A tripartite taxonomy of character

Full text

1
A Tripartite Taxonomy of Character
Angela Lee Duckworth, Eli Tsukayama, Sarah D. Patrick
University of Pennsylvania
Draft 4/03/14 – Please do not cite without permission
Author Note
Angela Lee Duckworth, Eli Tsukayama, and Sarah D. Patrick, Department of Psychology,
University of Pennsylvania.
The research reported here was supported by the Character Lab, the Spencer Foundation, and
K01-AG033182 from the National Institute on Aging.
Correspondence concerning this paper can be addressed to Angela Lee Duckworth, Department
of Psychology, University of Pennsylvania, 3701 Market Street, Suite 215, Philadelphia, PA 19104. E-
mail: duckworth@psych.upenn.edu.

2
Abstract
In the current investigation, we developed a character growth card for both applied (e.g., formative
assessment) and research purposes. In Study 1, in a sample of several hundred students from two
middle schools in the Northeast, both teachers (about five per student) and the students themselves
rated character skill items developed to represent zest, self-control (schoolwork and interpersonal),
gratitude, curiosity, optimism, grit, and social intelligence. Factor analyses indicated a three-factor
solution describing social, intellectual, and achievement character. Items loading on social character
included items from the gratitude, optimism, social intelligence, and interpersonal self-control scales.
Items loading on intellectual character included items from the zest and curiosity scales. Items loading
on achievement character included items from the grit, optimism, curiosity, and schoolwork self-
control scales. These three factors differentially predicted theoretically-related outcomes assessed one
year later, including objective academic performance and self-reported well-being and social
functioning. In Study 2, we replicated the three-factor structure and examined the relationships among
the three character factors and the Big Five personality traits.

3
A Tripartite Taxonomy of Character
Character, the disposition to act, think, and feel in admirable and beneficial ways, is widely
recognized as multidimensional (Peterson & Seligman, 2004). Under a variety of designations
(character strengths, virtues, life skills, soft skills, social and emotional skills, learning mindsets,
developmental assets, non-cognitive factors, etc.), character skills in young people are now attracting
new attention from scientists and practitioners alike. By whatever name, character skills long have
been considered essential for social, personal, and vocational success. Recently, growing numbers of
educators have become convinced that character skills are strongly implicated in academic success
(Tough, 2012). In parallel, there has been a surge of interest in the fields of developmental science,
social, positive, and personality psychology, and education in assessing and promoting the growth of
character skills (see, for example, Damon & Lerner, 2006; Damon, Lerner, Kuhn, Siegler, &
Eisenberg, 2011). Collectively, this empirical work suggests that character is largely independent of
general intelligence yet as important to academic, social, health, and economic outcomes (Heckman,
Humphries, & Kautz, 2014). In the current investigation, we developed a character growth card for
both applied (e.g., formative assessment) and research purposes.
Study 1
Method
Participants. Participants were 357 fifth through eighth grade students at two urban charter
middle schools in the Northeast region of the United States. About 51% were female; 65% were
Hispanic, were 34% Black, and 1% were from other ethnic backgrounds. About 81% were from low-
income families, as indicated by participation in the federal free lunch program.
Procedure and measures. Teacher ratings of student character were obtained each semester
during the 2011-2012 school year. The next school year (2012-2013), students completed online
questionnaires administered by school personnel during school hours assessing various aspects of their

