You are on page 1of 30

Intelligence 30 (2001) 71 – 100

When IQ is controlled, does motivation still


predict achievement?$
Françoys Gagnéa,*, François St Pèreb
a
Department of Psychology, University of Québec à Montréal, P.O. Box 8888,
Downtown Station, Montreal, Canada H3C 3P8
b
Clinique de consultation conjugale et familiale Poitras-Wright, Côté 32, rue St-Charles, ouest, suite 220,
Longuevil, QC, Canada, J4H 1C6
Received 10 January 2000; received in revised form 18 June 2000; accepted 18 August 2000

Abstract

Past research shows controversial results concerning relationships between cognitive abilities,
motivation, and achievement. The present study aims to assess the unique contribution of motivation
to academic achievement, after controlling for the predictive power of cognitive abilities. Over 200
female high school students completed two IQ tests, and three motivation-related measures (intrinsic,
extrinsic, and persistence) twice during a semester. One parent and two teachers also rated twice each
student’s three types of motivation. The results revealed among other things that (a) IQ and motivation
were not correlated; (b) the parents’ ratings were only slightly related to their children’s own
judgments; (c) the teachers’ ratings had doubtful predictive validity because they were strongly biased
by the teachers’ knowledge of the students’ achievements; (d) cognitive abilities were by far the best
predictor of school achievement; and (e) the students’ self-assessments of their motivation were not
related to their academic achievement. These results question the belief of most educators about the
crucial role of motivation as a determinant of scholastic achievement. D 2001 Elsevier Science Inc.
All rights reserved.

Keywords: Intelligence; Motivation; Intrinsic motivation; Volition; Persistence; Achievement; Academic


achievement; Predictive validity; Parents; Teachers

$
The study was conducted to become the second author’s PhD thesis.
* Corresponding author.
E-mail address: gagne.francoys@uqam.ca (F. Gagné).

0160-2896/01/$ – see front matter D 2001 Elsevier Science Inc. All rights reserved.
PII: S 0 1 6 0 - 2 8 9 6 ( 0 1 ) 0 0 0 6 8 - X
72 F. Gagné, F. St Père / Intelligence 30 (2001) 71–100

1. Introduction

In his famous book ‘‘Hereditary Genius,’’ Francis Galton (1869/1962) circumscribed the
roots of eminence in the following way:
By natural ability, I mean those qualities of intellect and disposition, which urge and qualify a
man to perform acts that lead to reputation. I do not mean capacity without zeal, nor zeal
without capacity, nor even a combination of both of them, without an adequate power of
doing a great deal of very laborious work. (p. 77)

In Galton’s terms, reputation (talent, eminence) will emerge from proper qualifications
(high capacities, gifts), urges and zeal (needs, passions), as well as the power for laborious
work (will power, persistence). These ‘‘ingredients’’ of exceptional performance identified by
Sir Francis Galton more than a century ago have reappeared again and again (a) in general
models of school learning (e.g., Bloom, 1976; Carroll, 1963; Walberg, 1984), (b) in more
specific models of talent emergence (e.g., Feldhusen, 1992; Gagné, 1993; Renzulli, 1986;
Tannenbaum, 1983), (c) in models of skill growth proposed by industrial psychologists (e.g.,
Atkinson, 1964; Kanfer & Ackerman, 1989; Vroom, 1964), and (d) in biographical analyses
of eminent historical figures (e.g., Simonton, 1994). Cognitive aptitudes and motivational
factors are probably the two most commonly mentioned determinants of academic achieve-
ment. Not only are they included in theoretical models, but they have appeared as
independent variables in thousands of empirical studies of school learning, skill training in
work settings, talent development in arts and sports, as well as longitudinal studies of
occupational achievement. The most common belief within the general population is that both
factors exert approximately equal causal influences on talent development (Gagné &
Blanchard, submitted for publication). Attribution studies also show that both effort and
ability are by far the two major causal attributions for both success and failure, not only in
academics (Good & Brophy, 1990), but also in music (Austin & Vispoel, 1998) and sports
(Biddle, 1993).
To what extent does empirical research support these ‘‘equalitarian’’ beliefs? First, the very
strong relationship between cognitive aptitudes, usually measured by some type of IQ test,
and academic achievement is well-documented. In a survey of close to 3000 empirical studies
of school learning, Walberg and his colleagues (see Walberg, 1984) computed an average
correlation of .71 between various IQ measures and academic achievement. It emerged as the
most powerful determinant, by far, among dozens of factors examined. Similarly, in her
synthesis of many meta-analytic surveys of the predictive validity of IQ tests in work settings,
Gottfredson (1997) argued (a) that the validity of intelligence measures applied to most
occupations, (b) that it rose with job complexity, and (c) that ‘‘g can be said to be the most
powerful single predictor of overall job performance’’ (p. 83). Recently, Schmidt and Hunter
(1998) published for the field of personnel psychology a synthesis similar to that of
Walberg’s. They analyzed the predictive validity of 19 distinct constructs and selection
methods, including general mental ability (GMA) tests, work samples, interviews, assessment
centers, job knowledge, job tryout, job experience, peer ratings, even graphology! In the case
of GMA tests, they computed validity coefficients of .51 and .56 with job performance and
job training, respectively. The relationship between motivation and achievement is also well-
F. Gagné, F. St Père / Intelligence 30 (2001) 71–100 73

documented. In the above-mentioned review, Walberg (1984) and his colleagues found an
average correlation of .34 between various indices of motivation and school learning. In terms
of explained variance, this contribution is about one-fourth of the predictive power of IQ
(r2 =.50 for IQ, and .12 for motivation). And it is probably a generous estimate since the text
does not specify if the correlations used for the motivation component were independent
contributions, that is after controlling for the predictive power of intelligence. In the Schmidt
and Hunter synthesis, 2 of the 19 predictors analyzed belong to the general area of
motivation, namely conscientiousness and vocational interests; their average validities
with regard to job performance are .31 and .10, respectively. Again, these are zero-
order correlations.
If cognitive abilities and motivation are both significant determinants of achievement, to
what extent are they related to each other? Schmidt and Hunter (1998) cited Ones,
Viswesvaran, and Schimdt (1993) to support their premise of a zero correlation between
conscientiousness and IQ; they also cited Holland (1986) to justify a similar lack of
correlation between interests and cognitive abilities. On the other hand, the gifted education
literature is replete with statements to the effect that gifted children are more motivated;
most checklists of ‘‘gifted’’ characteristics include curiosity, incessant questioning, and
similar indices of intrinsic motivation (IM) to learn (e.g., Clark, 1996; Davis & Rimm,
1989; Gallagher, 1985). Some research also supports that relationship, showing significant
differences between gifted children or adolescents and peers of average abilities on a variety
of measures of motivation: intellectual curiosity, IM, achievement motivation, task orienta-
tion, and so forth (Gottfried & Gottfried, 1996; Janos & Robinson, 1985; Lloyd &
Barenblatt, 1984; Vallerand, Gagné, Senécal, & Pelletier, 1994). But, other studies report
no correlation whatsoever between measures of these two constructs (e.g., Aspinwall &
Taylor, 1992; Joswig, 1994; Spence, Pred, & Helmreich, 1989; Wong & Csikszentmihalyi,
1991). These contradictory results brought Shore, Cornell, Robinson, and Ward (1991) to
express the following doubts: ‘‘There is insufficient explanation of how attention to intrinsic
motivation might be different for the gifted compared with others. The implication in much
of the literature that gifted children have more of it, or a superior kind, is not supported’’
(p. 215).
The prediction problem is compounded by a major terminological ambiguity, since there is
no agreed upon definition of the term gifted (Gagné, 1995). Consequently, the gifted subjects
in comparative studies are selected with a variety of procedures. Gagné argued that two types
of measures are used more regularly: IQ tests (group or individual) and academic achieve-
ment scores (grades or standardized tests). In the framework of Gagné’s Differentiated Model
of Giftedness and Talent, IQ tests pinpoint intellectually gifted (IG) students, whereas grades
identify academically talented (AT) students, or, in Gagné’s (1995) words, IGAT students.
These two groups overlap only partially. Consequently, many studies that compare ‘‘gifted’’
children with peers of average abilities are often comparing achieving gifted children. It
follows that many of the significant differences observed between the two groups could be
attributable as much to the superior cognitive abilities of these gifted children as to their
academic talent. This methodological problem speaks eloquently in favor of efforts to
disentangle the respective causal contributions of cognitive aptitudes (A) and motivation
(M) to achievement.
74 F. Gagné, F. St Père / Intelligence 30 (2001) 71–100

2. The multiple regression literature

The relationships involving these two factors can be represented by two distinct causal
models: an additive (A + M) model, and a multiplicative (A  M) model.

2.1. The additive causal model

The more common and simple additive model guiding the empirical analysis of the causal
relationships between aptitudes, motivation, and achievement underlies the question: ‘‘How
much does motivation add to the prediction of school achievement after the predictive power
of cognitive aptitudes has been controlled?’’ One might ask why precedence should be given
to aptitudes in that causal chain; why not examine first the predictive power of motivation,
then look for the proportion of the remaining variance explained by aptitudes? There are at
least two answers: a theoretical one and a statistical one, both closely related. There is virtual
unanimity among intelligence theorists that g reflects the ability to reason, solve problems,
think abstractly, and acquire knowledge (Snyderman & Rothman, 1987). Gottfredson (1997)
points out: ‘‘Intelligence is not the amount of information people know, but their ability to
recognize, acquire, organize, update, select, and apply it effectively. In educational contexts,
these complex mental behaviors are referred to as higher thinking skills’’ (p. 93). Similarly,
Gagné affirms that ‘‘any structured LTP [learning, training, practicing] program uses as its
starting point some appropriate NAT [natural] abilities, which it progressively transforms,
adapts, and improves to create the skills and competencies that are characteristic of a given
occupational field’’ (Gagné, 1999, p. 115). By contrast, motivation plays the role of a
facilitator or catalyst in the talent development process. In brief, the relationship between
aptitudes and achievement is more direct. As we have seen above (Schmidt & Hunter, 1998;
Walberg, 1984), that closeness is confirmed at the statistical level by much higher correlations
between IQ and achievement than between motivation and achievement. That difference
leads to the priority given to the stronger link in the causal chain.
Surprisingly, very few studies can be found in the education literature that address the
question of the independent additive contributions of these two major determinants of
academic achievement. The following survey will distinguish microlevel studies from
macrolevel ones. The former typically use an experimental design with small samples,
measure achievement on a short learning task, and manipulate motivation differently in
treatment and control groups. We will survey them first since their more artificial environ-
ment makes them somewhat less relevant to our goal of assessing the respective roles of
aptitudes and motivation in real-life, long-term learning situations. A series of microlevel
studies was conducted as part of a major research effort to validate Bloom’s (1976) model of
mastery learning. The model itself included three causal constructs: cognitive entry behaviors
(CEB), affective entry behaviors (AEB), and quality of instruction (QI). CEBs correspond
essentially to past mastered learning, but can include aptitude test results. That encompassing
definition of CEBs creates a problem similar to the one mentioned above with reference to the
nondissociation of ‘‘pure’’ intellectually gifted (IG) students from academically talented ones
(IGAT). AEBs include interests, attitudes toward school learning, self-views (e.g., confidence
in one’s ability to learn, in resources to overcome difficulties), and so forth. Based on an
F. Gagné, F. St Père / Intelligence 30 (2001) 71–100 75

