ArticlePDF Available

Abstract and Figures

Beck’s insight—that beliefs about one’s self, future, and environment shape behavior—transformed depression treatment. Yet environment beliefs remain relatively understudied. We introduce a set of environment beliefs— primal world beliefs or primals —that concern the world’s overall character (e.g., the world is interesting, the world is dangerous ). To create a measure, we systematically identified candidate primals (e.g., analyzing tweets, historical texts, etc.); conducted exploratory factor analysis ( N = 930) and two confirmatory factor analyses ( N = 524; N = 529); examined sequence effects ( N = 219) and concurrent validity ( N = 122); and conducted test-retests over 2 weeks ( n = 122), 9 months ( n = 134), and 19 months (n = 398). The resulting 99-item Primals Inventory (PI-99) measures 26 primals with three overarching beliefs— Safe, Enticing , and Alive (mean α = .93)—that typically explain ∼55% of the common variance. These beliefs were normally distributed; stable (2 weeks, 9 months, and 19 month test-retest results averaged .88, .75, and .77, respectively); strongly correlated with many personality and wellbeing variables (e.g., Safe and optimism, r = .61; Enticing and depression, r = −.52; Alive and meaning, r = .54); and explained more variance in life satisfaction, transcendent experience, trust, and gratitude than the BIG 5 (3%, 3%, 6%, and 12% more variance, respectively). In sum, the PI-99 showed strong psychometric characteristics, primals plausibly shape many personality and wellbeing variables, and a broad research effort examining these relationships is warranted.
Content may be subject to copyright.
Primal World Beliefs
Jeremy D. W. Clifton and Joshua D. Baker
University of Pennsylvania
Crystal L. Park
University of Connecticut
David B. Yaden
University of Pennsylvania
Alicia B. W. Clifton
Search for Common Ground, Washington, DC
Paolo Terni
California State University, Sacramento
Jessica L. Miller
Georgia State University
Guang Zeng
Tsinghua University
Salvatore Giorgi
University of Pennsylvania
H. Andrew Schwartz
Stony Brook University
Martin E. P. Seligman
University of Pennsylvania
Beck’s insight—that beliefs about one’s self, future, and environment shape behavior—transformed
depression treatment. Yet environment beliefs remain relatively understudied. We introduce a set of
environment beliefs—primal world beliefs or primals—that concern the world’s overall character (e.g.,
the world is interesting,the world is dangerous). To create a measure, we systematically identified
candidate primals (e.g., analyzing tweets, historical texts, etc.); conducted exploratory factor analysis
(N930) and two confirmatory factor analyses (N524; N529); examined sequence effects (N
219) and concurrent validity (N122); and conducted test-retests over 2 weeks (n122), 9 months
(n134), and 19 months (n398). The resulting 99-item Primals Inventory (PI-99) measures 26
primals with three overarching beliefs—Safe,Enticing, and Alive (mean ␣⫽.93)—that typically explain
55% of the common variance. These beliefs were normally distributed; stable (2 weeks, 9 months, and
19 month test-retest results averaged .88, .75, and .77, respectively); strongly correlated with many
personality and wellbeing variables (e.g., Safe and optimism, r.61; Enticing and depression, r⫽⫺.52;
Alive and meaning, r.54); and explained more variance in life satisfaction, transcendent experience,
trust, and gratitude than the BIG 5 (3%, 3%, 6%, and 12% more variance, respectively). In sum, the PI-99
showed strong psychometric characteristics, primals plausibly shape many personality and wellbeing
variables, and a broad research effort examining these relationships is warranted.
This article was published Online First October 8, 2018.
Jeremy D. W. Clifton and Joshua D. Baker, Department of Psychology,
University of Pennsylvania; Crystal L. Park, Department of Psychological
Sciences, University of Connecticut; David B. Yaden, Department of
Psychology, University of Pennsylvania; Alicia B. W. Clifton, Communi-
cations, Search for Common Ground, Washington, DC; Paolo Terni, Cen-
ter for College and Career Readiness, California State University, Sacra-
mento; Jessica L. Miller, Department of Psychology, Georgia State
University; Guang Zeng, Department of Psychology, Tsinghua University;
Salvatore Giorgi, Positive Psychology Center, University of Pennsylvania;
H. Andrew Schwartz, Department of Computer Science, Stony Brook
University; Martin E. P. Seligman, Department of Psychology, University
of Pennsylvania.
Research was partly supported by the Templeton Religion Trust. Over
sixty additional researchers volunteered numerous hours over four
years, including senior advisors: C. Dweck and J. Pawelski; other 2014
Primals Retreat attendees: A. Crum, R. DeRubeis, A. Fiske, R. Reeves,
P. Rozin, C. Sripada, and D. S. Wilson; literature reviewers: R. Oak-
erson, I. Verstegen, F. Wood, and A. Hao; interns: T. Kreiss, N.
Kulkarni, T. Jeong, S. Dominguez, C. Tsu, I. Garfunkel, S. Schimmel,
S. Kim, S. Felt, S. Witt, A. Brown, G. Leonardo, F. Chowdry, C.
Stollsteimer, J. Froberg, G. Parris, L. Petroucheva, M. Split, A. Zaarur,
I. Ugboaja, K. Carney, J. Spandorfer, O. Walsh, and B. Adams; Chinese
scholars: K. Peng, L. Cai, H. Ying, H. Li, S. Cai, Y. Zhou, Z. Ni, F. Xu,
and several others; factor analysis consultants: W. Ruch, S. B. Kauff-
man, P. McDermott, and K. Allred; additional specialists: E. Metz, C.
Campbell, C. Clements, A. Roepke, G. Bermant, and A. Potkay.
Correspondence concerning this article should be addressed to Jeremy
D. W. Clifton, Department of Psychology, University of Pennsylvania,
Philadelphia, PA 19104. E-mail:
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Psychological Assessment
© 2018 American Psychological Association 2019, Vol. 31, No. 1, 82–99
Public Significance Statement
Beliefs about a situation often impact thought and action. For example, in places seen as dangerous,
one is more alert. What if the whole world is seen as dangerous? This study identifies and creates a
tool to measure 26 fundamental beliefs about the world as a whole.
Keywords: beliefs, worldview, attribution, personality, primal world beliefs
Supplemental materials:
In the 1960s and 1970s, Beck based cognitive therapy on the
premise that beliefs about the self, the self’s future, and the self’s
external environment—what he called the Cognitive Triad—play a
fundamental role in the etiology of depression (e.g., 1979). These
beliefs concern basic judgments such as I’m worthless (a self-
belief) or my boss hates me (an environment belief). His work
inspired decades of improved clinical treatment for depression and
other disorders (Hofmann, Asnaani, Vonk, Sawyer, & Fang,
2012). Expanding this insight into the cognitive roots of behavior,
researchers found that shaping these beliefs increases aspects of
quality of life, such as wellbeing and skill development (Dweck,
Chiu, & Hong, 1995). However, while self-beliefs have received
considerable attention, environment beliefs remain relatively ne-
glected (Chen et al., 2016). This article introduces a class of
environment beliefs that may influence wellbeing and personality.
A Definition in Historical Context
One of the earliest debates in Western philosophy concerned the
world’s overall qualities. For example, Heraclitus argued the world
is defined by change, whereas Pythagoras suggested beauty (Kirk,
Raven, & Schofield, 2011). After Socrates’ turn inward to the
unexamined life more or less changed the topic for many centuries,
nineteenth-century thinkers came to see Weltanschauung (i.e.,
worldview)—an amalgam of religious, political, and moral doc-
trines—as determining behavior (Naugle, 2002). Koltko-Rivera
(2004) identified 41 such beliefs. Examples concern the mutability
of human nature and the purpose of sexual activity.