4
psychological well-being, physical well-being, and social functioning. In spring 2013, students also
completed a self-report version of the character growth card. At the conclusion of the school year,
academic outcome data were collected from school records. Measures included class conduct,
attendance, and GPA. See Table 1 for a summary of when the measures were assessed.
Character growth card. Teachers (about five per student) and the students themselves rated
items chosen to represent zest, self-control (schoolwork and interpersonal), gratitude, curiosity,
optimism, grit, and social intelligence (see Tables 2 and 3). The 5-point scale ranged from 1 = Almost
never to 5 = Almost always. Because teacher ratings were consistent (avg. α = .78), we averaged
teacher ratings at the item level for subsequent analyses.
Life satisfaction. Students completed a single item assessing life satisfaction, “Overall, how
satisfied or unsatisfied are you with your life?” using a 7-point scale where 1 = Extremely unsatisfied
and 7 = Extremely satisfied. Lucas and Donnellan (2012) demonstrated that single-item measures of
life satisfaction tend to be reliable (i.e., reliability estimates > .70).
Positive and negative affect. Students completed a modified version of the Scale of Positive
and Negative Experience (SPANE; Diener et al., 2010). Students were asked to rate “How often do
you experience these feelings?” on a 5-point scale, where 1 = Never and 5 = Always. For positive
affect, the items were happy, relaxed, and excited. For negative affect, the items were sad, worried, and
angry. The internal reliability coefficients were α = .78 and α = .64 for positive and negative affect,
respectively.
Class participation. Students completed three items assessing class participation: “When was
the last time you raised your hand in class?”; “When was the last time you knew the answer, but
DIDN'T raise your hand in class?” (reverse scored); and “When was the last time you volunteered to
write something on the board during class?” The response options were “Today or yesterday,” “Within

5
the last week,” “Within the last month,” and “More than a month ago.” The internal reliability
coefficient was .28.
Peer conflict. Students completed two items assessing peer conflict: “When was the last time
you argued with a friend?”; and “When was the last time you were mean to someone?” The response
options were “Today or yesterday,” “Within the last week,” “Within the last month,” and “More than a
month ago.” The internal reliability coefficient was .56.
GPA. We collected final course grades from school records. We calculated GPA for each
quarter and for final GPA by averaging grades from all major academic courses, including math,
science, language arts, and social studies classes.
Attendance. We collected data on unexcused absences from attendance records.
Conduct grades. As part of regular school practice, teachers rated student homework and
conduct in each class using a single 5-point scale, where 1 = Unsatisfactory, 2 = Needs improvement, 3
= Satisfactory, 4 = Good, and 5 = Excellent. We calculated conduct for each marking period and for
final conduct by averaging grades from all major academic courses, including math, science, language
arts, and social studies classes.
Socioeconomic status and demographic variables. We obtained data on gender, ethnicity,
birthdate, and free lunch status from school records.
Results
Analytic strategy. We conducted exploratory factor analyses on the teacher-report character
growth card items from the 2011-2012 school year, and confirmatory factor analyses on the student-
report character growth card items from spring 2013. Factor analyses indicated a three-factor structure.
Subsequently, we ran correlations to examine the relationship between the character skills factors and
the outcomes. In order to examine the unique one-year predictive relationship between each outcome
and the teacher-report character skills factors above and beyond the other character skills factors, we

6
conducted simultaneous multiple regression models for each outcome with social, intellectual, and
achievement character factors as well as gender, ethnicity, and SES.
Exploratory factor analyses. Exploratory factor analyses on the teacher-report data suggested a
three-factor solution describing social, intellectual, and achievement character. The pattern of results
was the same for the fall, spring, and the averaged fall and spring data. Therefore, we report the results
from the average scores. We set the minimum factor loading criterion to .40. Because we expected
domains of character to share common variance and therefore to be correlated, we used oblique
promax rotation (k = 4). To determine the number of factors to extract, we used scree tests (Cattel,
1966) and the Kaiser criterion (Kaiser, 1960). These tests suggested extracting two to three factors. We
examined both solutions and selected the three-factor solution because it was psychologically
meaningful. Furthermore, in confirmatory factor analyses not reported here, the three-factor model fit
the data significantly better than the two-factor or one-factor models (ps < .001).
The subscales for social, intellectual, and achievement character demonstrated good internal
consistency, α = .98, .95, and .98, respectively. The correlations among the subscales were r = .51 for
social and intellectual character, r = .87 for social and achievement character, and r = .65 intellectual
and achievement character, suggesting that these domains are strongly correlated.
As shown in Table 2, items loading on social character seem to facilitate harmonious
relationships with other people, taking items from the gratitude, optimism, social intelligence, and
interpersonal self-control scales. Items loading on intellectual character seem to facilitate learning,
consisting of items from the zest and curiosity scales. Items loading on achievement character seem to
facilitate the achievement of personal goals, largely consisting of items from the grit, optimism,
curiosity, and schoolwork self-control scales.
Confirmatory factor analyses. Confirmatory Factor Analyses (CFAs) on the student-report
character growth card items confirmed that the three-factor model fit the data better than a one-factor