extensive review of the school learning literature, Bloom theorized that the upper limits for
the predictive validity of the first two factors (cognitive and affective antecedents) would be
.70 (r2=.49) and .50 (r2=.25), respectively, with a combined predictive power of .80 (r2=.64).
These values indicate that Bloom recognized (a) the much stronger predictive power of
CEBs, (b) a significant overlap between the two constructs, and (c) a significant unique
contribution of about .15 (.64–.49) for AEBs in accounting for academic achievement. In
microstudies of mastery learning completed by five of his PhD students (see Bloom, 1976,
Table 7.3, p. 192), the results revealed that in both the control and mastery learning groups the
AEBs added virtually nothing to the predictive validity of CEBs. Because these students did
not use cognitive ability tests as their CEB measures, but past learning scores, they increased
significantly the predictive power of that component. Moreover, AEBs were assessed with a
very short measure of interest for the subject matter. Bloom acknowledges that ‘‘the
correlations for the cognitive plus affective entry characteristics are not much higher than
those for the cognitive measures only,’’ blaming the ‘‘very limited measures of affect used in
these brief studies’’ (p. 193). Still, it is very unlikely that more valid measures of AEBs
would have brought the observed correlations from zero to anywhere close to Bloom’s
theoretical predictions.
For their part, Grabe and Latta (1981) examined the predictive power of aptitudes (past
GPA and ACT score), achievement motivation, and effort, on end of semester achievement in
two college courses. They measured achievement motivation with Mehrabian’s Resultant
Achievement Motivation scale, while effort was assessed through the frequency of attempts at
bettering one’s score on weekly quizzes (by redoing them), depending on one’s initial
performance. They concluded that ‘‘appropriate effort was strongly correlated with student
achievement, even when differences in student aptitude were controlled’’ (p. 7). Unfortu-
nately, their measures of effort were themselves directly dependent on within-semester
achievement, thus confounding their dependent and independent variables, and consequently
inflating very substantially the causal relationship. The last microlevel study is a series of four
very short experiments by Millman, Bieger, Klag, and Pine (1983). They were designed to
test a corollary of Carroll’s (1963, 1989) model of school learning, namely that an increase in
persistence will not alter learning efficiency. The authors used various motivators (encourage-
ments, gifts, etc.) to manipulate the persistence of the children (Grades 2 to 4) in the treatment
groups as they tried to memorize pairs of words. Academic aptitude was controlled in the
selection process by asking teachers to propose students ‘‘who were most like each other in
terms of their reading level and verbal ability in general’’ (p. 426). They found that, ‘‘as
predicted by the model, none of the differences was statistically significant’’ (p. 425); but they
observed that the increased induced persistence affected the amount of time spent on the task
and, consequently, the amount learned (effectiveness).
Only five macrolevel studies were found that tested the additive model. Gottfried (1979)
administered a multiscale inventory of IM to 141 middle-school students (Grades 4 to 7).
Each scale focused on a specific subject matter (reading, math, social studies, and science).
After partialling out IQ, she found significant correlations, ranging from .20 to .41, between a
given subject matter IM measure and the corresponding achievement score on the Stanford
Achievement Test. The average r2 was close to .10. Because of these within subject matter
pairings, the author argued that her results supported a nontrait, specificity account of IM. No
76 F. Gagné, F. St Père / Intelligence 30 (2001) 71–100

gender differences were observed. Lloyd and Barenblatt (1984) gathered data from 455 high
school sophomores to test five hypotheses on the relationships between intrinsic intellectual
motivation (IIM), socioeconomic status, need achievement, intelligence, sex, and scholastic
achievement. IQ measures came from student files, in which different tests scores were
recorded (Lorge–Thorndike, Henmon–Nelson, Otis Beta). The custom-made IIM scale was
comprised of 44 Likert-type statements; need achievement scores came from a subscale of the
Edwards Personal Preference Schedule, and achievement was measured with the reading test
from the High School Battery of the Metropolitan Achievement Tests. The two motivation
measures, need achievement and IIM, were both equally correlated with IQ (.27), but only
IIM had a substantially significant correlation with achievement (.37). The IQ’s predictive
validity coefficient was .68. In a multiple regression, IQ explained 46% of the achievement
variance, and IIM added a significant, but small, unique contribution of 3.6%; this
corresponds to a 13:1 ratio between the two measures in terms of predictive power. Spence
et al. (1989) administered to four large samples of college students (N > 900) a 7-item
factorially independent subscale of the Jencks Activity Scale, called Achievement Strivings
(AS); its name suggests a close relationship with the need achievement construct. The authors
used SAT and GPA scores as measures of academic aptitude and scholastic achievement,
respectively. The AS and SAT scores were not correlated, but both contributed to the
prediction of GPA. The SAT’s unique contribution in terms of percentage of explained
variance varied between .11 and .26, whereas the AS’s impact varied from .07 to .13. The
SAT’s average contribution overall (.20) was about twice as large as that of the AS
scale (.11).
In their study of the relationships between personality, experience while studying, and
academic performance, Wong and Csikszentmihalyi (1991) measured a large set of variables
with close to 200 junior high school students: (a) scholastic aptitudes with the Preliminary
Scholastic Aptitude Test (PSAT), a junior-high analogue of the well-known SAT; (b) a factor
from the Personality Research Form, labeled Work Orientation, which ‘‘contained the
personality characteristics that were believed to be important in academic performance —
for instance, the motive to achieve, to control impulses, endure, etc.’’ (p. 547); (c) experiences
felt during studying (one question each for IM, happiness, satisfaction with work, concen-
tration, and unselfconsciousness) using the Experience Sampling Method (ESM); and (d)
scholastic achievement with end-of-year GPA. The ESM uses random paging (during waking
hours!); when paged, participants use a special form to indicate their ongoing activity, and
answer the five questions. Focusing on studying activities, Wong and Csikszentmihalyi found
that work orientation and IM were not related to each other or to aptitudes, and only the
former predicted academic achievement. In multiple regression analyses, the PSAT was by far
the best predictor, explaining approximately 20% of the variance in achievement, while work
orientation, a measure analogous to persistence and will-power, explained between 5% (girls)
and 10% (boys) of the GPA variance. Unselfconsciousness also had a unique, albeit small
(4%), predictive power, and level of concentration while studying also contributed (7%), but
only within the boys’ sample. It is important to point out that their nonhierarchical multiple
regression design deleted any covariance between the predictors, thus reducing the SAT’s
explanatory power. Finally, as part of a longitudinal study of the determinants of college
adaptation and achievement, Aspinwall and Taylor (1992) examined the predictive role of
F. Gagné, F. St Père / Intelligence 30 (2001) 71–100 77

measures of optimism, psychological control and self-esteem with regard to motivation and
subsequent achievement, while controlling initial aptitudes with the SAT. Motivation was
assessed 3 months after admission to college with a 15-item locally written scale that
combined questions about IM, persistence, school aspirations, and predictions of success in
college. GPA scores were collected 2 years later. The authors used an a priori structural
equation model with standardized parameter estimates; they found that the SAT had twice the
predictive power of motivation (.50 vs. .25). A low unique causal relationship (.12) was
observed between aptitudes and motivation.
If the information is scarce in education, it appears much more abundant in the field of
personnel psychology. One of the most extensive efforts at hierarchizing a large set of
predictors of job performance was realized within the U.S. Army during the 1980s. Over 70
different predictors were measured: (a) general cognitive, spatial, and perceptual–psycho-
motor abilities; (b) temperament and personality; (c) vocational interests; and (d) job outcome
preferences. The sample of over 4000 recruits covered nine different enlisted jobs. Two job
performance indices were collected (core technical proficiency and general soldiering
proficiency), as well as three measures of ‘‘personal achievement’’ (effort and leadership,
personal discipline, and physical fitness/military bearing). In the course of the analysis
process, the predictors were combined into two large composites: cognitive ability (11
variables) and temperament/interests (13 variables). The cognitive composite had validity
coefficients of .65 and .69 with the two job performance measures; these coefficients only
increased to .67 and .70 when the second composite was added, a substantially minuscule
improvement (McHenry, Hough, Toquam, Hanson, & Ashworth, 1990). In their synthesis of
the predictive validity of 19 predictors of job performance, Schmidt and Hunter (1998)
estimated the unique contribution of each of them, that is how much they added to the
prediction of job performance beyond the .51 average validity of GMA tests. Only three
predictors increased the predictive power by at least .10: work sample tests (.12); integrity
tests (.14), and structured employment interviews (.12). The two motivation-related con-
structs had smaller unique contributions: .09 for conscientiousness and .01 for vocational
interests. The authors argued that an 18% gain in validity (.09/.51, but still a 6:1 ratio in
predictive power) could be considered a substantial improvement.