Within worldview, we focus on a subset of beliefs specific to the
overall character of the world rather than particular topics within
it—that is, only beliefs about everything. Instead of doctrine, our
focus, like Beck’s, is basic character such as beautiful or danger-
ous.Koltko-Rivera (2004) noted that such beliefs lack terminol-
ogy. We call them primal world beliefs,orprimals.
We conceive of primals as falling along continuous conceptual
dimensions anchored by opposites such as interesting and boring.
We propose six criteria to define primals more precisely (all but
the fourth apply to beliefs in Beck’s Triad). (a) Simple: Though
linguistic and conceptual simplicity differ, primals usually take the
form “the world is X” where X is one or two basic words con-
cerning a single-faceted concept like beautiful or dangerous. (b)
Adjectival: Rather than describing how or why the world came to
be, primals describe what the world is like using “is” (i.e., predi-
cate adjective) statements ascribing a current quality. (c) Goal-
relevant: Dweck (2017, p. 698) uses the term goal-relevant to refer
to perceived environmental conditions essential to the individual’s
interests, needs, or values. Goal-relevant qualities, like “safe,” may
ground the psychological relevance of more specific qualities, like
“toxic.” (d) Maximally General: Primals concern the world as a
whole, and thus what is typical of most things and situations.
Precise physical boundaries, such as this jungle or that solar
system, are misleading. Instead, primals concern an individual’s
broadest psychologically meaningful habitat. (e) Automatic:
Though sometimes explicit, primals are expected to operate pri-
marily beneath cognitive awareness in the “System 1” family of
cognitions (see Kahneman & Frederick, 2002). (f) Active: Like
other beliefs, primals are expected to dynamically direct attention;
organize, simplify, filter, and fill in information; and guide action.
In sum, primals are one’s implicit answers to the question “what
sort of world is this?” The next section suggests the value of
studying primals and a tentative model for establishing the validity
of a primals inventory.
Theory and Literature
A half-century of research clearly indicates that various beliefs
shape behavior and wellbeing— often dramatically (Hofmann et
al., 2012). Researchers have also identified mechanisms, such as
those associated with schemas and priors, which govern how
beliefs shape behavior via ambiguity interpretation (Janoff-
Bulman, 1989;McNamara, Green, & Olsson, 2006). In the context
of fruitful literatures examining beliefs, Dweck (2008,2017) re-
cently suggested that as-yet unidentified beliefs may even shape
major personality traits as well. She calls for the identification of
promising beliefs followed by initial correlational research exam-
ining critical empirical benchmarks that—if met—would justify a
broader phase examining causation, development, and other im-
portant issues. This measurement article introduces primals, ex-
amines these critical benchmarks, and leaves causation and other
questions for a later date.
In our tentative model, primals operate like other beliefs. For
example, some may see the world as a negative place. This view
would inform a base-rate of negativity that might alter ambiguity
interpretation toward seeing situations as miserable, meaningless,
and getting worse. In worlds where fortunate events are considered
exceptions, pessimism appears prudent and optimism naïve. Con-
versely, some may see the world positively: most situations are
enjoyable, safe, and naturally tend to work out—the unknown
hides further wonders. When misfortune is rare, pessimism ap-
pears profitless or paranoid, and optimism sensible. In sum, both
pessimists and optimists might theoretically be realists who hap-
pen to disagree about what is real.
In addition to pessimism and optimism, many other personality
and wellbeing-related variables may be shaped by primals. Rau-
thmann et al. (2014) and Dweck (2017) suggest that various
environment beliefs and reactions to those beliefs could— over
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
time—aggregate into major personality traits. For example, the
belief that a situation is dangerous should increase neurotic behav-
iors. If so, a belief that the world as a whole is dangerous should
increase neuroticism scores. Likewise, curiosity, which makes
little sense in contexts offering low return on attentional invest-
ment, may be part of a reaction to a belief that the world is actually
full of fascinating things. Similarly, resistance to cultural change
(i.e., social conservatism) may be part of a reaction to a belief that
the world is deteriorating. Using the intuitive logic they suggested
(given Xenvironment, Yis prudent/natural; Rauthmann et al.,
2014), we identified hundreds more hypotheses concerning how
primals might shape outcomes. To establish convergent validity,
we provide 30 such hypotheses in Table 2 along with correlational
results from studies below.
All such hypotheses rest on the (perhaps counterintuitive) as-
sumption that labels such as neuroticism,curiosity, and so forth
may express less who we are and more where we think we are. The
attribution literature commonly distinguishes between disposi-
tional and situational influences (Harvey, Madison, Martinko,
Crook, & Crook, 2014). Personality traits have historically been
attributed to disposition (i.e., a biological or innate quality shaped
by experience; e.g., Eysenck, 1963). Kelley’s covariation model
(1967) suggests why: dispositional theories appeal most when
interpreting behavioral patterns that (a) persist across situations
(i.e., low-distinctiveness); (b) remain consistent over time (i.e.,
high-consistency); and (c) vary across individuals (i.e., low-
consensus). Put more plainly, when individuals behave consis-
tently across situations, we assume the action expresses an innate
person characteristic. Yet if primals are stable across situations and
time— but vary across persons—this assumption is unjustified. In
theory, individuals could spend decades compensating for (or
capitalizing on) habitat-wide constraints described by one’s pri-
mals while sharing actual habitats with those reacting to entirely
different perceptions. Then, because (a) individuals may not real-
ize others hold different primals; (b) individuals are biased by the
Fundamental Attribution Error, which causes one to overestimate
the role of disposition; and (c) patterns of behavior associated with
personality traits exhibit Kelley’s (1967) hallmarks of disposition
(i.e., low-distinctiveness, etc.); individuals could en masse misin-
terpret the actions of others as an expression of who they are rather
than the reasonable reaction to where they happen to think they
are. In other words, the tentative model informing our measure-
ment assumptions—and animating our interest in this topic—
holds that many personality variables and wellbeing outcomes are
driven in part by the (perceived) external situation rather than
internal disposition. For this theory to be plausible (and for a broad
research phase examining it to be justified), Dweck (2008,2017)
suggests certain critical benchmarks should be met. Primals must
(a) be as stable within-persons as personality trait scores; (b) vary
Table 1
Standardized Cronbach’s for 26 PI Subscales in All Studies
S1 S2
S3 S4 S5 S6 2-week S6 9-month S6 19-month Total
N930 524 529 219 122 122 134 398 2,454
Good .97 .96 .97 .97 .97 .97 .97 .97 .97
Safe .96 .93 .96 .95 .96 .95 .96 .96 .96
Enticing .95 .94 .94 .94 .94 .94 .95 .94 .95
Alive .90 .87 .86 .87 .91 .90 .89 .90 .89
Secondary Mean .94 .91 .92 .92 .94 .93 .93 .93 .93
About Me .85 .67 .83 .78 .85 .86 .87 .86 .84
Abundant .85 .80 .83 .84 .82 .70 .87 .83 .84
Acceptable .83 .77 .77 .84 .75 .74 .85 .82 .81
Beautiful .84 .80 .82 .83 .79 .74 .80 .81 .82
Changing .79 .73 .75 .83 .79 .76 .83 .82 .79
Cooperative .84 .77 .84 .75 .81 .73 .81 .85 .84
Funny .90 .83 .87 .83 .81 .83 .86 .89 .88
Harmless .91 .82 .89 .89 .89 .88 .90 .90 .90
Hierarchical .77 .76 .77 .77 .69 .67 .83 .80 .78
Improvable .82 .79 .79 .80 .77 .79 .87 .83 .81
Intentional .87 .79 .85 .82 .84 .82 .88 .88 .86
Interconnected .87 .84 .81 .87 .82 .70 .88 .85 .85
Interesting .85 .85 .85 .85 .86 .82 .87 .87 .86
Just .82 .71 .74 .83 .85 .81 .87 .83 .81
Meaningful .87 .86 .84 .82 .86 .81 .90 .88 .86
Needs Me .91 .85 .92 .88 .90 .87 .90 .90 .91
Pleasurable .91 .84 .88 .85 .88 .85 .89 .87 .89
Progressing .92
.91 .91 .89 .89 .90 .92 .91
Regenerative .86 .83 .83 .84 .79 .80 .89 .85 .85
Stable .82 .80 .77 .79 .78 .70 .84 .83 .81
Understandable .81 .80 .80 .78 .82 .76 .80 .78 .80
Worth Exploring .81 .80 .69 .77 .63 .64 .78 .81 .77
Tertiary Mean .85 .80 .82 .83 .81 .78 .86 .85 .84
Study Mean .87 .82 .84 .84 .83 .80 .87 .86 .86
Because Study 2 used an initial PI version (e.g., Progressing was measured with only one item), data from its 524 participants were not included in totals.