7
model, Δχ²(3) = 181.55, p < .001. In the three-factor model, items were allowed to load freely on their
respective factor, the factor loadings with other factors were set to zero, and the covariances between
the factors were freely estimated. In the one-factor models, all items were allowed to load freely on a
single factor. Factors were scaled by setting the variance equal to 1.0. All factor loadings were
significant at p < .001 (see Table 3). The hypothesized three-factor model fit the data adequately.
Following recommendations suggested by Kline (2004), we considered CFI values greater than .90,
RMSEA values less than .08, and SRMR values less than .10 to indicate good fit. Model χ²(206) =
181.55, p < .001, CFI was .86, the RMSEA was .073 (90% confidence interval was .066 to .080), and
the SRMR was .063. Allowing one item to dual load (“I recognize and show appreciation for my
opportunities” on the intellectual factor in addition to the social factor), and allowing three error
covariances (“I keep my temper in check” with “I remain calm even when criticized or otherwise
provoked,” “I actively participate” with “I am eager to explore new things,” and “I pay attention and
resist distractions” with “I believe that effort will improve my future”) brought the CFI above .90.
Finally, we tested for measurement invariance across gender by estimating a multiple-group
CFA model and constraining the factor loadings to be equal across groups (i.e., males vs. females).
Using ΔCFI ≤ .01 as a guideline (see Cheung & Rensvold, 2002), we found that the same factor
structure held across gender (ΔCFI = .005).
Teacher test-retest correlations. At the item level, the average test-retest correlation for the
teacher-report character growth card from fall 2011 to spring 2012 was r = .83. The subscale test-retest
correlations were .89, .85, and .91 for social, intellectual, and achievement character, respectively.
Student-teacher correlations. The correlations between teacher-report and self-report ratings
were .40, .45, and .34 for social, intellectual, and achievement character, respectively (see Table 4).
These associations compare favorably to the meta-analytically derived average correlation of r = .22
between child self-report and informant ratings by Achenbach, McConaughy, and Howell (1987).

8
Longitudinal teacher-report character skills prediction of the outcomes. As shown in Table
5, social character predicted less class participation (β = -.28, p = .045), lower peer conflict (β = -.60, p
< .001), and lower GPA (β = -.33, p < .001) one year later when controlling for the other character
factors and demographics. Intellectual character predicted more class participation (β = .36, p < .001),
and better attendance (β = .17, p = .046) one year later when controlling for the other character factors
and demographics. Achievement character predicted more peer conflict (β = .41, p = .016), better GPA
(β = 1.05, p < .001), and better conduct (β = .21, p < .01) one year later when controlling for the other
character factors and demographics.
Social character was the best predictor of lower peer conflict (β = -.60), and achievement
character was the best predictor of GPA (β = 1.05; ps < .05 for the differences in betas). For the rest of
the outcomes, two or more character skills did not differ significantly in predicting the outcome.
Study 2
In Study 1, we found evidence that character strengths clustered into three broad categories:
social, intellectual, and achievement. In Study 2, we sought to replicate and extend these findings by
situating the strengths in a taxonomy of personality: the Big Five Personality Traits.
Method
Participants. Participants were 1,549 fifth through eighth grade students at nine charter and
two private middle schools in the United States. About 52% were female; 13% were White, 33% were
Black, 45% were Hispanic, 5% were Asian, and 4% were from other ethnic backgrounds. About 80%
were from low-income families, as indicated by participation in the federal free lunch program.
Procedure and measures. Teacher and self-report ratings of student character were obtained
during the first semester of the 2013-2014 school year. Students also completed questionnaires
administered by school personnel during school hours assessing personality.