2.2. The multiplicative causal model

The multiplicative (A  M) model gives motivation the role of a moderator variable, so


that the relationship between aptitudes and motivation becomes equivalent to the interaction
effect in a two-way ANOVA. In other words, individual differences in motivation can modify
the strength of the predictive validity of cognitive aptitudes with regard to achievement. The
multiplicative model assumes that both aptitudes and motivation are essential for perform-
ance, so that there is a definite limit to the compensatory power of any of the two: if one of
them is very low, performance will also be very low. Anastasi and Urbina (1997) reviewed
past efforts to identify significant moderator variables of the predictive validity of IQ with
regard to academic achievement, identifying the 1950s and 1960s as the most active period.
The only consistent finding, according to them, was a small sex effect, with slightly higher
predictive validity coefficients for women than men, especially at the college level. About the
78 F. Gagné, F. St Père / Intelligence 30 (2001) 71–100

global effort to find moderator variables (including interests and motivation) of the strong
relationship between aptitudes and academic achievement, Anastasi and Urbina concluded:
‘‘in the light of present knowledge, no variable can be assumed to moderate validities in the
absence of explicit evidence for such an effect’’ (p. 156).
Terborg (1977) also reviewed efforts to validate the multiplicative model, focusing on its
application in the training of occupational skills. He pointed out that the A  M model was
hypothesized by many scholars (e.g., Atkinson, 1964; Gagné & Fleishman, 1959; Heider,
1958, Vroom, 1964; all cited in Terborg, 1977), and that the additive and multiplicative
sources were not mutually exclusive; in fact, both can be measured simultaneously, and
account independently for a significant percentage of the variation in a performance measure.
Terborg reviewed 14 empirical studies that attempted to verify the presence of both additive
and multiplicative causal influences of performance; he stated that ‘‘two studies found clear
support (. . .), six studies reported mixed findings, and six studies found no evidence for an
interaction’’ (p. 192). Terborg conducted his own experiment, assessing ability, effort, role
definitions, and the aptitude by motivation (effort) interaction. Abilities were measured by a
composite score from (a) the Otis–Lennon, (b) a mechanical aptitudes test, (c) a test of study
skills, and (d) a test of paragraph meaning from an achievement battery. Effort was measured
through time-lapse movie films, and was operationally defined as ‘‘the percentage of time a
person spent working at the material’’ (p. 199). He obtained a multiple correlation of .83 with
his three variables (including interaction), but observed ‘‘that ability alone correlated r = .81
with performance and additional consideration of effort and role definitions increased the
multiple R by only .02 units (. . .) inclusion of the multiplicative term did not significantly
increase predicted performance variance’’ (p. 204).

3. Planning the new study

3.1. Critique of past research

Whether one uses an additive or a multiplicative model of the relationship between


aptitude and motivation, motivation’s independent contribution to the prediction of scholastic
or occupational achievement appears limited. It is frequently nonexistent (Anastasi & Urbina,
1997; Bloom, 1976; McHenry et al., 1990; Terborg, 1977) or much less powerful than the
independent contribution of cognitive abilities. At best, the ratio of these two contributions
reaches 2:1 (Aspinwall & Taylor, 1992; Wong & Csikszentmihalyi, 1991); at worst, it can be
as low as 13:1 (Lloyd & Barenblatt, 1984). The 4:1 and 6:1 ratios respectively extracted from
Schmidt and Hunter’s (1998) and Walberg’s (1984) syntheses, probably upper-limit esti-
mates, are more or less equidistant from the two extremes. Could this limited influence of
motivation be explained by methodological problems? We purposely introduced in the
literature review details that highlight some of these problems, the most common being the
inconsistent operationalization of the variables. The problem was smaller in the case of
aptitudes, although Bloom’s use of past achievement as CEB introduced an artificially high
correlation with his dependent variable. Grabe and Latta (1981) did the same, mixing past
GPA with the ACT to measure aptitudes. Such measures can hardly be labeled aptitude
F. Gagné, F. St Père / Intelligence 30 (2001) 71–100 79

measures, at least in the more commonly acknowledged meaning of that concept (Anastasi &
Urbina, 1997; Snow, 1992).
There was also much variability in the way the motivation construct was defined and
assessed. Measures included IM scales (Bloom, 1976; Lloyd & Barenblatt, 1984; Wong &
Csikszentmihalyi, 1991), need achievement indices (Grabe & Latta, 1981; Spence et al.,
1989), extrinsically induced motivation leading to persistence (Millman et al., 1983), effort
defined as time on task (Terborg, 1977) or multiple trials (Grabe & Latta, 1981), and
persistence included in composite measures (Aspinwall & Taylor, 1992; Wong & Csikszent-
mihalyi, 1991). Some used two distinct measures of motivation (Grabe & Latta, 1981; Lloyd
& Barenblatt, 1984; Millman et al., 1983; Wong & Csikszentmihalyi, 1991), but the rationale
for their inclusion was not always specified. Some instruments clearly circumscribed their
target construct, like Spence et al.’s (1989) AS scale, Wong and Csikszentmihalyi’s (1991)
work orientation factor, Schmidt and Hunter’s (1998) conscientiousness construct, or Lloyd
and Barenblatt’s (1984) IIM scale; others looked more like a hodgepodge of weakly related
concepts (e.g., Aspinwall & Taylor, 1992; McHenry et al., 1990). With such diversity, it is
hardly surprising that few consistent results were found. Still, one trend emerges, namely the
generally nonsignificant predictive power of IM (Bloom, 1976; Lloyd & Barenblatt, 1984) as
opposed to the much larger impact of measures associated with the need to achieve and the
will to persist (Aspinwall & Taylor, 1992; Schmidt & Hunter, 1998; Spence et al., 1989;
Wong & Csikszentmihalyi, 1991).
These divergent results fit perfectly well with Lyn Corno’s work; she distinguishes
predecisional (decision making) and postdecisional (implementation) components in any
goal-seeking process, labeling them motivation and volition, respectively. Broadly para-
phrasing Corno and Kanfer (1993, pp. 303–305), these two constructs can be described
as follows. Motivation comprises constructs and processes that affect decision making and
choice with respect to an individual’s goals. The major determinants include individual
differences in preferences, beliefs, expectancies, perceptions of outcome value, patterns of
attribution, goal orientation, self-efficacy judgments, self-worth, and so forth; they have
been used to distinguish between extrinsically and intrinsically motivated choice behavior
and actions. Volition can be considered in terms of three broad construct/process clusters:
(a) individual differences in action control processes, which refer to knowledge and
strategies used to manage cognitive and noncognitive resources for goal attainment; (b)
use of goal-related cognitions and flexible strategies for self-monitoring, self-evaluation, &
self-regulation; (c) individual differences in dispositionally based volitional styles derived
from factor-analytic views of personality (e.g., the will power or persistence factor
mentioned in Digman, 1990). For their part, Deci and Ryan (1985) elaborated on the
polarity between intrinsic (IM) and extrinsic (EM) motives, based on their self-determi-
nation theory.
Intrinsically motivated behaviors are engaged in for their own sake — for the pleasure and
satisfaction derived from their performance. (. . .) Extrinsically motivated behaviors, on the
other hand, are instrumental in nature. They are performed not out of interest but because they
are believed to be instrumental to some separable consequence. (. . .) Self-determination
theory posits that the four types of extrinsic motivation [external, introjected, identified,
integrated] result from the internalization processes having been differentially effective. The
80 F. Gagné, F. St Père / Intelligence 30 (2001) 71–100

resulting regulatory styles thus fall at different points along an autonomy continuum that
describes the extent to which they have been internalized and integrated (Deci, Vallerand,
Pelletier, & Ryan, 1991, pp. 328– 329).
That quote suggests a single continuum of self-determination, with extrinsic motives
occupying the negative end, and intrinsic motives the positive end. The above tripartite
distinction guided our efforts to improve the assessment of the motivation components.
Because of the lack of instruments to assess the volitional component, we chose to focus
on one recurring construct clearly related with volition, namely persistence. The literature
review shows much ambiguity between the concepts of effort and persistence. A survey of
various definitions suggests that effort is usually associated with using one’s abilities to
their fullest, or doing one’s best (Holloway, 1987), whereas persistence adds both a
duration and an obstacle component: maintaining effort over time in order to overcome
obstacles to the sought goal (Kanfer & Ackerman, 1989). Persistence is recognized as a
stable dispositional trait, hence its presence in many personality tests (Digman, 1990). In
spite of Corno’s work, persistence remains generally identified with motivation. This is
why we decided to adopt the label ‘‘motivation-related constructs’’ to designate, as a group,
the three independent variables (IM, EM, P) chosen to represent the motivational aspect of
this study.

3.2. Design particularities

The following methodological inputs were introduced to reinforce the correlational


design: (a) two measures of cognitive abilities were used; (b) well-validated instruments
were chosen to measure the three motivation-related constructs; (c) three sources of
information (students, parents, and teachers) participated in the assessment of the
motivation-related constructs, and (d) their ratings were collected twice during the
semester; (e) a large sample was sought to maximize statistical power. Based on previous
work by Rushton, Brainerd, and Pressley (1983), as well as Epstein (1986), we hoped to
improve the reliability of both independent and dependent variables by aggregating data
not only over time, but also over sources and instruments. Three hypotheses were
formulated: (a) academic aptitudes will be, by far, the best predictor of scholastic
achievement; (b) indices of motivation will add a statistically significant, but substantially
small contribution to that predictive power; (c) aggregating data over, sources, constructs,
and especially time, will slightly increase the unique predictive power of the motivation-
related constructs.

4. Method

4.1. Subjects

Two hundred eight female students from nine classes in Grade 8 of an all-girl high school
in the Greater Montreal area participated in the study; the students’ age varied from 12.3 to
14.10, with a mean of 13.5 (S.D. = 0.38). One parent of each student and a group of 15 of
F. Gagné, F. St Père / Intelligence 30 (2001) 71–100 81

their teachers were also involved; parents had to assess the motivation and persistence of their
child, while teachers did the same for groups of 25 to 30 participating students. The research
design called for independent evaluations by two different teachers.

4.2. Instruments

4.2.1. Intelligence
Two IQ measures were collected, using (a) Raven’s Progressive Matrices (Raven &
Summers, 1986) and (b) the Otis–Lennon Mental Ability Test: Intermediate Form (Otis &
Lennon, 1969). The Raven is a nonverbal test of inductive reasoning, recognized as one of the
purest measures of g (Snow, Kyllonen, & Marshalek, 1984). It is not timed; its administration
requires approximately 30 min. Good coefficients of reliability and validity have been
reported (Raven & Summers, 1986). The Otis–Lennon is a well-known group test of
academic aptitude (mostly verbal), with 80 multiple-choice items and a time limit of 45 min;
adequate reliability and validity indices have been found (Otis & Lennon, 1969).

4.2.2. Motivation
We chose a recent self-administered test of intrinsic–extrinsic school motivation, based on
Deci and Ryan’s (1985) theoretical framework. Created by Vallerand, Blais, Brière, and
Pelletier (1989), it is called Échelle de Motivation en Éducation: Études Secondaires [School
Motivation Scale: High School form]. Its 28 items are divided into seven 4-item subscales
(see Appendix A for sample items), measuring three types of intrinsic (IM) motivation
(S: stimulation, K: knowledge, A: accomplishment), three types of extrinsic (EM) motivation
(Ex: external, Ij: introjected, Id: identified), as well as nonmotivation (Nm). Respondents
judge the self-applicability of the statements with a 7-point Likert-type scale (1 = Does not
correspond to me at all; 7 = Is just like me). It takes approximately 15 min to complete. The
authors developed the following equation to create a global Relative Autonomy Index that
ranges from 18 to + 18: [(2(S + K + A)/3) + Id] [(Ex + Ij )/2 + (2Nm)] (see Vallerand &
Bissonnette, 1992, for a slightly modified version of the equation). That equation assumes a
positive correlation between one EM subscale (Id ) and the three IM subscales, as well as
negative correlations between the IM group and the other EM subscales plus the Nm subscale.
In other words, the IM and EM constructs are not conceived as independent, but as the two
poles of a continuum. The higher the score, the more the student is judged to attend school
out of sheer pleasure, interest, and personal choice. The scale’s wording was slightly modified
for the parent form. Because the teachers had to evaluate a whole group of students, a much
shorter form was needed for this scale. Three items were kept, one for IM, EM, and Nm,
respectively (see Appendix A); the same response scale was proposed.