Values falling below .70 have been bolded.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Table 2
30 Hypotheses Suggesting Primals’ Potentially Pervasive Influence and Relevant Study 2 Results, Controlling for Age, Sex,
Education, and Income (Standardized s; N 524)
Rationale Hypotheses Result
In environments seen asX... ...Yisprudent/natural. In worlds seen as (IV; primal)... ...(DV) will be higher/lower.
Personality Psychology
1. Positive, safe, and
Optimism Good Optimism .66
Safe .59
Enticing .58
Alive .47
2. Abundant, beautiful,
fascinating, and worth
Curiosity and openness Enticing Curiosity (VIA) .61
Curiosity (CEI-II) .46
Openness (BFI) .37
3. Dangerous and miserable Worry and neuroticism Safe Neuroticism .43
Pleasurable .45
4. Fascinating, fun,
meaningful, and safe
Extraversion Safe Extraversion .37
Enticing .39
5. Meaningful and responsive
to effort
Tenacity and hard work Meaningful Grit .38
Improvable .29
6. Abundant, beautiful, and
made by someone to be
thankful to
Gratitude Enticing Trait gratitude .69
Abundant .63
Alive .44
Clinical Psychology
7. Miserable, boring,
meaningless, and with no
role for you
Depression and withdrawal Pleasurable Depression .52
Enticing .51
Needs Me.44
8. Threatening, miserable,
and punishing
Fear and anxiety Safe Anxiety .27
Stress .39
9. Unexpectedly unsafe Extreme stress Safe Post-traumatic stress
Positive Psychology
10. Safe, meaningful, and
Pursuing friends and intimacy Good More friends .24
Closer friends .51
11. Fun, meaningful, and
worth exploring
Engaging ones surroundings Enticing Engagement in life .57
Meaningful .49
12. Vile, dangerous, and
Dissatisfaction Safe Life satisfaction .51
Regenerative .48
13. Seeking your help Achievement Needs Me Accomplishment .52
14. Beautiful and full of
intention and purpose
Having a sense of meaning Beautiful Meaning in life .45
Alive .53
15. Positive, enjoyable, and
Wellbeing and positive emotions Good PERMA wellbeing .64
Understandable .40
16. Exciting Zest Enticing Zest No data
17. Positive and malleable Hope Regenerative Hope No data
18. Fascinating Love of Learning Interesting Love of Learning No data
Political Psychology
19. Cooperative rather than
cutthroat or Darwinian
Trusting people, institutions, etc. Safe Interpersonal trust .52
Cooperative .50
20. Declining Resisting change Progressing Conservatism .27
21. Hierarchical, fair, and
where the best win
Right-wing authoritarianism Hierarchical Conservatism .48
Just .34
22. Hierarchical, fair, and
where the worst fail
Preferring harsh punishment Hierarchical Pro-harsh sentencing .27
Just .14
23. Interconnected Holistic concerns Interconnected Fearing climate change .19
24. Hopeful and responsive to
change efforts
Supporting political campaigns Good Pro-political activism .34
Improvable .20
Vocational Psychology
25. Worth exploring,
meaningful, and
Better teaching outcomes and
Enticing Teaching success No data
Meaningful No data
Understandable No data
26. Dangerous, bleak, and
barren (vs. safe, inviting,
and abundant)
Success in low-failure/high-risk jobs
(vs. high failure/low-risk)
Safe Policing success No data
Abundant Sales success No data
Safe No data
Abundant No data
27. Fair and responsive Hard work Just Hard work
28. Those in authority see it Advancement The boss sees it Chance of promotion No data
(table continues)
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
across persons; and (c) substantially covary with variables they are
proposed to influence.
Studies below shed light on these benchmarks, as do two pre-
viously studied constructs that clearly qualify as primals. Belief in
a just world (BJW) holds that the world is a place where one gets
what one deserves and deserves what one gets (Lerner, 1980).
Over 30 studies using several validated scales suggest significant
relationships between BJW and many salient variables (e.g., Lip-
kus, 1991). These studies show that, causal or not, individuals act
in ways that appear optimal, given their primals. In general, those
high in BJW are more hardworking (since the world rewards
effort), more prosocial (since the world rewards kindness), more
successful (since they work harder and are nicer), and more likely
to blame victims (since suffering results from sloth or unkindness).
Over the years, strategies for measuring BJW scales have im-
proved. Early scales involved a relatively large number of items
concerning specific domains, such as sports or school (e.g., Rubin
& Peplau, 1975). However, Lipkus (1991) found that using items
concerning the world’s more general character increased scale
brevity and internal reliability.
The literature on belief in a dangerous world (BDW) is smaller,
not connected to the BJW literature, and primarily focused on
political ideology, especially right-wing authoritarianism. For ex-
ample, BDW predicts conservatism (Duckitt, 2001). However,
Perry, Sibley, and Duckitt’s (2013) scale items focus on societal
decline. More immediate aspects of danger, such as physical
threat, are absent. Still, a few BDW studies shed light on primals
more broadly. Two articles on split-second decision-making sug-
gest BDW is a System 1 (i.e., automatic) process measurable via
self-report (Miller, Zielaskowski, & Plant, 2012;Schaller, Park, &
Mueller, 2003). Schaller, Park, and Faulkner (2003) found that
darkness moderates BDW, suggesting ambiguity increases reliance
on primals. Finally, the relationship between beliefs about the
ubiquity of germs and one’s personal susceptibility to infection
suggests individuals think their own primals describe everyone’s
reality (Murray & Schaller, 2012).
Depending on standards for conceptual and nomological overlap,
other constructs may be considered primals-related. Candidates in-
clude general self-efficacy (Chen, Gully, & Eden, 2001), mastery
(Pearlin & Schooler, 1978), and constructs measured by Janoff-
Bulman’s (1989) World Assumptions Scale; Perry et al.’s (2013)
Competitive Worldview scale; and Dweck et al.’s (1995) mindset
scale. Studies below examine overlap between primals and these
In sum, though primals could theoretically play the pervasive role
Dweck (2008,2017) envisions, we know little. This is partly because
primals have not been systematically identified. This article seeks to
(a) systematically identify candidate primals; (b) determine their
structure using factor analysis; (c) validate a Primals Inventory; (d)
distinguish primals from other constructs; and (e) shed light on above
benchmarks (variability, stability, and covariance).
Item Generation
Candidate Primals Identification
To identify, as nearly as practical, an exhaustive list of exclusive
candidate primals, we used several mixed-method approaches to
catalogue instances of world description across (a) groups—who
said it—(b) era/location—when/where it was said—and (c) me-
dia—how/why it was said. In addition to explicit statements, we
identified implicit options via a lexical approach. We selected
methods based on coverage and convenience. Supplement 1 pro-
vides further details.