9
Character growth card. Teachers (about five per student) and the students themselves rated a
modified version of the character growth card used in Study 1 (see Table 6). We changed the response
scale to a 7-point scale ranging from 1 = Almost never to 7 = Almost always, made slight vocabulary
changes to several of the items to make them more appropriate for middle-school children, and added a
gratitude item: "I did something nice for someone else as a way of saying thank you." For the teacher-
rating scale, teachers rated each of the eight character strengths once with the items as bullet point
descriptions of the strengths. The 7-point scale ranged from 1 = Almost never to 5 = Almost always.
Student-report subscale alphas ranged from .76 to .85 (avg. = .81).
Big Five personality. Students completed selected items from the Big Five Inventory (John &
Srivastava, 1999). Specifically, we selected four items for each trait and added words in parentheses to
clarify potentially difficult vocabulary (list of items available upon request). Students rated the items
on a 7-point scale ranging from 1 = Strongly disagree to 7 = Strongly agree. Subscale alphas ranged
from .65 to .84 (avg. = .77).
Results
Analytic strategy. Because teacher and self-report ratings of student character were correlated
(avg. r = .24), we created composites scores by averaging teacher and self-report student character
subscales. We then conducted confirmatory factor analyses on these composite scores. Subsequently,
we examined the relationship between the character skills factors and the Big Five Personality Traits.
We examined correlations as well as simultaneous multiple regression models predicting each of the
Big Five with social, intellectual, and achievement character factors as well as gender, ethnicity, and
SES.
Confirmatory factor analyses. Confirmatory Factor Analyses (CFAs) confirmed that a three-
factor model fit the data better than a one-factor model, Δχ²(3) = 218.47, p < .001. In the three-factor
model, items were allowed to load freely on their respective factor, the factor loadings with other

10
factors were set to zero, and the covariances between the factors were freely estimated. In the one-
factor models, all items were allowed to load freely on a single factor. Factors were scaled by setting
the variance equal to 1.0. All factor loadings were significant at p < .001 (see Table 7). The
hypothesized three-factor model fit the data adequately. Model χ²(17) = 511.66, p < .001, CFI was .93,
the RMSEA was .147 (90% confidence interval was .137 to .157). Although the RMSEA was greater
than expected, this indication of poor fit may have resulted from small model size (Kenny & McCoach,
2003) and large factor loadings (Browne, MacCallum, Andersen, & Glaser, 2002; Miles & Shevlin,
2007), rather than actual model misspecification.
Finally, we tested for measurement invariance across gender by estimating a multiple-group
CFA model and constraining the factor loadings to be equal across groups (i.e., males vs. females).
Using ΔCFI ≤ .01 as a guideline (see Cheung & Rensvold, 2002), we found that the same factor
structure held across gender (ΔCFI < .001).
Student-teacher correlations. The correlations between teacher-report and self-report ratings
were .35, .20, and .26 for social, intellectual, and achievement character, respectively (see Table 8).
These associations compare favorably to the meta-analytically derived average correlation of r = .22
between child self-report and informant ratings by Achenbach, McConaughy, and Howell (1987).
Character skills factors relation to the Big Five personality traits. As shown in Table 9, all
three of the character factors were associated with all of the Big Five personality traits (ps < .001).
Among the character factors, intellectual character (r = .52) was the best predictor of openness to
experience, achievement character (r = .68) was the best predictor of conscientiousness, intellectual
character (r = .47) was the best predictor of extraversion, social character (r = .62) was the best
predictor of agreeableness, and intellectual character (r = .17) was the best predictor of emotional
stability (ps < .001 for the difference in rs).