4.2.3. Persistence
A 10-item scale, called Ma Fac˛on de Travailler à l’École [How I Work at School], was
created (see Appendix A for sample items). It was inspired by a similar scale in Le Test de
Tendance Personnelle, a French Canadian adaptation of the Edwards Personal Preference
Schedule (Gauthier, 1964). The same 7-point Likert-type self-applicability response scale was
used. The scale takes about 5 min to complete; a global score is computed, namely the mean
82 F. Gagné, F. St Père / Intelligence 30 (2001) 71–100

of the 10 answers after inverting responses to negatively worded statements. Again, minor
wording changes were made for the parent form. The teacher questionnaire included one item
to measure persistence (see Appendix A). The scale was pretested with a group of 70 students
from a different grade in the same high school. The item means and variances were judged
very satisfactory, as well as homogeneity (Cronbach’s a=.82). A principal components
analysis confirmed its unidimensionality (St Père, 1997).

4.2.4. Academic achievement


The four major subject matters taught in that grade were chosen: French (mother tongue),
Math, English, and History. The correlations between them ranged between .45 and .74, with
a mean of .56, and justified pooling them into a single academic achievement score expressed
in percentages.

4.3. Procedure

About 360 students in the same grade were invited to participate in the study; they and one
parent had to agree to complete the motivation and persistence scales twice during the
semester, and allow access to IQ and achievement data. The participation rate was 58%
(N = 208). Participating students received the Raven during a class period about a month after
the beginning of the fall semester; the Otis–Lennon IQ was taken from the students’ files.
Soon after, they completed both the motivation and persistence scales, and were given copies
to be completed by a parent. The teachers also received the short 4-item questionnaire with
which they had to evaluate the participating students in their class; 3 teachers out of 15 had to
evaluate two groups. As planned, we obtained for each student two independent teacher
ratings. Two months later, toward the end of the semester, the students, the parents, and the
teachers again completed the motivation and persistence scales. At the end of the data
collection, we had complete data for 205 students, 181 parents (9% loss), and 14 teachers.
Academic achievement data were extracted from the school files for two consecutive
assessment periods, one in the middle of the semester, the other at the end.

5. Results

We will first examine basic statistical information concerning the various instruments;
because of its structural difference, the teacher questionnaire will be analyzed separately. We
will then look at the relationships between the ability- (IQ and achievement) and motivation-
related constructs before assessing the separate contributions of aptitude and motivation to
school achievement.

5.1. Descriptive analyses

5.1.1. Determining independent variables


Table 1 shows descriptive statistics for all the variables included in the data analyses.
Percentiles were chosen for both the Otis–Lennon and Raven tests; they reveal a well-
F. Gagné, F. St Père / Intelligence 30 (2001) 71–100 83

Table 1
Basic descriptive and psychometric statistics
Variables Mean S.D. N Homogeneity
Abilities
Raven Matrices 74.7 22.08 203 –
Otis – Lennon 81.8 19.72 197 –
Academic achievement 1 79.4 9.12 208 .80
Academic achievement 2 80.1 8.36 208 .86

Motivation-related variables
Student data
IM 1 4.08 1.12 208 .91
IM 2 3.88 1.21 205 .94
EM 1 5.46 0.99 208 .77
EM 2 5.52 0.98 205 .80
P1 4.80 1.00 208 .81
P2 4.68 1.06 205 .86
Parent data
IM 1 4.74 1.05 200 .92
IM 2 4.63 1.03 176 .94
EM 1 5.47 1.01 200 .76
EM 2 5.52 1.02 176 .85
P1 5.26 1.02 203 .87
P2 5.28 1.01 181 .88
Teacher data (1 item per construct)
IM 1 4.89 0.99 207 –
IM 2 4.80 1.11 207 –
EM 1 4.66 0.97 207 –
EM 2 4.67 1.04 207 –
P1 4.97 1.11 207 –
P2 4.72 1.16 207 –
Intrinsic motivation (IM), extrinsic motivation (EM), and persistence (P). 1, 2 = time periods. Teacher data
correspond to pooled independent assessments by two teachers.

above average population, with about 45% of the scores on both tests exceeding the 90th
percentile. In the case of the school motivation scale, correlation matrices of the subscales
using the four data sets (two groups, two time periods) revealed that the authors’ equation
for the computation of a global score could not be used because the assumption of
negative correlations between the three IM subscales and two EM subscales (Ij, Ex) was
not met. Four independent principal components analyses with Varimax rotations were
then performed using the default options of SPSSx (Norusis, 1983), revealing three
recurring factors. The first was called IM and grouped 16 items (S, K, A, and Id); the
second was called EM because of high loadings from the 8 Ij and Ex items; the last one
brought together the 4 Nm items. Due to very low item means and variances (see Fig. 1),
that last factor was not retained. Thus, two scores, IM and EM, were created by averaging
responses to the relevant items. The Table 1 data show a definite leniency effect, with
most means about one unit above the scale’s midpoint of 4.00; we also observe good
84 F. Gagné, F. St Père / Intelligence 30 (2001) 71–100

Fig. 1. Mean ratings of two groups for the seven IM/EM subscales, and of the three groups for the three
motivation-related constructs (Times 1 and 2 pooled).

variation (S.D.), and satisfactory homogeneity coefficients (Cronbach’s a) for group


comparison purposes.
Results from the persistence scale reveal a similar leniency leaning. The homogeneity
coefficients are equally satisfactory. Four separate principal components analyses, again using
the default options of SPSSx and Varimax rotations, were performed (two groups, two times);
the unipolar structure of the P scale was confirmed. Descriptive data for the 4-item teacher
questionnaire appear at the bottom of Table 1. We first checked the comparability of the
ratings given by the two teachers, and found just a few significant mean differences, as well
as moderate correlations (from .18 to .51); teachers did agree better on their IM and P ratings
than their EM ratings (St Père, 1997). The two sets of ratings were judged similar enough to
be pooled; Table 1 shows the pooled data. In line with the student and parent data, very low
means and lack of variance in the rating of nonmotivation forced its exclusion from
further analyses.

5.1.2. Stability and validity


With its three sources of information on three different motivation-related constructs, as
well as repeated measures over two time periods, the research design makes it possible to
answer important questions concerning the reliability and validity of the data. How good is
the short-term stability of the respondents’ ratings? How good is the convergent validity,
which corresponds to the correlation between measures of a given construct by different
‘‘methods’’ (here sources of ratings)? How good is the discriminant validity, which
corresponds to the correlation between measures of different constructs (here IM, EM, P)
by the same source? Answers to these three psychometric questions appear in Table 2, which
adopts the format of Campbell and Fiske’s (1959) multitrait – multimethod matrix of
F. Gagné, F. St Père / Intelligence 30 (2001) 71–100 85

Table 2
Partial group by construct multitrait – multimethod matrix of correlations
Students Parents Teachers
IM EM P IM EM P IM EM P
Students
IM (.77) .13 .43 .33 .01
EM .11/.03 (.76) .02 .23 .15
P .45/.44 .12/ .05 (.76) .22 .00

Parents
IM .32/.42 (.78) .36 .39 .07
EM .12/.20 .35/.30 (.68) .01 .15
P .26/.38 .48/.44 .12/.09 (.74) .10

Teachers
IM .15/.14 .19/.15 (.70) .36 .81
EM .07/.03 .19/ .08 .50/.56 (.46) .01
P .18/.24 .39/.22 .86/.80 .38/.51 (.70)
Values within parentheses along the main diagonal are short-term (2 months) stability coefficients. Validity
coefficients from time periods 1/2 data are placed under the diagonal; the aggregated values (Times 1 + 2) appear
above the diagonal. Discriminant validity coefficients surround the diagonal, whereas convergent validity
coefficients appear in italics. N = 169. Significance levels: r  .15, P < .05; r  .20, P < .01.

correlations. Short-term stability coefficients are placed, within parentheses, on the main
diagonal. Only the lower triangle would be used normally to insert validity coefficients; but,
because of our repeated measures, we placed Times 1 and 2 coefficients below the diagonal,
and the pooled (1 + 2) coefficients above. The three blocks clinging to the main diagonal
correspond to discriminant validity coefficients, what Campbell and Fiske called heterotrait–
monomethod data. Then, further away from the diagonal, there are, on each side, three trios of
diagonally placed values, called monotrait – heteromethod coefficients, or measures of
convergent validity. To avoid cluttering the table, we did not include the heterotrait–
heteromethod data that would have filled the table; as expected, almost all of them were
statistically nonsignificant or just barely so ( P < .05). Absent from Table 2, the short-term
stability of the achievement scores was .89. The short-term stability coefficients are much
lower than the corresponding homogeneity coefficients (see Table 1), but, except for the
teachers’ EM scores, they are high enough for group comparison purposes. They are
somewhat lower in the case of the teacher questionnaire, but still surprisingly high for
correlations between single items.
Ideally, discriminant validity coefficients should be as close to zero as possible, except
when there is reason to believe that a pair of constructs are conceptually related. The three
groups behave somewhat differently; but each group’s pattern remains almost the same
over the two time periods. The students’ EM scores entertain no relationship whatsoever
with either IM or P, but these last two are partially correlated (.45/.44). In continuity with
their children’s responses, the parents’ IM and P scores are moderately correlated, while
EM and P are independent; but, the parents perceive a low-to-moderate positive relation-
ship between intrinsic and extrinsic motives (.35/.30). In contrast, the teachers’ discrim-
86 F. Gagné, F. St Père / Intelligence 30 (2001) 71–100

inant validity coefficients are much higher; EM scores are moderately correlated with both
IM and P, whereas the IM/P correlations are so high (.86/.80) that the two measures
become undistinguishable. Ideally, the convergent validity coefficients should be almost as
high as reliability allows; indeed, since the three groups are assessing the same motiva-
tional phenomena — the students’ IM, EM, and P, — these coefficients could be almost
regarded as scorer reliability measures. The results are far from ideal. Whatever the time
period, the three EM measures entertain no relationship between them. The coefficients for
the IM and P measures are larger, but only a few of them barely reach the moderate level
(r2  .10). The student–parent correlations are the least weak on average, especially with
regard to IM.