Textual analyses. We examined all mentions of world,uni-
verse,everything,nothing, and life in many of the world’s most
influential historical sources including 14 sacred texts, 100 novels,
100 films, 100 speeches, and 71 treatises. For each, we compiled
chunks (i.e., quotes) comprised of the sentence surrounding the
keyword, and coded chunks by candidate primal(s) via interrater
agreement. The result was 1,727 instances of world description,
sorted by candidate primal, source, era, geography, and so forth
Next, we gleaned 80,677 tweets beginning with The world is,The
universe is,orEverything is from a 2010 –2013 “random 1%
stream” database of 2.24 billion tweets (i.e., microblogs on twitter
.com). We used natural language processing tools to identify the
objects of these phrases and fed objects into Latent Dirichlet
Allocation (Blei, Ng, & Jordan, 2003). This yielded 50 clusters
(i.e., groups of similar phrases in similar contexts). To avoid
reliance on explicit statements, we analyzed the 840 most fre-
quently used adjectives in the Corpus of Contemporary American
English, a 450-million-word database of 190,000 texts curated
from five genres from 1990 –2012. After removing words irrele-
Table 2 (continued)
Rationale Hypotheses Result
In environments seen asX... ...Yisprudent/natural. In worlds seen as (IV; primal)... ...(DV) will be higher/lower.
Basic Information Processing
29. Imbued with purpose and
meaning by an active
designer or force (See
Spunt & Adolphs, 2015)
Social network activation w/o social
Alive Social network activity in
nonsocial contexts
No data
30. Numerous physical threats
(See Eberhardt et al.,
2004) Increased false threat percepts Safe False threat percepts No data
Note. Standardized s can be interpreted as rs controlling for age, sex, education, and income. We do not correct for multiple comparisons because (a)
our aim is to explore the scale of primals’ influence and not draw conclusions about any one variable and (b) with effect sizes this large pis less instructive.
We selected examples for illustrative purposes only; many additional (and stronger) effects are omitted.
See the BJW literature and Janoff-Bulman (1989) for data relevant data to these hypotheses.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
vant to world description, we organized the 469 that remained into
17 topics.
Expert retreats and interviews. Ten American experts on
beliefs, depression, and related topics met for three days at the
University of Pennsylvania to discuss and identify candidate pri-
mals. Chinese partners at Tsinghua University identified Confu-
cianism, Buddhism, Daoism, and traditional philosophy as central
Chinese traditions; identified experts in each; interviewed them;
and hosted a retreat in Beijing for these experts and Tsinghua
Focus group discussions. We conducted 10 focus groups
among Baltimore-area self-identified adherents of the world’s four
major religions (Christianity, Buddhism, Islam, and Hinduism).
Tsinghua partners conducted two focus groups among Chinese
graduate students in Beijing. We used semistructured, facilitated
discussions to prompt spontaneous expressions of candidate pri-
Literature and theory review. To identify previously stud-
ied primals outside of psychology, experts wrote reviews in
philosophy, cultural anthropology, art history, political science,
and comparative religion. Additionally, discovering little psy-
chological research identifying primals conducive to wellbeing,
we used a set of steps to systematically identify candidate
primals conducive to wellbeing, defined as 34 increased char-
acter strengths and emotions. For example, one step involved
reviewing all measures for items suggesting an associated pri-
mal (e.g., I find the world a very interesting place from Peterson
& Park, 2009).
Conceptual analysis. We synthesized these inputs in phases.
In each phase we reviewed new inputs, edited a working classifi-
cation of candidate primals, shared drafts with scholars, and in-
corporated feedback. Some drafts were merely lists. Others ad-
opted creative conceptual architectures, like the periodic table of
elements, to spot gaps. A list of 38 primals emerged in phase 4.
Phases 5 and 6 identified no new candidate primals, suggesting
saturation. To evaluate coverage, we estimated that our list cap-
tured 80% of world descriptions in each input. For example,
98% of the 1,727 gathered historical quotes we considered suffi-
ciently represented. In sum, though inputs and lists of candidate
primals can be interpreted and organized differently, our work
suggested low return on further efforts to identify candidate pri-
Item Creation
Measuring 38 constructs at once was infeasible. Also, some
candidate primals were infrequent, conceptually similar, or strug-
gled to satisfy criteria. Using these considerations, we prioritized
25 candidate primals for initial measurement. We also stopped
using acronyms (e.g., BJW) and began referring to candidate
primals by adjectives (Just,Beautiful, etc.).
Adopting an approach similar to Lipkus (1991) discussed above,
we generated an initial item pool of 500 items (20 per candi-
date primal) based on bounded definitions of each candidate pri-
mal; in situ language from the exploratory activities; items gleaned
from other scales (e.g., It’s a dog-eat-dog world . . . from Perry et
al., 2013); and generic repeatable formats (e.g., “On the whole, the
world is X”). Then, drawing on coauthors’ scale-building and
marketing backgrounds, we made assumptions regarding chal-
lenges particular to each candidate primal. For example, because
we expected few to disagree with items such as On the whole, the
world is a beautiful place, we increased difficulty for Beautiful
items (e.g., There is beauty everywhere, no matter where we look).
Aiming to reduce items to fewer than 250, we then shared items
with scholars mentioned in the acknowledgments and piloted with
11 online and 13 in-person participants. The result was the 234
items administered in Study 1 (9 per candidate primal, with 4
reverse-scored). We wrote scale instructions to frame items as
beliefs, rather than values, desires, or goals (see Table 4). Pilot
participants interpreted instructions as intended.
Research Studies
What follows are descriptions of six studies approved by the
University of Pennsylvania Institutional Review Board. Study 1, 2,
and 3 employ factor analyses to explore and confirm the measure-
ment model. Measuring an array of personality, clinical, and
wellbeing variables, Study 2 also examines convergent, divergent,
and incremental validity. Study 4 examines primals’ connection to
affect in response to a validity concern raised by Study 1 results.
Study 5 examines concurrent, convergent, and discriminant valid-
ity. Study 6 examines test–retest reliability and stability over 2
weeks, 9 months, and 19 months.
We administered surveys on Qualtrics from October 2015 to
March, 2017. Each included 5–10 attention checks. We discarded
responses for failing validation with 1 incorrect check or dupli-
cative IP addresses. Save for Study 3, we recruited samples via
mTurk, which Paolacci and Chandler (2014) consider suitable for
initial scale building since it provides cost-effective access to
diverse samples. We ipsatized primals data to correct for agree-
ment bias (Hicks, 1970). In checks, ipsatization aided factor inter-
pretability, simple structure, and internal reliability while ipsatized
versus nonipsatized factor scores correlated .97. Reverse-scoring
of items was done prior to analyses. See Supplement 2 for further
factor analysis results; Supplement 3 for the final version of the
Primals Inventory (PI-99); and Supplement 4 for further validity
Study 1: Exploratory Factor Analysis
The goal of Study 1 was to determine the factor structure of 234
items via exploratory factor analysis (EFA). We had two broad
expectations. First, since many candidate primals involve desirable
(or undesirable) states, a general Good factor might explain sub-
stantial variance. Through item diversity, we hoped the most
defining qualities of a Good implicit reality— beauty, justice, or
something else—might reveal themselves. Second, we expected
234 items would collapse into a handful of factors (5). We
expected few of the 25 candidate primals to emerge as their own
factor for seven reasons. (a) Items for each candidate primal
concerned conceptual opposites that may not hang together. For
example, we formulated the opposite of Beautiful as “ugly,” but it
could actually be something else like “disgusting” or “mundane.”