11
As shown in Table 10, social character was associated with more agreeableness (β = .50, p <
.001) when controlling for the other character factors and demographics. Intellectual character
predicted more openness to experience (β = .52, p < .001), conscientiousness (β = .23, p < .001),
extraversion (β = .62, p < .001), and agreeableness (β = .10, p = .031), and less emotional stability (β =
-.13, p < .019) when controlling for the other character factors and demographics. Achievement
character predicted more conscientiousness (β = .55, p < .001) and agreeableness (β = .13, p = .018)
when controlling for the other character factors and demographics.
Differences in betas generally paralleled the differences in zero-order correlations. Among the
character factors, intellectual character (β = .52) was the best predictor of openness to experience,
achievement character (β = .55) was the best predictor of conscientiousness, intellectual character (β =
.62) was the best predictor of extraversion, and social character (β = .50) was the best predictor of
agreeableness (ps < .05 for the difference in βs). However, character did not differentially predict
emotional stability (ps > .22 for the difference in βs).
We also examined personality predicting character controlling for other personality traits and
demographics (see Table 11). Among the personality traits, agreeableness (β = .35, p < .001) and
conscientiousness (β = .30, p < .001) were the best predictors of social character; conscientiousness (β
= .35, p < .001) was the best predictor of intellectual character; and conscientiousness (β = .56, p <
.001) was the best predictor of achievement character (ps < .01 for the difference in βs).
General Discussion
In two studies, factor analyses supported a three-factor structure for character distinguishing
between social, intellectual, and achievement character. These character factors differentially related to
a variety of outcomes, including objective academic performance, self-reported well-being, social
functioning, and personality. For instance, social character had the strongest relationship with peer

12
conflict, intellectual character had the strongest relationship with class participation, and achievement
character had the strongest relationship with GPA.
Limitations and Future Directions
As with all empirical studies, the current investigation has several limitations that suggest
future directions. First, we were surprised to observe a negative relationship between social character
and class participation, a positive relationship between achievement character and peer conflict, and a
negative relationship between social character and GPA in multiple regression models controlling for
the other character skills in Study 1. These bivariate relationships (see Table 4) were in the
theoretically-predicted direction indicating statistical suppression in the regression models. Although
the variance inflation factors (VIFs) in the models were all less than 7, Cohen, Cohen, West, and Aiken
(2003) note that “the values of the multicollinearity indices at which the interpretation of regression
coefficients may become problematic will often be considerably smaller than traditional rules of thumb
guidelines such as VIF = 10” (p. 425). Because multicollinearity is due to insufficient information in
the data (Berry & Feldman, 1985), we are collecting data on more participants in order to obtain more
precise estimates in future studies. Furthermore, we are simultaneously assessing teacher-report and
student-report character in order to create a composite character report card measure. Because the
correlations between the student-report character skill factors were lower than those between the
teacher-report character skill factors, a composite measure should exhibit less multicollinearity.
Second, as with all correlational studies, we cannot infer causality. Specifically, we do not
know if these character skills cause the outcomes. Random-assignment interventions to increase
character skills are needed to establish causal relationships with outcomes. To assess interventions,
however, valid measures to assess character growth are needed. The current investigation is one step
toward generating such measures to assess character growth.

13
Finally, the low reliability of some of the outcomes presents another limitation of the current
study. In the future, we plan to conduct new analyses using structural equation modeling with latent
variables to assess relationships between the character skills factors and outcomes. Future studies
should use more reliable outcomes.
Conclusion
These results converge with other classifications of character in current theoretical research.
For instance, the same tripartite taxonomy has been identified by the National Research Council of the
National Academies (2012) in their review of skills and knowledge necessary to function in the 21st
century economy as well as by a taskforce on novel measures of competence in school-age children
convened by the Spencer Foundation. This study provides further empirical evidence that such a
tripartite model exists, and additionally, that these character skills can contribute to personal
achievement. Moreover, the current study adds to existing literature by focusing on an understudied
population that may stand to benefit most from the promotion and development of character skills
related to positive academic and life outcomes.