5.1.3. Group comparisons


The means in Table 1 indicate that the three groups might differ quite significantly in terms
of their evaluations of the three motivation-related constructs, and that there might be also
time effects. To answer these questions, the 18 IM, EM, and P variables in Table 1 were
analyzed with a 3  3  2 (Group  Construct  Time) ANOVA, the three factors designated
repeated measures. The group variable was considered a within subjects variable for the
simple reason that the parents and teachers were assessing the students’ motivations, thus
making the three sets of data interrelated with each other. The results appear in Table 3. The
major main effect, a global difference between the means of the three motivational constructs,

Table 3
Repeated measures ANOVA comparing groups, constructs, and time periods
Source df F h2
Between subjects
The groups were introduced as a within subjects variable; see text.

Within subjects
Groups (G) 2 17.61*** .10
S within-group error 334 (2.49)
Constructs (C) 2 77.57*** .32
S within-group error 334 (1.61)
Time (T) 1 9.54** .05
S within-group error 167 (0.41)
GC 4 66.92*** .29
S within-group error 668 (0.94)
CT 2 7.12*** .04
S within-group error 334 (0.27)
GCT 4 2.73* .02
S within-group error 668 (0.27)
Values enclosed in parentheses represent mean square errors. Only statistically significant interactions are shown.
* P < .05.
** P < .01.
*** P < .001.
F. Gagné, F. St Père / Intelligence 30 (2001) 71–100 87

can be deduced from the three curves on the right side of Fig. 1. Higher ratings are given to
EM items (5.22) than to P items (4.96), whereas IM items receive the lowest ratings (4.53).
To show the reasons for the higher popularity of EM over IM, we have included on the left
part of Fig. 1 the means of the seven subscores — aggregated over time — from both the
student and parent questionnaires. EM.Id (going to school to find a career I like and become
competent in it) is judged by both groups to be the most relevant motive, whereas IM.S
(learning for the fun and excitement) receives the lowest ratings from both groups. All other
motives share generally comparable ratings.
The second most significant main effect involves the three groups, with the parents
expressing higher ratings overall (5.15) than either the teachers (4.82) or the students (4.76).
The last main effect confirms that Time 1 ratings were slightly more positive than those made
2 months later (4.94 vs. 4.87). Fig. 1 shows the strongest interaction effect, the group by
construct interaction. For each construct, a different group distinguishes itself from the two
others, students in the case of IM, teachers in the case of EM, and parents in the case of
persistence. The statistically significant, but not very substantial (h2=.04) construct by time
effect is caused by an identical drop over time ( .12) in the IM and P ratings, counter-
balanced by a small rise (.02) of the EM ratings. These changes can be easily extracted from

Table 4
Correlations of the three motivation-related constructs, by group, with the three ability measures, by time
Raven Otis – Lennon Academic achievement
Time 1 Time 2 Time 1 Time 2 Time 1 Time 2
Aptitudes
Otis – Lennon .32*** .53*** .56***
Raven .37*** .36***

Students
IM .02 .02 .04 .03 .03 .06
EM .05 .03 .00 .00 .05 .01
P .12 .16 .07 .08 .25** .19*

Parents
IM .08 .06 .07 .10 .21** .21**
EM .05 .04 .07 .02 .12 .01
P .10 .18* .12 .15 .36*** .35***

Teachers
IM .22** .17* .25** .33*** .61*** .62***
EM .21** .15 .26** .35*** .43*** .43***
P .14 .06 .27** .25** .61*** .57***
N = 156. Times 1 and 2 for the Raven and the Otis – Lennon refer to the corresponding measures of IM, EM, and P.
In the case of academic achievement, the Times 1 and 2 measures are paired with the corresponding motivation-
related measures.
* P < .05.
** P < .01.
*** P < .001.
88 F. Gagné, F. St Père / Intelligence 30 (2001) 71–100

the Table 1 data. Finally, the significant G  C  T interaction effect is due to the fact that the
IM drop over time is stronger in the students’ ratings, whereas the drop in the P ratings comes
mostly from the teachers’ ratings (see Table 1).

5.2. Predicting school achievement

As a first descriptive survey of the relationships between the independent variables and
academic achievement, Table 4 shows the correlations between the three sets of motivation-
related measures and the three ability measures. First, the well-documented relationship
between cognitive abilities and academic achievement is evident, especially in the case of the
Otis–Lennon. Note the barely moderate correlation between the two cognitive ability tests,
announcing possible independent contributions. Second, the students’ self-assessments of
their motivation and persistence at two points in time show no relationship whatsoever with
their cognitive abilities; moreover, only the persistence measures weakly predict academic
achievement. The two sets of parent judgments are also unrelated to the child’s academic
aptitudes, but the IM and P scores are significantly correlated with the students’ grades.
Surprisingly, it is the teachers’ ratings, made with a psychometrically very fragile instrument,
which produce by far the ‘‘best’’ results, not only as correlates of cognitive aptitudes, but
even more as predictors of academic achievement. Based on (a) these suspiciously high
correlations, (b) the moderate to high correlations between the three motivation-related
measures obtained from the teachers (see Table 2), and (c) the lack of convergent validity
between the students’ and teachers’ assessments, it became clear that the teachers’ ratings
were unduly influenced by their knowledge of the students’ achievements. The construct
validity of these ratings appeared so doubtful that we decided to exclude them from the last
analytical step.
To verify the three hypotheses mentioned earlier, a series of multiple regression analyses
were performed. After excluding the teacher data, two sets of six motivation-related variables
(3 constructs  2 groups) were available for each time period, plus the two academic aptitude
tests. A semihierarchical model was adopted (Norusis, 1983). The Raven and Otis–Lennon
scores were entered as Step 1, in that sequence. We judged that the purest available measure
of cognitive abilities should have priority over an instrument that had been specifically
designed to maximize its predictive validity with regard to scholastic achievement (Anastasi
& Urbina, 1997). Then, assuming that the students were the best judges of their own
motivation and persistence, we entered the students’ IM, EM, and P scores as a second step,
without specifying a particular order among them. The last step brought in the three parental
measures of motivation, again in no specific order. We did not use the stepwise procedure in
SPSSx (Norusis, 1983), but the ‘‘forward’’ command so that a previously retained variable
would not be deleted later, in the same step or in a subsequent step. Four successive multiple
regressions were performed, first for Times 1 and 2 separately, then with the student and
parent data separately aggregated over time, and finally by also aggregating the two aptitude
measures. The basic assumptions of regression analysis (linearity, normality, and homo-
scedasticity) were checked and found to be respected; no outliers were discovered
using the Mahalanobis distance index (Tabachnick & Fidell, 1989). The results appear
in Table 5.
F. Gagné, F. St Père / Intelligence 30 (2001) 71–100 89

Table 5
Multiple hierarchical regressions of ability and motivation-related indices on school achievement
Variables B b sr2 R & R2
Time 1
1. Raven Matrices 0.085 .204 .13*** R = .63
2. Otis – Lennon 0.185 .388 .17*** R2 = .40
3. Persistence (students) 1.120 .120 .04* adj. R2 = .39
4. Persistence (parents) 2.390 .271 .06***

Time 2
1. Raven Matrices 0.067 .171 .14*** R = .65
2. Otis – Lennon 0.205 .466 .22*** R2 = .42
3. Persistence (parents) 2.136 .257 .06*** adj. R2 = .41

Times pooled
1. Raven Matrices 0.071 .179 .15*** R = .68
2. Otis – Lennon 0.203 .455 .21*** R2 = .46
3. Persistence (students) 0.684 .075 .03 adj. R2 = .45
4. Persistence (parents) 2.555 .288 .07***

Times and aptitudes pooled


1. Raven + Otis – Lennon 0.263 .506 .33*** R = .66
2. Persistence (students) 0.618 .068 .03 R2 = .43
3. Persistence (parents) 2.547 .288 .07*** adj. R2 = .42
* P < .05.
*** P < .001.

The first regression, using the Time 1 data, shows that (a) the two aptitude measures account
for 30% of the academic achievement variance, (b) only persistence scores (students and
parents) add any significant contribution to the prediction equation, and (c) the ratio of their
respective contributions is 3:1. Note that the parents’ assessment of their child’s persistence
appears a more powerful predictor of academic achievement than the students’ own judgment,
even though we gave priority to the students’ scores. The second regression equation, which
uses end of semester measures of motivation, persistence, and achievement, generally
confirms the first one, except that the students’ P scores no longer contribute significantly
to the prediction of academic achievement. Here, the aptitude/motivation ratio for predictive
power is 6:1. The aggregation of the two successive measures slightly improves the predictive
validity; the squared multiple correlation increases from .42 to .46. This is mainly attributable
to the reapparition of the students’ P scores among the predictors. But, note that while their
contribution had to be statistically significant when these scores were entered in Step 3, the
significance disappears when the parents’ P scores are included in the equation. In other words,
if we had used the stepwise procedure instead of the forward one, the students’ P scores would
have been deleted from the final regression equation. The fourth multiple regression attempted
to verify if we could improve the predictive validity by aggregating the two aptitude measures;
the result was negative. Similarly, we tried aggregating the student and parent data and found
no significant change in predictive power (R2 change from .43 to .42). In brief, the best
90 F. Gagné, F. St Père / Intelligence 30 (2001) 71–100

predictive equation, with both times combined, gives aptitudes at least five times (.36/.07)
more predictive power than any combination of motivation-related measures.

6. Discussion

We will first discuss the results related to the three hypotheses, then examine some
peripheral yet interesting observations.

6.1. The three hypotheses

We predicted (a) that cognitive aptitudes would be a strong predictor of school achieve-
ment, (b) that motivation would also contribute, but to a much smaller degree, to the
prediction equation, and (c) that some slight improvement in predictive power would occur
thanks to data aggregation. Only the first hypothesis was clearly confirmed.

6.1.1. Aptitudes
Together, the Raven Matrices and Otis–Lennon produced a multiple correlation in the .55
to .60 range depending on the data set. The hierarchical method advantages the Raven since
it receives the variance shared with the Otis–Lennon; in a stepwise approach, it ranks
second, well behind the Otis–Lennon, and its contribution in terms of explained variance
drops slightly below 10%. Still, that unique contribution remains at least as important as that
of any motivation-related variable assessed in this study. As noted in the literature review,
such results are quite common, especially in the case of the Otis–Lennon, whose mostly
verbal items were chosen to maximize its ability to predict school achievement. The fact
that the nonverbal inductive reasoning measured by the Raven Matrices contributes
significantly, albeit more marginally, to the prediction was hoped for, but not strongly
expected; that positive effect reinforces the strength of general cognitive abilities as a
predictor of academic achievement.