(b) We adopted a conservative approach of excluding factors
lacking at least one salient reverse-scored item. After all, if the
world is improving does not exclude the world is declining, valid-
ity is suspect. (c) Vague domain-general items may not allow
numerous independent data patterns because such items were used
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
to measure many primals at once (Lipkus (1991) had measured one
predefined primal). (d) Several candidate primals were conceptu-
ally similar (e.g., Improvable and Progressing). (e) We prioritized
wording differences among items measuring each candidate pri-
mal, while items across candidate primals often shared similar
language. (f) Item order was randomized between-participants. (g)
Finally, randomization spread items over a large pool (234 items)
concerning 50 ideas (25 pairs of conceptual opposites). Indeed,
some of these obstacles were erected to purposefully minimize
factor proliferation. Consequently, if EFA identified any pre-
defined candidate primals as independent, reliable factors, it would
imply an actual primal, and the factor should be retained.
Method and Participants
One thousand and ten participants completed 234 primals items;
a 20-item affect measure (PANAS; Watson, Clark, & Tellegen,
1988); a 13-item social desirability scale (Li & Sipps, 1985); and
16 demographic questions. 930 Americans age 18 to 74 (M36.7,
SD 11.7) passed validation and received $2.25. Of these, 57%
were female, 55% full-time workers, 79% white, 46% Christian,
41% nonreligious, 52% Democrat, 25% Republican, 51% college
graduates, and 58% with household incomes $40,000.
Data Analysis
To determine how many factors to extract, we used Minimum
Average Partial analysis, parallel analysis, scree analysis, and
“Bass-Ackwards” analysis (Goldberg, 2006). The latter involves
serially extracting and analyzing factor solutions until one extracts
the maximum number of reliable (.70) and meaningful fac-
tors. In “Bass-Ackwards” analysis and subsequent EFAs we used
PROMAX rotation, EQUAMAX prerotation, salience of .295 or
.395, and maximized hyperplane count; these methods typically
produced simpler structure. Oblique rotation was also considered
appropriate given expected worldview coherence pressures. Be-
cause above analyses disagreed, we explored the possibility of a
multilevel measurement model. Once we identified a factor struc-
ture (which required a second exploratory study), we retained
items to optimize (a) factor representation, (b) (all items must
increase ), and (c) brevity. We also considered item-total corre-
lations, means (i.e., difficulty), kurtosis, face validity, language
variety, item parameter slope estimates, item characteristic curves,
item information curves, test information curves, and relationships
with affect and social desirability.
Exploratory factor analysis suggested factor solutions at three
levels of granularity. Scree analysis of all items (see Figure 1)
suggested a one- or three-factor solution explaining 33% or 46% of
the common variance, respectively. Compared to two- and four-
factor solutions, they had fewer multiloading items, fewer non-
loading items, more reliable scores, and more meaningful factors.
We labeled the general factor Good. Its top-loading item was On
the whole, the world is an uncomfortable and unpleasant place
(reverse-scored) and Pleasurable items loaded highly. We labeled
factors in the three-factor solution Safe,Enticing, and Alive.
However, Minimum Average Partial analysis, parallel analysis, and
“Bass-Ackwards” analysis, pointed to numerous tertiary factors (28,
19, and 27, respectively). Seeking insight into how many factors to
extract at this more granular level, we separately analyzed reverse-
and forward-scored items. Again, scree analysis suggested the same
one or three superordinate factors and the other analyses indicated
many tertiary factors. Lacking indicator agreement on how many
tertiary factors to extract, we used the above criteria and additional
conservative criteria described below to create a preliminary subset of
items for further exploratory purposes. This subset resulted from
separate analyses of each of the 25 item sets associated with each
candidate primal in order to select only four items—at least one of
which had to be reverse-scored—per internally reliable, nonredundant
candidate primal. Standardizing the number of items per candidate
primal was helpful because low-loading and nonloading items were
unequally shared among candidate primals in the full 234-item bat-
tery. We judged that only four items per candidate primal would
discourage proliferation while still allowing factor emergence given
strong signal from a latent variable. To ensure factors were not
artifacts of highly correlating items, we deprioritized internal reliabil-
ity and removed 17 top-loading items with item-total correlations
above or near .80. These items could be added back once factor
structure was clarified.
Due to these analyses, one candidate primal was dropped for
failing to produce a set of items with .70 (Characterizeable);
two candidate primals were merged for being statistically re-
dundant (Pleasurable and Good); and one candidate primal was
dropped for lacking both salient forward- and reverse-scored
items (For Me). Thus, the resulting subset had 88 items (four items
per 22 candidate primals). For this subset, scree analysis again
suggested one and three factors; Minimum Average Partial anal-
ysis suggested 15; parallel analysis suggested 17; and “Bass-
Ackwards” analysis suggested 22. The 22-factor solution also
produced simple structure, defined as 0 multiloaders, 0 nonloaders,
and all factors with .70. These factors reflected the 22
remaining predefined candidate primals. We further explored pri-
mals’ superordinate structure by analyzing the 22 most represen-
tative items in each 4-item scale (scree analysis suggested 1 and
three factors; Minimum Average Partial analysis: 3; parallel anal-
ysis: 4) and 22 scores on the 22 scales (i.e., a hierarchical EFA;
scree analysis: 1 and 3 factors; Minimum Average Partial analysis:
3; parallel analysis: 6). Across 26 primals (one primary three
secondary 22 tertiary scales), pairwise relationships between
primals and social desirability were small (Mr.18, range: .08
to .31). Some relationships with affect were larger (positive affect:
1 3 5 7 9 11131517192123252729313335
Figure 1. Scree plot of 234 primals items in Study 1.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Mr.27, range: .01 to .45; negative affect: Mr⫽⫺.22,
range: .39 to .10).
Primals emerged at three levels of granularity. Each level makes
steeper tradeoffs between variance explained and parsimony, with
new factors emerging at more granular levels (and some at less).
The primary primal Good (␣⫽.97), largely defined by Pleasur-
able items—not Beautiful or Just explained 45% of common
variance among the 88-item subset. Secondary primals Safe (␣⫽
.96), Enticing (␣⫽.95), and Alive (␣⫽.90) explained 65%.
Twenty-two tertiary factors (mean ␣⫽.87; SD .05) slightly
overextracted, explaining 105%.
As noted, we had not expected so many tertiary factors to
emerge. To identify the exact number to extract, we created a
subset of items for exploratory purposes. This involved various
techniques that weakened tertiary factors to increase flexibility in
the correlation matrix. (Once factor structure was clarified, high-
performing items would be added back to strengthen psychometric
characteristics.) These analyses required numerous decisions—a
weakness of the approach—and brought clarity. Despite obstacles,
a 22-factor solution produced simple structure with each factor
including only four items, at least one of which was opposite-
scored. Because one- and three-factor solutions underextracted
variance, ignored meaningful complexity, lacked simple structure
with numerous nonloading items, and poorly represented several
tertiary factors—Interconnected,Understandable,Acceptable,Hi-
erarchical, and Changing—we concluded that the most accurate
solution was likely the improbable 22-factor solution. However,
we were worried by slight overextraction, the possibility of over-
fitting, and low incremental validity for tertiary primals. We also
wondered if the tertiary model would fall apart if only 4 items were
administered per factor. Thus, with key questions remaining, we
chose to delay settling on a final version of the Primals Inventory.
Instead, we would use the 88-item subset that achieved simple
structure for further exploratory purposes in Study 2. Study 1
results also raised a validity question concerning affect that Study
4 examines further.
Study 2: Confirmatory Factor Analysis and Validity
The main goal of Study 2 was to clarify factor structure so
item-retention decisions could be finalized. For easier reading,
Study 2 analyses, results, and discussions are organized by topic:
(a) factor analyses; (b) an examination of the relative predictive
utility of primary versus secondary versus tertiary primals; (c)
convergent and divergent validity; and (d) incremental validity.
Topics are followed by a general discussion of Study 2’s main
goal: finalizing the measurement model.