14
References
Achenbach, T. M., McConaughy, S. H., & Howell, C. T. (1987). Child/adolescent behavioral and
emotional problems: Implications of cross-informant correlations for situational specificity.
Psychological Bulletin, 101, 213-232.
Berry, W. D., & Feldman, S. (1985). Multiple Regression in Practice. Thousand Oaks: Sage.
Cattel, R. B. (1966). The scree test for the number of factors. Multivariate Behavioral Research, 1,
245-276.
Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing measurement
invariance. Structural Equation Modeling, 9, 233-255.
Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied Multiple Regression/Correlation
Analysis for the Behavioral Sciences (3rd ed.). New York: Routledge.
Damon, W., & Lerner, R. (2006). Handbook of Child Psychology: The Sixth Edition (Vol. 1-4). New
York: John Wiley and Sons.
Damon, W., Lerner, R., Kuhn, D., Siegler, R., & Eisenberg, N. (2011). Child and Adolescent
Development: An Advanced Text. New York: John Wiley and Sons.
Diener, E., Wirtz, D., Tov, W., Kim-Prieto, C., Choi, D.-w., Oishi, S., & Biswas-Diener, R. (2010).
New well-being measures: Short scales to assess flourishing and positive and negative feelings.
Social Indicators Research, 97, 143-156.
Heckman, J. J., Humphries, J. E., & Kautz, T. (Eds.). (2014). The Myth of Achievement Tests: The
GED and the Role of Character in American Life. Chicago: University of Chicago Press.
John, O. P., & Srivastava, S. (1999). The Big Five Trait taxonomy: History, measurement, and
theoretical perspectives. In L. A. Pervin & O. P. John (Eds.), Handbook of personality: Theory
and research (2nd ed.). (pp. 102-138). New York, NY, US: Guilford Press.
Kaiser, H. F. (1960). The application of electronic computers to factor analysis. Psychometrika, 20,
141-151.
Kline, R. B. (2004). Principles and Practice of Structural Equation Modeling (2nd ed.). New York:
Guilford.
Lucas, R., & Brent Donnellan, M. (2012). Estimating the reliability of single-item life satisfaction
measures: Results from four national panel studies. Social Indicators Research, 105, 323-331.
doi: 10.1007/s11205-011-9783-z
NRC. (2012). Education for life and work: Developing transferable knowledge and skills in the 21st
century. Washington, D.C.: The National Academies Press.
Peterson, C., & Seligman, M. E. P. (2004). Character strengths and virtues: A handbook and
classification. Washington, DC: American Psychological Association.
Tough, P. (2012). How children succeed: Grit, curiosity, and the hidden power of character. New
York: Houghton Mifflin Harcourt.

15
Table 1
Timeline of Assessments in Study 1
2011-2012 School Year
2012-2013 School Year
Fall
Spring
Fall
Spring
Teacher-Report Character Skills


Student-Report Character Skills

Student-Report Outcomes


Academic Outcomes



16
Table 2
Exploratory Factor Analysis Loadings for 2011-2012 Teacher-Report Character Growth Card Items in
Study 1
Factor
Social
Intellectual
Achievement
I keep my temper in check
1.06
I remain calm even when criticized or otherwise provoked
1.03
I get over frustrations and setbacks quickly
1.00
I am polite to adults and peers
0.90
I demonstrate respect for the feelings of others
0.89
I allow others to speak without interruption
0.79
I am able to find solutions during conflicts with others
0.76
I recognize and show appreciation for other
0.57
I recognize and show appreciation for my opportunities
0.48
I show enthusiasm
1.10
I invigorate others
0.95
I actively participate
0.86
I ask and answer questions to deepen understanding
0.84
I am eager to explore new things
0.78
I finish whatever I begin
0.89
I come to class prepared
0.87
I work independently with focus
0.87
I get to work right away rather than procrastinating
0.78
I remembers and follow directions
0.75
I believe that effort will improve my future
0.59
I try very hard even after experiencing failure
0.54
I actively listen to others
0.41
Note. Factor loadings are from oblique promax solutions (promax k = 4). Factor loadings less than .40
are not displayed.

17
Table 3
Three-Factor Confirmatory Factor Analysis Loadings for Spring 2013 Student-Report Character Growth
Card Items in Study 1
Factor
Social
Intellectual
Achievement
I keep my temper in check
0.66
I remain calm even when criticized or otherwise provoked
0.55
I get over frustrations and setbacks quickly
0.44
I am polite to adults and peers
0.67
I demonstrate respect for the feelings of others
0.71
I allow others to speak without interruption
0.64
I am able to find solutions during conflicts with others
0.60
I recognize and show appreciation for my opportunities
0.62
I recognize and show appreciation for others
0.70
I show enthusiasm
0.71
I invigorate others
0.42
I actively participate
0.61
I ask and answer questions to deepen understanding
0.55
I am eager to explore new things
0.61
I finish whatever I begin
0.61
I come to class prepared
0.56
I work independently with focus
0.67
I get to work right away rather than procrastinating
0.53
I remembers and follow directions
0.68
I believe that effort will improve my future
0.60
I try very hard even after experiencing failure
0.71
I actively listen to others
0.67
Notes. Factor loadings are from oblique two-factor models. Factor loadings are significant at p < .001.