6.1.2. Motivation
The second hypothesis was just barely confirmed, thanks to a statistically significant
contribution of the persistence measures, especially the parents’ P scores. Their unique
contribution ranged between 6% and 10%, approximately equivalent to that of the Raven
Matrices. Compared to the explanatory power of aptitudes, it is, at best, three or four times
smaller. Again, such a result is perfectly in line with a majority of the past studies reviewed
(e.g., Aspinwall & Taylor, 1992; McHenry et al., 1990; Schmidt & Hunter, 1998; Spence
et al., 1989; Wong & Csikszentmihalyi, 1991). But, there is a problem; the construct validity
of that small significant contribution of the parents’ persistence scores is doubtful. Let us
assume, as we did for the hierarchical multiple regressions, that the students’ P scores are the
most valid representation of the ‘‘real’’ persistence of these high school students; it is a very
plausible assumption in view of the distancing process between parents and children at the
start of adolescence. Keeping that assumption in mind, we observe first that the students’ and
parents’ persistence measures, at both points in time, are not even moderately correlated (see
F. Gagné, F. St Père / Intelligence 30 (2001) 71–100 91

Table 2). Moreover, as shown in Table 4, it is the parents’ P ratings that better predict school
achievement; in terms of percentage of (zero-order) explained variance, that difference is
large: about 5% for the students’ P scores, as opposed to 12% for the parents’ P scores. When
both variables are entered in the multiple regression, the students’ P scores, even though they
are given priority, do not contribute significantly (except slightly in the Time 1 data set) to the
prediction of academic achievement; that priority makes them benefit from any shared
predictive power with the parents’ P scores.
Why do the parents’ P ratings predict school achievement better than the students’
theoretically more valid self-ratings? The most plausible hypothesis is a contamination of
the parents’ perception of their child’s motivation by their knowledge of her achievement, an
effect identical to the much stronger one observed in the teacher data. There is a very strong
common sense belief that persistent effort and success are closely related. It shows not only in
popular sayings (e.g., Edison’s famous quote ‘‘Genius is 1% inspiration and 99% perspi-
ration,’’ or ‘‘If one has the will, one can succeed’’), but also in the scientific literature. For
instance, as mentioned earlier, research in the field of attribution theory has confirmed
repeatedly the major role of effort as a perceived cause of success. Consequently, if the
parents observe that their child is achieving, they will tend to readjust their persistence
judgments to a slightly higher level, doing the opposite when the child fails. These influences
are sufficient to inflate the correlation between their persistence ratings and achievement, and
create a low-level artificial causal relationship. Without such knowledge of the dependent
variable, these scores would probably correlate no better with school achievement than the
students’ own P judgments. That bias is quite common in psychometrics; its exact opposite,
called criterion contamination, happens when raters of job performance are influenced by
their knowledge of the ratee’s performance on selection tests (Anastasi & Urbina, 1997). In
summary, these results show no significant contribution of intrinsic or extrinsic motives,
whether assessed by the students themselves or rated by their parents, to the prediction of
academic achievement. Academically talented students are not more intrinsically motivated
nor are they more (or less) extrinsically motivated than their peers who achieve at average or
below average levels. Even the slightly significant impact of persistent efforts was questioned
because it came from the less valid parents’ ratings instead of the students’ self-perceptions.

6.1.3. Aggregation
The design of this study made possible three forms of data aggregation: over time, over
sources (students, parents, and teachers), and over constructs (aptitudes). When we
hypothesized that aggregation would slightly improve the overall predictive validity of the
motivation-related constructs, we targeted time — because of usually strong stability
coefficients — as the most promising of the three types. Aggregation works best when
the covariance is large and reflects the same true score, while the rest of the variance is just
random measurement error instead of the nonrandom measure of a different construct.
Pooling the three successive motivation-related measures for both students and parents did
increase the multiple correlation from .65 to .68, a substantially limited improvement. The
two other aggregation attempts proved unsuccessful. When the two aptitude scores were
merged into a global aptitude percentile, that construct’s predictive power slightly decreased
(see Table 5); a similar decrease was observed when we pooled the students’ and parents’
92 F. Gagné, F. St Père / Intelligence 30 (2001) 71–100

measures of the three motivation-related variables. It must be pointed out that in both cases
the correlations were low (usually < .35), barely high enough, according to Epstein (1986),
to warrant such aggregation. Moreover, the strong bias in the parents’ ratings did not help.
All in all, these results indicate that the success of an aggregation strategy rests on rather
demanding criteria.

6.1.4. Methodological considerations


It is not the goal of the present study to explain why motivation or persistence contribute so
little to the prediction of scholastic achievement; we leave such reflections to motivation
theorists. We only aimed to show, as so few others did before us, that the predictive power of
these constructs was limited. And the results are clear, especially in view of efforts to ensure
reliable and valid measures of all the relevant constructs. Granted that the teacher data lack
validity, and that the parental information is also partly questionable. But, these two sources
were included as independent estimates of the ‘‘real thing,’’ namely the students’ motivation
and persistence. Consequently, their exclusion does not jeopardize the methodological
strength of this study. In fact, the comparisons between sources lead to important observations
about the value of outside judgments (see below). The core data are the aptitude measures and
the students’ own judgments of their motivation and persistence; and we believe that our
instruments stand comparison with the best ones used in past studies. First, the Otis–Lennon
and Raven Matrices are both well-validated tests of cognitive abilities. Some could argue that
we had a skewed distribution with a large percentage of high scores, but reduced variation
attenuates correlation coefficients. So, we could argue that our multiple correlation of .60 is a
lower-bound estimate of the true relationship. Second, the IM, EM, and P scores show good
reliability; the short-term stability coefficients are somewhat lower than the corresponding
homogeneity coefficients, but still high enough for group comparisons. Third, in terms of
content validity, the IM and EM scales were created by Vallerand et al. (1989) from a
respected theory of school motivation (Deci & Ryan, 1985). Similarly, the P scale was
inspired by a valid scale in a well-recognized personality test (Edwards, 1959; Gauthier,
1964). Finally, the discriminant validity data confirmed the independence of the constructs,
except for a theoretically defensible moderate relationship between IM and persistence; it
makes sense that as students find more pleasure in academic learning, they will tend to invest
more energy to understand and master the more difficult concepts. In brief, we doubt that the
lack of predictive power could be attributable to invalid measurement of the key concepts.

6.2. Additional observations

6.2.1. The nature of school motivation


In their own judgment, students attend school more for extrinsic motives than for intrinsic
ones (see Fig. 1); not that learning and discovering are not interesting to some extent for many
of them, but these motives are judged less important than the more practical goals of finding a
career that will be interesting (EM.Id) and well-paid (EM.Ex). Among intrinsic motives, IM.S
(the excitement of learning) gets the lowest mean; nonmotivation is almost nonexistent, but a
large S.D. (1.03) suggests that some students might be seriously questioning their schooling
process. The parents’ ratings follow a very similar pattern, but they perceive their child as
F. Gagné, F. St Père / Intelligence 30 (2001) 71–100 93

more motivated intrinsically and more persistent on average than the students perceive
themselves. The more positive parental assessments of IM and persistence can be interpreted
in different ways. First, they could represent the observable part of the child’s motivation,
what she decides to show publicly to her parents. It is understandable that these students
would be hesitant to manifest outwardly too much lack of interest or persistence, not only to
preserve their own self-esteem, but also as a form of thoughtfulness toward parents who are
meeting the added cost of a private education. Second, it could be the result of a parental
clemency bias brought out by their limited knowledge of their child’s motivation level; unsure
about the ‘‘real’’ level, they would give her the benefit of the doubt. Third, it might also
reflect selective perception on the part of the parents who want to believe that the choice of a
costlier education has a positive impact on their child’s IM and persistence.
The tendency to identify more readily with extrinsic than with intrinsic motives is not
specific to this sample, but was observed by the authors of the instrument (Vallerand et al.,
1989). In view of the relative loftiness of intrinsic motives compared to extrinsic ones — as
shown so clearly in Deci and Ryan’s (1985) use of them as endpoints on a self-determination
continuum, with IM placed at the ‘‘good’’ end, — one would expect that the students,
influenced by social desirability pressures, would have given higher ratings to IM motives.
The fact that they did not does not necessarily mean that social desirability had no impact.
The students’ ratings might still be somewhat inflated; only a more in-depth approach would
reveal the extent of that potential bias. Finally, the significant but small decrease in motivation
over the course of the semester is a phenomenon well known to teachers and professors.
Because of its minor importance, we did not check whether it had been scientifically
documented. It is interesting that the students report a decrease in interest (IM), whereas the
teachers perceive a decrease in persistence. Could the teachers’ focus result from the better
observability of persistence in daily school life?

6.2.2. Motivation and aptitudes


The total lack of relationship between the students’ IQ and their motivational level is
another important result of this study. Not a single one of the 12 correlation coefficients (see
Table 4) attains minimal statistical significance, not to say minimal substantial significance
(r2  .05). These results contradict a very common belief among educators, namely that
intellectually gifted students show more IM and task commitment than average ability
students. At the same time, they confirm Shore et al.’s (1991) questioning of that belief.
Persistence is the only variable that predicts, and just barely, academic achievement. On
second thought, that lack of relationship can be easily explained. It is a well-known fact that
the high school curriculum offers little challenge to bright students, a fact that has been amply
demonstrated in the U.S. (Archambault et al., 1993; Reis et al., 1993); now, the Quebec
curriculum largely overlaps its U.S. counterpart. The lack of challenge impacts on both IM
and persistence. It decreases the desire to learn in those students who were hoping for new
and fascinating knowledge presented at a fast rate. In the framework of Vygotsky’s (1978)
theory, they are seldom working at the upper limit of their zone of proximal development. On
the other hand, bright students can achieve very well without even trying. It is clear that the
above observations do not apply to all students. Some of them maintain IM even in dreary
school environments, in the same way that some high achievers show high persistence and
94 F. Gagné, F. St Père / Intelligence 30 (2001) 71–100

effort even when they could achieve almost as well with little involvement. In other words,
there are complex phenomena in action as well as important individual differences, and their
effect is to minimize the relationship between IQ and motivation, as well as between
motivation and academic achievement.