Method and Participants
Following primals items, 562 participants completed measures
of BIG 5 personality traits (44-item BFI; John, Donahue, & Kentle,
1991); life satisfaction (5-item SWLS; Diener, Emmons, Larsen, &
Griffin, 1985); depression, anxiety, and stress (21-item DASS-21;
Antony, Bieling, Cox, Enns, & Swinson, 1998); positive emotions,
negative emotions, engagement, relationships, meaning, accom-
plishment, health, and overall wellbeing (23-item PERMA-
Profiler; Butler & Kern, 2016); curiosity (10-item CEI-II; Kashdan
et al., 2009; 10-item VIA; Peterson & Park, 2009); grit (8-item
Grit-S; Duckworth & Quinn, 2009); optimism (6-item LOT-R;
Scheier, Carver, & Bridges, 1994); belief in a just world (6-item
GBJWS; Lipkus, 1991); gratitude (6-item GQ-6; McCullough,
Emmons, & Tsang, 2002); Machiavellianism, narcissism, and psy-
chopathy (12-item “Dirty Dozen”; Jonason & Webster, 2010);
growth mindset (4 items; Dweck et al., 1995); 24 political ques-
tions; mystical experiences (5-item DT subscale; Hood & Morris,
1983); and 27 demographic questions.
Five hundred twenty-four Americans, age 18 to 75 (M36.9,
SD 11.5), passed validation and received $2.10. Of these, 97%
were native English speakers, 51% female, 78% white, 40% mar-
ried, 40% Christian, 40% agnostic or atheist, 43% Democrat, 18%
Republican, 51% college graduates, and 59% with household
incomes $40,000. 48 of 50 states were represented.
Confirmatory Analyses
We examined fit for one-, three-, and 22-factor models.
Analysis. As Schreiber, Nora, Stage, Barlow, and King
(2006) recommend, we assessed fit with Confirmatory Factor
Analysis (CFA) using thresholds of .95 for Non-Normed Fit
Index (NNFI) and Comparative Fit Index (CFI), and .08 for
Root Mean Square Error of Approximation (RMSEA). How-
ever, CFAs suggest artificially poor fit given (a) lengthy sur-
veys (30 items); (b) items correlating with 1 factor; and
(c) complex models (Floyd & Widaman, 1995). Though we
expected mediocre fit for 1- and 3-factor models (since they
ignore tertiary variance), these three problems could be over-
come by parceling items to distribute Study 1 PROMAX load-
ings and dimensionality, as Kishton and Widaman (1994) ad-
vise. However, since this was impossible for the 22-factor
model, fit statistics were deemed uninformative. (The common
1:10 parameter to observation guideline (e.g., Schreiber et al.,
2006) calls for 4,290 participants). Thus, we supplemented CFA
with exploratory structural equation modeling; identical EFA
procedures; and comparisons of item parameter slope estimates,
item characteristic curves, item information curves, and test
information curves. Administering only 4 items per tertiary
primal also allowed another chance for tertiary primals to fail.
If remained acceptable, it would support tertiary-level extrac-
Results. CFA indicated acceptable fit for 1- and 3-factor mod-
els and other analyses supported tertiary extraction. NNFI, CFI,
and RMSEA were, respectively, .980, .984, and .061 for the
1-factor model and .956, .964, and .072 for the 3-factor model. A
22-factor exploratory structural equation model would not con-
verge, suggesting the need for a larger sample. EFA revealed 1-,
3-, and 22-factor solutions nearly identical to those in Study 1.
Minimum Average Partial analysis and parallel analysis again
suggested many factors (15 and 23) while scree analysis again
suggested 1 and 3. A 1-factor solution explained 39% of the
common variance, a 3-factor solution 57%, and a 22-factor solu-
tion 96%. Mean was .96 for primary, .91 for secondary, and .80
for tertiary scales (range: .67 to .86, SD .07; only About Me fell
below .70).
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Discussion. CFA indicated adequate fit for one- and three-
factor models. The other analyses largely confirmed the 22-
factor solution, which no longer overextracted common vari-
ance. Also, despite obstacles, remained acceptable (.70) for
nearly all four-item tertiary factors. These results suggested a
measurement model should include primary, secondary, and
tertiary factors.
Predictive utility of tertiary versus secondary versus pri-
mary primals. We were interested in the possibility of dropping
the primary, secondary, or tertiary levels from our model for the
sake of simplicity. Thus, we compared the predictive utility of
each level in exploratory regression analyses examining a variety
of dependent variables (DVs).
Analysis. We created several three-way prediction contests
using all 21 nonprimals scales as DVs and three atheoretical
methods for predictor variable selection. DVs included personality
and wellbeing indicators (extraversion, agreeableness, conscien-
tiousness, neuroticism, openness, curiosity measured in two ways,
Machiavellianism, psychopathy, narcissism, optimism, grit, trait
gratitude, growth mindset, interpersonal trust, life satisfaction,
transcendent experiences, overall PERMA wellbeing, depression,
anxiety, and stress). Predictor variables were selected from either
1 primary, 3 secondary, or 22 tertiary primals, as well as age, sex,
race, education, marital status, employment status, family income,
personal income, parental SES, personal SES, religiosity, spiritu-
ality, religion, conservatism, party affiliation, number of friends,
English as a second language, U.S.A. resident, and Body Mass
Index. The first of our atheoretical predictor-variable selection
methods was the nonparsimonious kitchen sink method, in which
we simultaneously entered primals and all other predictor vari-
ables. Second, in a reduced method, we included only significant
variables from the kitchen sink model (p.05). Third, in a ridge
regression method, we iteratively entered and removed variables
from the kitchen sink method to identify the model that optimized
parsimony according to the “lambda 1-SE” criterion from the R
package glmnet (Friedman, Hastie, & Tibshirani, 2010). Using
Bayesian Information Criterion (BIC), a fit index that penalizes
overfitting, we ranked primary, secondary, or tertiary models 1st,
2nd, and 3rd by how well DVs were predicted. Raftery (1995)
suggests BIC reductions of 0 –2, 2– 6, 6 –10, and 10 denote weak,
positive,strong, and very strong evidence of improved fit, respec-
Results. No level of granularity underperformed the others.
Across contests (21 DVs 3 methods of predictor variable selec-
tion 63 contests), primary models provided the best fit 22 times,
second best 23 times, and worst 18 times. Secondary models were
best 10 times, second best 36 times, and worst 17 times. Tertiary
models were best 31 times, second best 9 times, and worst 23
times. Mean BIC difference between best and second-best models
was 15.6 (SD 19.1) and between best and worst was 45.3 (SD
28.0), denoting very strong evidence of improved fit.
Discussion. Our examination of relative predictive utility
found no basis for discarding any level of granularity. Though BIC
typically provided very strong evidence that one level of granu-
larity was considerably more useful in predicting specific DVs
than others, primary models were worst about as often as they were
best or second best; tertiary models were most often best and
worse; and secondary primals were a compromise, usually not
worst or best. All were useful.
Convergent and Divergent Validity
Analysis. To assess convergent and divergent validity, we
examined pairwise correlations between the 26 PI scales and 115
personality, clinical, wellbeing, and demographic variables.