18
Table 4
Summary Statistics and Bivariate Correlations between the Character Skills Factors and the Outcomes in Study 1
Measures
M
SD
1
2
3
4
5
6
Character Strengths
1. Teacher-Report Social
3.82
0.55
-
2. Teacher-Report Intellectual
3.60
0.51
.51***
-
3. Teacher-Report Achievement
3.75
0.60
.87***
.65***
-
4. Student-Report Social
3.58
0.70
.40***
.25***
.35***
-
5. Student-Report Intellectual
3.60
0.75
-.04
.45***
.10
.37***
-
6. Student-Report Achievement
3.67
0.69
.31***
.33***
.34***
.72***
.54***
-
Student-Report Outcomes
8. Life Satisfaction
4.85
1.59
.09
.12†
.11
.34***
.27***
.36***
9. Positive Affect
3.50
0.82
-.01
.05
-.04
.36***
.36***
.32***
10. Negative Affect
2.69
0.73
.04
.08
.06
-.14*
-.02
-.08
13. In-Class Participation
2.74
0.61
-.06
.25***
.06
.10
.40***
.25***
14. Peer Conflict
2.49
0.92
-.22**
.01
-.08
-.34***
-.04
-.21***
Academic Outcomes
16. GPA
77.67
8.20
.55***
.51***
.73***
.28***
.16**
.32***
17. Attendance
38.58
5.32
.38***
.32***
.40***
.24***
.08
.31***
18. Conduct
59.51
19.15
.19***
.22***
.21***
.06
.06
.05
Demographics
19. Female
51%
.24***
-.03
.26***
-.03
-.01
.06
20. Hispanic
34%
.17**
.10†
.17**
.05
-.08
.06
21. Black
65%
-.17**
-.10†
-.17**
-.13†
.11
-.04
22. Other Ethnicity
1%
-.04
.01
-.04
-.04
-.05
-.02
23. Free Lunch
81%
.05
-.05
.07
.10
.04
.06
Note. † < .10. *p < .05. **p < .01. ***p < .001.

19
Table 5
Standardized Coefficients from Longitudinal Multiple Regression Models Predicting Outcomes from Teacher-Report Social, Intellectual,
and Achievement Character in Study 1
Predictor
Life
Satisfaction
Positive
Affect
Negative
Affect
Class
Participation
Peer
Conflict
GPA
Attendance
Conduct
Social Character
.04a
.22a
-.09a
-.28*b
-.60***a
-.33***c
.19a
.01a
Intellectual Character
.06a
.11a
.13a
.36***a
.02b
-.01b
.17*a
.10a
Achievement Character
.09a
-.26a
-.05a
.06ab
.41*b
1.05***a
.09a
.21**a
Note. The absolute values of coefficients with different superscripts within a column are significantly different from each other.
† < .10. *p < .05. **p < .01. ***p < .001.

20
Table 6
Revised Items in Study 2
Subscale / Item
Zest
I actively participated.
I showed enthusiasm.
I approached new situations with excitement and
energy.
Grit
I finished whatever I
began.
I worked independently with
focus.
I tried very hard even after experiencing
failure.
I stayed committed to my goals.
I kept working hard even when I felt like
quitting.
Self-Control (School Work)
I came to class
prepared.
I remembered and followed
directions.
I got to work right away rather than procrastinating (waiting until the
last minute).
I paid attention and resisted
distractions.
Self-Control (Interpersonal)
I remained calm even when criticized or otherwise
provoked.
I allowed others to speak without
interruption.
I was polite to adults and peers.
I kept my temper in

21
check.
Optimism
I believed that effort would improve my
future.
When bad things happened, I thought about what I could do to make it better
next time.
I stayed motivated, even when things didn’t
go well.
I believed that I could improve on things I’m not
good at.
Gratitude
I recognized what other people did for me.
I showed appreciation for the good things I have in
my life.
I expressed appreciation by saying thank
you.
I did something nice for someone else as a way of saying
thank you.
Social Intelligence
I was able to find solutions during conflicts with
others.
I demonstrated respect for the feelings of
others.
I adapted to different social
situations.
Curiosity
I was eager to explore new things.
I asked questions to deepen my
understanding.
I took an active interest in
learning.

