. The g factor we
4Asalreadyexplained,forthemodeldisplayedinFig.2,weusedthe
“best” 42 items (6 per category) from the 90-item test of current events
knowledgegiveninSession2.However,becausethe48excludeditems
tapped relevant content, we re-ran the model with the total score on the
test as the single indicator for new knowledge. The results were similar
to those displayed in Fig. 2: there were positive effects of Prior CE
Knowledge (.62) and News Exposure (.37) on the total score.
5We scaled the ability factors by constraining the variance of each
factor to 1.0 instead of setting the factor loading of one indicator per
factorto1.0.Wedidthisbecause,hadweusedthelatterapproach,forone
of the three sets of ability indicators (e.g., Gc), it would have been
necessary to use one indicator to scale the specific ability and then
another to scale the g factor, leaving only one freely estimated factor
loading.Thisworkedwell,exceptthatfortheanalysisofsexdifferences,
it created a problem: when a latent factor is scaled by setting its variance
to 1.0, it cannot be treated as an endogenous factor that is predicted by
some variable (e.g., sex); it can only be exogenous (see Kline, 2005). To
get around this problem, for the analysis of sex differences, we left all of
the ability variables in the model, but we specified only g and Gc. We
scaled the g factor by setting the factor loading of one Gf indicator to 1.0
and theGc factor by setting the factorloadingof oneGc indicator to 1.0.
275D.Z. Hambrick et al. / Intelligence 36 (2008) 261–278