Results. Results suggest convergent validity. Strong examples
include the four relationships proposed in the theory section: Good
and optimism, r.67, p.001; Safe and neuroticism, r⫽⫺.44,
p.001; Enticing and VIA curiosity, r.61, p.001; and
Progressing and conservatism, r⫽⫺.14, p.001. Table 2
highlights 40 additional examples. Results also suggest divergent
validity. Examples include Alive and Body Mass Index (r⫽⫺.03)
and Funny and family income (r.01). No relationships violated
directional assumptions (,, or orthogonal). A few surprises
occurred when correlations were higher or lower than expected. In
the most extreme cases, Machiavellianism covaried with Cooper-
ative less than expected (r⫽⫺.11 p.01) and trait gratitude
covaried with Good slightly more than expected, r.68, p
Discussion. Across the board, study participants reported be-
haviors that, causal or not, appear optimal given their primals. In
addition to validating our PI interpretations, these results shed light
on one of Dweck’s (2008,2017) critical benchmarks for whether
a belief could exert a pervasive influence on personality: does the
belief substantially covary with variables it theoretically influ-
ences? Though such relationships provide no evidence of causal-
ity, substantial and largely expected patterns of covariance were
observed (see Table 2 for many examples).
Incremental Validity
We have proposed as part of our tentative measurement model
that primals, like other beliefs, may influence behaviors associated
with major personality traits yet are distinct from those traits. To
explore overlap with these traits (and primals’ relative predictive
power at the primary, secondary, and tertiary level), we pitted
primals against the well-known Big Five Inventory in a series of
exploratory regression analyses.
Analysis. For each non-BIG 5 DV used above, we fit a base-
line model using the 19 remaining demographic variables as pre-
dictors. Adding to this baseline, we fit a BIG5-only model, three
primals-only models (at each level of granularity), and three BIG5-
and-primals models (at each level of granularity) with all variables
added simultaneously. We compared model fit using Bayesian
Information Criterion (BIC), likelihood ratio tests, and standard-
ized r
Results. Primals usually predicted DVs better than the BIG 5
or improved on the BIG 5. Likelihood ratio tests (all p’s .05)
showed (a) 47 of 48 primals-only models fit better than demo-
graphic baseline models (secondary primals did not help predict
narcissism); and (b) 43 of 48 BIG5-and-primals models fit better
than BIG5-only models (no primals models helped predict grit;
only the tertiary model helped predict Machiavellianism). As
Table 3 shows, 21 of these 43 models had BIC differences 10.
Finally, based on BIC, primals-only models outperformed the
BIG5-only models outright in 11 of 48 cases. Notably, Good alone
explained 1.6% more variance in growth mindset and 3.3% more
in life satisfaction (BIC differences were 10.9 and 49.0, respec-
tively). Safe,Enticing, and Alive explained 11.5% more variance in
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
trait gratitude and 6.2% more in interpersonal trust (BIC differ-
ences were 134.8 and 56.5, respectively).
Discussion. Primals are different from—and in some cases
much better predictors than—the BIG 5. Primals explained sub-
stantial variance over and above the BIG 5 when predicting 90%
of DVs, including several well-studied variables like wellbeing
and depression. Notably, primary and secondary primals-only
models outperformed the BIG 5 when predicting life satisfaction,
trait gratitude, growth mindset, and interpersonal trust.
Study 2 General Discussion
As noted, we had created Study 2’s version of the Primals
Inventory for further exploratory purposes in a new sample. In
light of evidence supporting the primary, secondary, and tertiary
factor models and convergent, divergent, and incremental validity
at each level of granularity, we decided to retain all levels. Thus,
returning to Study 1 data, we made final item-retention decisions
based on original criteria, such as prioritizing . This resulted in (a)
exchanging some items for higher-loading, higher-intercorrelating
items and (b) adding fifth items to a few weaker tertiary scales. To
finalize superordinate scales, we examined one- and three-factor
solutions of the full 234-item battery to ensure top-loading items
with variance specific to superordinate primals were included. As
a result, we retained one item for Good and two for Enticing
(top-loading items for Safe and Alive were already among tertiary
scale items). The result was the final 99-item version of the
Primals Inventory (PI-99) in Table 4. Among final items, the
22-factor solution still had 0 multiloading items and 0 nonloading
items (i.e., simple structure; salience was .295). The 3-factor
solution had 0 multiloaders and 27 nonloaders and the 1-factor
solution had 28 nonloaders (salience was .395). Hyperplane count
was optimized at 1,874 for the 22-factor solution (k3) and 91
for the 3-factor solution (k3). For PROMAX loadings, see
Supplement 2. To gauge if final decisions strengthened psycho-
metric characteristics without altering factor structure or meaning,
studies below include another CFA, a test-retest comparing PI-99
scores to Study 2 scores, and further validity checks.
Study 3: A Second Confirmatory Factor Analysis
Because mTurk has limitations as a sample, Study 3 aimed to
replicate results elsewhere.
Method and Participants
We copied items, validation checks, and format from Study 1.
580 participants participated for no compensation via authen-, where one of 8 rotating banner headlines
invited study participants over 18 months. 529 Americans, age 18
to 77 (M43.4, SD 15.2) passed validation. Of these, 88%
were native English speakers, 77% female, 48% full-time workers,
74% white, 42% Christian, 33% agnostic or atheist, 35% Demo-
crat, 11% Republican, 65% college graduates, and 76% with
household incomes $40,000.
We conducted confirmatory procedures identical to Study 2.
Results largely mirrored Study 1 and 2. Mean for 26 PI-99
subscales was .84 (range: .69 to .97). NNFI, CFI, and RMSEA
were .963, .970, and .090 for the one-factor solution and .968,
.974, and .067 for the three-factor solution. EFA produced ex-
pected one-, three-, and 22-factor solutions (though without simple
structure) that explained 36%, 50%, and 86% of common variance,
respectively. Minimum Average Partial analysis and parallel anal-
ysis again suggested numerous factors (between 15 and 56). Scree
Table 3
Incremental Validity of Primals When Added to BIG 5 Based on Bayesian Information Criterion (BIC) Decrease and Percentage of
Additional Variance Explained
16 Dependent variables
BIC Decrease % Additional variance explained°
Primary Secondary Tertiary Primary Secondary Tertiary
Depression 44.8 33.2 41.8 4.9 4.7 6.7
Anxiety 1.9 8.1 92.6 .4 .9 2.5
Stress 1.6 5.6 96.1 .3 .9 1.5
Curiosity (CEI-II)
22.2 12.7 72.5 3.1 3.2 4.4
Curiosity (VIA)
55.5 57.0 22.9 6.0 7.1 8.6
Trait Gratitude 120.5 147.4 66.5 11.6 14.5 15.7
Grit 2.9 13.8 119.5 .1 .1 .5
Growth Mindset 10.0 4.8 59.1 2.5 3.3 8.3
Life Satisfaction 60.3 49.3 31.6 7.2 7.2 8.8
Machiavellianism 6.2 13.8 100.2 .2 .2 1.6
Psychopathy 1.8 3.8 74.0 .4 1.1 3.6
Narcissism .7 .3 83.8 .6 2.0 3.7
Optimism 98.1 86.7 4.6 7.7 7.7 8.7
Overall Wellbeing 86.5 78.0 10.3 6.9 7.0 9.0
Transcendence 2.1 21.1 15.7 .5 5.7 18.4
Trust 40.7 80.9 4.5 6.2 12.4 14.7
Note. All models include demographics. See Supplement 4 for full results.
° Based on standardized r
We used the Curiosity and Exploratory Inventory-II (CEI-II) and the Values in Action (VIA) curiosity subscale.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Table 4
The Final 99 Item Primals Inventory (PI-99)
Instructions: Below are very general statements about the world—not the world we wish we lived in, but the actual world as it is now. Please share
your sense of agreement or disagreement. When in doubt, go with what initially feels true of the real world. There are no wrong answers. There’s no
need to overthink. [Item order was randomized for each participant. Response options were strongly agree,agree,slightly agree,slightly disagree,
disagree, and strongly disagree.]
About Me vs. not about me
- Whatever is happening around me often feels related to me or
something I’ve done.
- Much of what happens around me feels like it’s because of
me or related to me somehow.
- My first instinct about events happening around me is that
they’re unrelated to me or anything I’ve done.