22
Table 7
Three-Factor Confirmatory Factor Analysis Loadings for Character Growth Card Scales in Study 2
Factor
Social
Intellectual
Achievement
Interpersonal Self-Control
0.74
Social Intelligence
0.82
Gratitude
0.76
Zest
0.75
Curiosity
0.78
Optimism
0.84
Work Self-Control
0.80
Grit
0.88
Notes. Factor loadings are from oblique two-factor models. Factor loadings are significant at p < .001.

23
Table 8
Student
Teacher
Social
Intellectual
Achievement
Social Character
.35***Aa
.15***Cab
.26***Ba
Intellectual Character
.20**Ab
.20***Aa
.15**Ab
Achievement Character
.25***Ab
.13**Bb
.26***Aa
Note. Coefficients with different superscripts are significantly different from each other (upper case for rows and lower case for columns).
**p < .01. ***p < .001.

24
Table 9
Summary Statistics and Bivariate Correlations between the Character Skills Factors and Personality in Study 2
Measures
M
SD
1
2
3
4
5
6
7
8
Character
1. Composite Social
5.68
0.9
-
2. Composite Intellectual
5.57
0.91
.69***
-
3. Composite
Achievement
5.6
0.92
.79***
.73***
-
Personality
4. Openness
6.07
0.91
.35***
.52***
.38***
-
5. Conscientiousness
5.85
0.93
.55***
.62***
.68***
.46***
-
6. Extraversion
5.88
0.95
.24***
.47***
.23***
.51***
.39***
-
7. Agreeableness
5.92
0.95
.62***
.51***
.54***
.45***
.59***
.37***
-
8. Emotional Stability
4.04
1.57
.10***
.17***
.12***
-.01
.12***
.05*
.05*
-
Demographics
9. Female
52%
.09**
-.01
.08**
.09***
.07*
.06*
.09**
-.21***
10. White
13%
.22***
.18***
.18***
.05†
.10**
.06*
.08**
.06*
11. Black
33%
-.35***
-.16***
-.30***
.02
-.06*
.10**
-.14***
.04
12. Hispanic
45%
.15***
.01
.12***
-.03
-.04
-.08**
.05
-.08**
13. Asian
5%
.01
-.02
.02
-.07*
.02
-.09**
.01
-.04
14. Other Ethnicity
4%
.09**
.07*
.10**
.03
.05†
-.04
.06*
.06†
15. Free Lunch
80%
-.25***
-.23***
-.23***
-.14***
-.08*
-.07†
-.15***
-.11**
Note. † < .10. *p < .05. **p < .01. ***p < .001.

25
Table 10
Standardized Coefficients from Multiple Regression Models Predicting Personality from Social, Intellectual, and Achievement Character in
Study 2
Openness
Conscientiousness
Extraversion
Agreeableness
Emotional Stability
Social Character
-.01a
.06a
-.06a
.50***b
.04a
Intellectual Character
.52***b
.23***b
.62***b
.10*a
-.13*a
Achievement Character
.05a
.55***c
-.10†a
.13*a
-.10a
Note. The absolute values of coefficients with different superscripts within a column are significantly different from each other.
† < .10. *p < .05. **p < .01. ***p < .001.

26
Table 11
Standardized Coefficients from Multiple Regression Models Predicting Social, Intellectual, and Achievement Character from Personality in
Study 2
Character
Social
Intellectual
Achievement
Openness
.04bc
.22***b
.06†b
Conscientiousness
.30***a
.35***a
.56***a
Extraversion
-.04c
.19***bc
-.07*c
Agreeableness
.35***a
.06†d
.14***b
Emotional Stability
.05†b
.11***cd
.07*b
Note. Coefficients with different superscripts are significantly different from each other.
†p < .10. *p < .05. **p < .01. ***p < .001.