6.2.3. The intrinsic/extrinsic continuum


As quoted earlier, Deci and Ryan (1985) present IM as the most desirable form of
motivation, corresponding to self-determination and maturity; EM is judged to be a less
evolved or mature form. Following that theoretical position, the authors of the scale we used
computed from their instrument a single global score, called a self-determination score or
relative autonomy index, in which the three IM subscales (A, K, S ), plus the EM.Id subscale,
had positive weights, whereas the two other EM subscales (Ex, Ij) and Nm were given
negative weights. In other words, the more self-determined a person is, the more importance
she should give to intrinsic motives. The fact that the EM.Id items side with those of the three
IM subscales to create an IM factor already signals some ambiguity in the IM vs. EM
distinction. It seems clear from the participants’ answers, both students and parents, that
going to school to find an interesting occupation and become competent in it (EM.Id) is
closely related to finding pleasure and interest in acquiring new knowledge (IM.K), and
wanting to become more competent and achieving as a learner (IM.A). The means-to-end
orientation measured by EM.Id clearly appeals more to both students and parents. It is also
interesting that the IM and EM constructs, contrary to theory, are not negatively correlated.
The students implicitly see them as independent, while the parents’ ratings make them
positively correlated. These results show that it is possible for intrinsic and extrinsic motives
to be pursued in parallel, for the intrinsic pleasure of learning to coexist with the desire to find
a well-paying job (EM.Ex) or to prove to oneself that one can attain the goal of a high school
diploma (EM.Ij). This is not the place for a full-fledged reexamination of Deci and Ryan’s
self-determination continuum, but our results clearly question that central aspect of their
theory. Moreover, other recent empirical work (e.g., Hoekman, McCormick, & Gross, 1999;
Nicholls, 1992) supports that questioning.

6.2.4. The validity of external judges


The present study has shown that using outside evaluators to assess a person’s internal
state of mind can be a risky business, the more so when the evaluators have limited direct
information about the traits or behaviors they are asked to assess. We were hoping for
moderately high correlations (>.50) between pairs of scores, but found that most of them were
not even statistically significant at the .01 level (see Table 2). Teachers were especially poor at
assessing both motivation and persistence; their judgments had virtually nothing in common
with the students’ self-perceptions nor with the parents’ judgments. The situation was even
worse when the measures were pooled over time. Note for example that the moderate
correlations between the teachers’ EM and P ratings (.38/.51) disappear when the Times 1 and
2 data are pooled (see Table 2), a sure sign of low reliability. Even though we had to expect
reduced stability coefficients because only one question per construct was used, we did not
expect (a) the unduly high discriminant validity coefficients observed, (b) the quasi-total lack
of convergent validity with the student data, as well as (c) suspiciously high correlations with
F. Gagné, F. St Père / Intelligence 30 (2001) 71–100 95

achievement measures. These results, totally divergent with those of the two other sources,
brought us to hypothesize that the teachers, because of their limited information about the
students’ thoughts and feelings, relied heavily on the only clear knowledge available, namely
school grades, to infer the students’ level of motivation (IM and EM), as well as persistence.
Indeed, if we had included the teachers’ scores in a stepwise multiple regression analysis,
their IM and P scores would have dethroned the two IQ measures from their top rank as
predictors of academic achievement (see Table 4). In summary, it is a clear case of
‘‘hindsightful’’ prediction; and we doubt that a longer instrument would have much improved
their ratings. Such bad results might be partly explained by the fact that our judges were
specialist teachers and met their students for only a few hours (3 to 8) each week. Still, we
doubt that elementary school teachers, who work full-time with the same students, would fare
much better as judges of these complex constructs. In a nutshell, there is much ground for
caution in using teachers’ or parents’ assessments of the students’ internal states, like mood,
needs, or goals, in fact anything that cannot be easily observed.

7. Conclusion

The nonsignificant contribution of motivation to the prediction of school achievement


found in the present study corroborates results from many past investigations in both school
and work settings (e.g., Bloom, 1976; McHenry et al., 1990; Schmidt & Hunter, 1998;
Terborg, 1977). More significant, the students’ own judgments of their motivation and
persistence did not predict achievement even when cognitive aptitudes were not controlled.
In fact, the motivation-related measures were not even related to intelligence, contradicting a
common belief among educators and scholars (Gagné & Blanchard, submitted for publication;
Janos & Robinson, 1985). These results should be extrapolated with caution. First, the study
was conducted in an all-girls high school; boys might behave differently. Anastasi and Urbina
(1997) mentioned a small, but consistent moderating sex effect of the predictive validity of
aptitude tests. They speculated that male students might find more interest in nonacademic
activities; ‘‘these interest differences would introduce additional variance in their course
achievement and would make it more difficult to predict achievement from aptitude test
scores’’ (p. 183). Second, the sample came from a private school with somewhat selective
admission criteria, as shown by the high IQ scores. More regular school settings could lower
not only the aptitude measures, but also the motivation scores, creating more variance and,
possibly, increasing the predictive power of motivation-related constructs. Third, these results
were obtained in a school setting. Compared to other fields of talent development, like arts,
sports or business, schools have one special characteristic: attendance is compulsory, thus has
little to do with interest. By contrast, most youngsters do choose to practice sports, learn a
musical instrument, develop a talent in visual arts, and so forth. If they become bored, or do not
perform at an expected level, they can always abandon and switch to another activity.
Consequently, studies in school settings need to be compared with parallel ones in these
other fields, with special attention given to designs that control for field-related aptitudes.
Fourth, these results should not be extrapolated to talent development at exceptional levels of
excellence; at such levels, motivation and persistence seem to play a much more significant
96 F. Gagné, F. St Père / Intelligence 30 (2001) 71–100

role (Hemery, 1986; Simonton, 1994). For instance, Lewis Terman compared the 150 most
successful and 150 least successful participants of his immense longitudinal study of talent
development, when they attained mid-life; he found that they did not differ to any extent in
intelligence as measured by tests. But, there were significant differences: ‘‘the four traits on
which they differed most widely were ‘persistence in the accomplishment of ends,’ ‘integra-
tion toward goals,’ ‘self-confidence,’ and ‘freedom from inferiority feelings.’ In the total
picture the greatest contrast between the two groups was in all-round emotional and social
adjustment, and in drive to achieve’’ (Terman & Oden, 1959, pp. 148–149).
Finally, this study exemplifies the importance of intelligence as a causal agent in human
behavior. That causality has been demonstrated well beyond the school walls, in a large
variety of social phenomena, from unemployment, incarceration, and illegitimate births
(Gottfredson, 1997), to single motherhood, HIV infection, and credulity (Gordon, 1997). Yet,
IQ is almost never included as a covariate in social science research. Lubinski and
Humphreys (1997) made a strong plea for its regular inclusion, contending ‘‘that, if consulted
more often, the construct of general intelligence would contribute to understanding many
puzzling human phenomena, because successive gradations of intelligence reflect successive
degrees of risk’’ (p. 159). More scholars in education should heed that advice, and examine
the antecedent causal role of intelligence with regard to popular psychoeducational con-
structs, like academic self-concept, self-efficacy, confidence, self-regulation, and many
others. It would be interesting to see how much of their demonstrated relationship with
school achievement would remain if cognitive abilities were controlled. Such controlled
studies are hard to find; yet, they are badly needed if we want to refine our understanding of
the determinants of achievement and success, not only in school settings, but in all other
domains where youths and adults alike strive toward competence and excellence.

Acknowledgments

The authors sincerely thank the administrators, faculty, and students (as well as
their parents) of College Regina Assumpta, in Montreal, for their precious help in the
data collection.

Appendix A

A.1. Sample items from the school motivation scale

I go to school because. . .
S: I really like going to school.
K: I find much pleasure and satisfaction in learning new things.
A: of the pleasure I get from surpassing myself in my studies.
Id: high school studies will prepare me better for the career I have chosen.
Ij: I want to prove to myself that I can complete with success a high
school program.
F. Gagné, F. St Père / Intelligence 30 (2001) 71–100 97

Ex: I need at least a high school diploma if I want to find a well-paying job later.
Nm: honestly, I don’t know; I really feel like I’m losing my time in school.

A.2. The 4-item teacher questionnaire

IM: This student goes to school for the pleasure she feels when she is learning new
things; she shows much interest for school subjects.
EM: This student goes to school to get rewards (e.g., good grades, a better salary
later, a present promised by the parents).
Nm: This student does not know why she is going to school; she really feels that
she is losing her time.
P: When facing difficulties in school, this student persists; she keeps at it and puts
in more efforts to overcome these difficulties and attain her goal.

A.3. Sample items from the persistence scale

(1) When I begin a complex homework, I always give my best to complete it. (3) When I
meet with some difficulties in my school work, I abandon rapidly and do something else. (5)
When I have some difficulty with a homework, I take all the time it takes to complete it well.
(8) Sometimes, I do not complete difficult homework.

References

Anastasi, A., & Urbina, S. (1997). Psychological testing (7th ed.). Upper Saddle River, NJ: Prentice-Hall.
Archambault, F. X. Jr., Westberg, K. L., Browns, S. W., Hallmark, B. W., Emmons, C. L., & Zhang, W. (1993).
Regular classroom practices with gifted students: results of a national survey of classroom teachers. Storrs,
CT: National Research Center of the Gifted and Talented.
Aspinwall, L. G., & Taylor, S. E. (1992). Modeling cognitive adaptation: a longitudinal investigation of the impact
of individual differences and coping on college adjustment and performance. Journal of Personality and Social
Psychology, 63, 989 – 1003.
Atkinson, J. W. (1964). An introduction to motivation. New York: Van Nostrand Reinhold.
Austin, J. R., & Vispoel, W. P. (1998). How American adolescents interpret success and failure in classroom music:
relationships among attributional beliefs, self-concept and achievement. Psychology of Music, 26, 26 – 45.
Biddle, S. (1993). Attribution research and sport psychology. In: Singer M. Murphey, & L. K. Tennant (Eds.),
Handbook of research on sport psychology ( pp. 437 – 464). New York: Macmillan.
Bloom, B. S. (1976). Human characteristics and school learning. New York: McGraw-Hill.
Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant validation by the multitrait – multimethod
matrix. Psychological Bulletin, 56, 81 – 105.
Carroll, J. B. (1963). A model for school learning. Teachers College Record, 64, 723 – 733.
Carroll, J. B. (1989). The Carroll model. Educational Researcher, 18 (1), 26 – 31.
Clark, B. (1996). Growing up gifted (3rd ed.). Columbus, OH: Charles E. Merrill.
Corno, L., & Kanfer, R. (1993). The role of volition in learning and performance. In: L. Darling-Hammond (Ed.),
Review of research in education, vol. 19 ( pp. 301 – 341). Washington, DC: American Educational Rese-
arch Association.
Davis, G. A., & Rimm, S. B. (1989). Education of the gifted and talented (2nd ed.). Englewood Cliffs, NJ:
Prentice-Hall.
98 F. Gagné, F. St Père / Intelligence 30 (2001) 71–100

Deci, E. L., & Ryan, R. M. (1985). Intrinsic motivation and self-determination in human behavior. New
York: Plenum.
Deci, E. L., Vallerand, R. J., Pelletier, L. G., & Ryan, R. M. (1991). Motivation and education: the self-determi-
nation perspective. Educational Psychologist, 26, 325 – 346.
Digman, J. M. (1990). Personality structure: emergence of the five-factor model. In: M. R. Rosenzweig, & L. W.
Porter (Eds.), Annual review of psychology, vol. 41 ( pp. 417 – 440). Palo Alto, CA: Annual Reviews.
Edwards, A. L. (1959). Edwards personal preference schedule. New York: Psychological Corporation.
Epstein, S. (1986). Does aggregation produce spuriously high estimates of behavior stability? Journal of Person-
ality and Social Psychology, 50, 1199 – 1210.
Feldhusen, J. F. (1992). Talent identification and development in education (TIDE). Sarasota, FL: Center for
Creative Learning.
Gagné, F. (1993). Constructs and models pertaining to exceptional human abilities. In: K. A. Heller, F. J. Mönks, &
A. H. Passow (Eds.), International handbook of research and development of giftedness and talent ( pp. 69 – 87).
Oxford: Pergamon Press.
Gagné, F. (1995). From giftedness to talent: a developmental model and its impact on the language of the field.
Roeper Review, 18, 103 – 111.
Gagné, F. (1999). My convictions about the nature of human abilities, gifts and talents. Journal for the Education
of the Gifted, 22, 109 – 136.
Gagné, F., & Blanchard, D. (submitted for publication). Comparing beliefs of educators and students in regular
education, music, and sports, concerning the major determinants of talent emergence.
Gallagher, J. J. (1985). Teaching the gifted child (3rd ed.). Boston: Allyn and Bacon.
Galton, F. (1869/1962). Hereditary genius: an inquiry into its laws and consequences. New York: Meridian Books.
Gauthier, G. (1964). Test de tendance personnelle. (Translation of Edward’s Personal Preference Schedule).
Montréal: Institut de Recherches Psychologiques.
Good, L. T, & Brophy, J. E. (1990). Educational psychology (4th ed.). New York: Longman.
Gordon, R. A. (1997). Everyday life as an intelligence test: effects of intelligence and intelligence context.
Intelligence, 24, 203 – 320.
Gottfredson, L. S. (1997). Why g matters: the complexity of everyday life. Intelligence, 24, 79 – 132.
Gottfried, A. E. (1979). Role of intrinsic motivation in children’s school achievement. Paper presented at the 87th
Annual Convention of the American Psychological Association, New York. ERIC Document Reproduction
Service No. ED 182 637. (Septenber) .
Gottfried, A. E., & Gottfried, A. W. (1996). A longitudinal study of academic intrinsic motivation in intellectually
gifted children: childhood through early adolescence. Gifted Child Quarterly, 40, 179 – 183.
Grabe, M., & Latta, R. M. (1981). Cumulative achievement in a mastery instructional system: the impact of
differences in resultant achievement motivation and persistence. American Educational Research Journal, 18
(1), 7 – 13.
Hemery, D. (1986). The pursuit of sporting excellence: a study of sport’s highest achievers. London: Willow Books.
Hoekman, K., McCormick, J., & Gross, M. U. M. (1999). The optimal context for gifted students: a preliminary
exploration of motivational and affective considerations. Gifted Child Quarterly, 43, 170 – 193.
Holland, J. (1986). New directions for interest testing. In: B. S. Plake, & J. C. Witt (Eds.), The future of testing
( pp. 245 – 267). Hillsdale, NJ: Lawrence Erlbaum Associates.
Holloway, S. D. (1987). Concepts of ability and effort in Japan and the United States. Review of Educational
Research, 58, 327 – 345.
Janos, P. M., & Robinson, N. M. (1985). Psychosocial development in intellectually gifted children. In: F. D.
Horowitz, & M. O’Brien (Eds.), The gifted and talented: developmental perspectives ( pp. 149 – 195). New
York: American Psychological Association.
Joswig, H. (1994). The connection between motivational and cognitive components of the personality of gifted
pupils. European Journal for High Ability, 5, 153 – 162.
Kanfer, R., & Ackerman, P. L. (1989). Motivation and cognitive abilities: an interactive/aptitude – treatment
interaction approach to skill acquisition. Journal of Applied Psychology, 74, 657 – 690.
Lloyd, J., & Barenblatt, L. (1984). Intrinsic intellectuality: its relation to social class, intelligence, and achieve-
ment. Journal of Personality and Social Psychology, 46, 646 – 654.
F. Gagné, F. St Père / Intelligence 30 (2001) 71–100 99

Lubinski, D., & Humphreys, L. G. (1997). Incorporating general intelligence into epidemiology and the social
sciences. Intelligence, 24, 159 – 201.
McHenry, J. J., Hough, L. M., Toquam, J. L., Hanson, M. A., & Ashworth, S. (1990). Project a validity results: the
relationship between predictor and criterion domains. Personnel Psychology, 43, 335 – 354.
Millman, J., Bieger, G. R., Klag, P. A., & Pine, C. K. (1983). Relation between perseverance and rate of learning:
a test of Carroll’s model of school learning. American Educational Research Journal, 20, 425 – 434.
Nicholls, J. G. (1992). Students as educational theorists. In: D. Schunk, & J. Meece (Eds.), Student perceptions in
the classroom ( pp. 267 – 285). Hillsdale, NJ: Lawrence Erlbaum Associates.
Norusis, M. J. (1983). Introductory statistics guide. New York: McGraw-Hill.
Ones, D. S., Viswesvaran, C., & Schmidt, F. L. (1993). Comprehensive meta-analysis of integrity test validities:
findings and implications for personnel selection and theories of job performance. Journal of Applied Psy-
chology Monograph, 78, 679 – 703.
Otis, A. S., & Lennon, R. T. (1969). Otis – Lennon mental ability test: technical handbook. New York: Harcourt,
Brace and World.
Raven, J. C., & Summers, B. (1986). Manual for Raven’s progressive matrices and vocabulary scales — research
supplement no. 3. London: Lewis.
Reis, S. M., Westberg, K. L., Kulilowich, J., Caillard, F., Hébert, T., Plucker, J., Purcell, J. H., Rogers, J. B., Smist,
J. M. Why not let high ability students start school in January?: the curriculum compacting study. Storrs, CT:
National Research Center of the Gifted and Talented.
Renzulli, J. S. (1986). The three-ring conception of giftedness: a developmental model for creative productivity.
In: R. J. Sternberg, & J. E. Davidson (Eds.), Conceptions of giftedness ( pp. 53 – 92). New York: Cambridge
Univ. Press.
Rushton, J. P., Brainerd, C. J., & Pressley, M. (1983). Behavioral development and construct validity: the principle
of aggregation. Psychological Bulletin, 94 (1), 18 – 38.
Schmidt, F. L., & Hunter, J. E. (1998). The validity and utility of selection methods in personnel psychol-
ogy: practical and theoretical implications of 85 years of research findings. Psychological Bulletin, 124,
262 – 274.
Shore, B. M., Cornell, D. G., Robinson, A., & Ward, V. S. (1991). Recommended practices in gifted education:
a critical analysis. New York: Teachers College Press.
Simonton, D. K. (1994). Greatness: who makes history and why. New York: Guilford Press.
Snow, R. E. (1992). Aptitude theory: yesterday, today, and tomorrow. Educational Psychologist, 27, 5 – 32.
Snow, R. E., Kyllonen, P. C., & Marshalek, B. (1984). The topography of ability and learning correlations. In: R.
J. Sternberg (Ed.), Advances in the psychology of human intelligence, vol. 2 ( pp. 47 – 104). Hillsdale, NJ:
Lawrence Erlbaum Associates.
Snyderman, M., & Rothman, S. (1987). Survey of expert opinion on intelligence and aptitude testing. American
Psychologist, 42, 137 – 144.
Spence, J. T., Pred, R. S., & Helmreich, R. L. (1989). Achievement strivings, scholastic aptitude, and academic
performance: a follow-up to ‘‘impatience versus achievement strivings in the type a pattern’’. Journal of
Applied Psychology, 74, 176 – 178.
St Père, F. (1997). La relation entre les aptitudes intellectuelles, la motivation et la persistance, ainsi que leur rôle
prédictif vis-à-vis du rendement scolaire [The relationship between intellectual aptitudes, motivation, and
persistence, as well as their predictive role with regard to school achievement]. Unpublished PhD thesis,
Université de Québec à Montréal, Montreal, Canada.
Tabachnick, B. G., & Fidell, L. S. (1989). Using multivariate statistics. New York: Harper Collins.
Tannenbaum, A. J. (1983). Gifted children: psychological and educational perspectives. New York: Macmillan.
Terborg, J. R. (1977). Validation and extension of an individual differences model of work performance. Organ-
izational Behavior and Human Performance, 18, 188 – 216.
Terman, L. M., & Oden, M. H. (1959). Genetic studies of genius, the gifted group at mid-life. V: Stanford, CA:
Stanford Univ. Press.
Vallerand, R. J., & Bissonnette, R. (1992). Intrinsic, extrinsic, and amotivational styles as predictors of behavior:
a prospective study. Journal of Personality, 60, 599 – 620.
Vallerand, R. J., Blais, M. R., Brière, N. M., & Pelletier, L. G. (1989). Construction et validation de l’échelle de
100 F. Gagné, F. St Père / Intelligence 30 (2001) 71–100

motivation en éducation (Construction and validation of the school motivation scale). Canadian Journal of
Behavioral Science, 21, 323 – 349.
Vallerand, R. J., Gagné, F., Senécal, C., & Pelletier, L. G. (1994). A comparison of the school intrinsic motivation
and perceived competence of gifted and regular students. Gifted Child Quarterly, 38, 172 – 175.
Vroom, V. H. (1964). Work and motivation. New York: Wiley.
Vygotsky, L. S. (1978). Mind in society: the development of higher psychological processes. Cambridge, MA:
Harvard Univ. Press.
Walberg, H. J. (1984). Improving the productivity of America’s schools. Educational Leadership, 41, 19 – 27.
Wong, M. M., & Csikszentmihalyi, M. (1991). Motivation and academic achievement: the effect of personality
traits and the quality of experience. Journal of Personality, 59, 539 – 574.

You might also like