- When unsure why something is happening, I often suspect
it’s got something to do with me.
- My first instinct about things happening around me is that
they have to do with me or something I’ve done.
Abundant vs. barren
- The world is an abundant place.
- Life overflows with opportunity and abundance.
- The world feels like a barren place with few opportunities.
- The world is an abundant place with tons and tons to offer.
Acceptable vs. unacceptable
- The world needs to be continually improved rather than
- Rather than accepting things as they are, the world needs to
be improved as much as possible.
- It’s usually better to accept a situation than try to change it.
- Most situations in life need to be improved, not accepted.
Beautiful vs. ugly
- Nearly everything in the world is beautiful.
- Though some things are incredibly beautiful, they’re few and
far between.
- There is beauty everywhere, no matter where we look.
- In life, there’s way more beauty than ugliness.
Changing vs. static
- Everything feels like it’s shifting and changing.
- Everything feels like it’s constantly moving, changing, and up
in the air.
- Everything feels like a whirl of constant change.
- I feel like everything changes all the time.
- The world is a place where most things stay pretty much the
Cooperative vs. competitive
- Instead of being cooperative, life is a brutal contest where
you got to do whatever it takes to survive.
- For all life—from the smallest organisms, to plants, animals,
and for people too—everything is a cut-throat competition.
- Instead of being cooperative, the world is a cut-throat and
competitive place.
- The world runs on trust and cooperation way more than
suspicion and competition.
Funny vs. not funny
- The world is hilarious; if we aren’t laughing, we aren’t
paying attention.
- Laughing a ton makes sense because life is hilarious and
humor is everywhere.
- While some things are humorous, most of the time the world
is not that funny.
- There’s humor in everything.
Harmless vs. dangerous
- On the whole, the world is a safe place.
- Real danger is everywhere; even if we don’t notice it.
- Most things and situations are harmless and totally safe.
- I tend to see the world as pretty safe.
- On the whole, the world is a dangerous place.
Hierarchical vs. nonhierarchical
- Most things can be organized into hierarchies, rankings, or pecking
orders that reflect true differences among things.
- Humans, animals, plants, and pretty much everything else can be
organized by how important or good they are.
- Most things aren’t better or worse. It’s hard to organize the world into
hierarchies, rankings, or pecking orders that reflect true differences.
- Most things in the world could be ranked in order of importance.
- Things are rarely equal. Most plants and animals, and even people, are
better or worse than one another.
Improvable vs. too hard to improve
- It’s possible to significantly improve basically anything encountered in
- Most situations seem really difficult if not impossible to improve.
- No matter who you are, you can significantly improve the world you live
- In most situations, making things way better is absolutely possible.
- Most things and situations are responsive, workable, and totally possible
to improve.
Intentional vs. unintentional
- Events happen according to a broader purpose.
- What happens in the world is meant to happen.
- Everything happens for a reason and on purpose.
- Events seem to lack any cosmic or bigger purpose.
- The universe doesn’t care if events happen one way or another.
Interconnected vs. atomistic
- Every single thing is connected to everything else.
- Most things are basically unconnected and independent from each other.
- Though things can appear separate and independent, they really aren’t.
Instead, all is one.
- The world is a place where everything is completely interconnected.
Interesting vs. boring
- The world is a somewhat dull place where plenty of things are not that
- Most things in life are kind of boring.
- It feels like interesting and exciting things surround us all the time.
- While some things are interesting, most things are pretty dull.
Just vs. unjust
- On the whole, the world is a place where we get what we deserve.
- Life will find ways to reward those who do good and punish those who
do bad.
- The world is a place where working hard and being nice pays off.
- If someone is generous and kind, the world will be kind back.
- The world is a place where we rarely deserve what we get.
Meaningful vs. meaningless
- The world is a place where most everything matters.
- Nothing really matters all that much.
- Most things are pointless and meaningless.
- The world is a place where things just don’t matter.
Needs Me vs. doesn’t need me
- The universe needs me for something important.
- Life has an important part for me to play.
- It feels like the world doesn’t really need me for anything.
- The world needs me and my efforts.
Pleasurable vs. miserable
- Life offers more pain than pleasure.
- On the whole, the world is a good place.
- Life in this world is usually pain and suffering.
- Life offers way more pleasure than pain.
- Most things in the world are good.
(table continues)
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
plot analysis suggested three and perhaps four. A four-factor
solution included Safe,Enticing, and Alive and a factor of Hier-
archical, Interconnected, and Changing items with PROMAX
loadings ranging from .49 to .62 (␣⫽.77).
Study 3 replicated the PI-99’s factor structure in a non-mTurk
sample that participated without monetary incentive and was com-
paratively older, less likely to be employed, wealthier, more edu-
cated, and had a higher percentage of women and non-native
English speakers. Exceptions included a RMSEA on the one-factor
model slightly above threshold; an for one of 26 primals (Worth
Exploring) slightly below .70 at .69; and an intriguing (if remote)
possibility of a fourth secondary primal concerning nonhierarchy,
connection, and flux.
Study 4: Validity—Exploring the Relationship
Between Primals and Affect
Are PI scores partially driven by ambient emotion? Perhaps
Good or Safe scores merely reflect the sort of day one is having or
the mood one is in. As noted, Study 1 found small and some
moderate correlations with affect. Among potential explanations,
we explored the possibility that, since PANAS followed the PI in
Study 1, perhaps reflecting on primals alters affect rather than
affect altering PI scores. Indeed, some study participants noted that
taking the PI-99 altered their mood; reflection on evaluative beliefs
should theoretically alter affect; and the covariance pattern among
primals was consistent with this view. For example, if ambient
emotion drove PI scores, Good should presumably be most related
to affect. Instead, among all 26 primals, Needs Me best predicted
positive affect despite attenuation (less reliable scales are inher-
ently less able to correlate highly with other variables). Yet, Needs
Me was comparatively less related to Good (15th in Study 1) and
third variables likely to serve a mediational role (see Study 2).
Thus, we designed Study 4 is an exploratory randomized experi-
ment examining the role of sequence effects on positive affect and
PI scores, especially Good and Needs Me.
Method and Participants
Two hundred thirty participants answered the PI-99, PANAS,
and 10 demographic questions. 219 participants age 19 to 70 (M
38.4, SD 12.3) passed validation and received $0.75. Of these,
55% were female, 79% white, 47% Christian, 44% secular, 39%
Democrat, 23% Republican, and 63% with household in-
comes $40,000. We randomly assigned 120 participants to a
PI-first condition and 99 to a PANAS-first condition. Compared to
the PI-First condition, the PANAS-first condition was more Chris-
tian (55% to 41%) and specifically Catholic (32% to 11%); white
(85% to 74%); and Republican (30% to 17%).
After checking if Study 1 results replicated in the PI-first con-
dition, we compared relationships between primals and positive
affect in the PI-first condition to those in the PANAS-first condi-
tion, focusing on Good and Needs Me. Since only effect-size
differences .24 would be statistically significant (p.05),
nonsignificant results may yield insights, especially if consistent
across primals.
Among those taking the PI-99 first, there was typically a
larger relationship between affect and PI-99 scores. Results
from the PI-first condition largely replicated Study 1 results,
including substantial covariance between positive affect and
(Good,r.46, p.001 and Needs Me,r.47, p.001). In
Table 4 (continued)
Progressing vs. declining
- On the whole, the world is getting worse.
- It feels like the world is going downhill.
- Though the world has problems, on the whole things are
definitely improving.
- It feels like the world is getting better and better.
Regenerative vs. degenerative
- Though sometimes situations get worse, usually they get
- Most things have a habit of getting worse.
- The usual tendency of most things and situations is to get
better, not worse.
- Over time, most situations naturally tend to get worse, not
Stable vs. fragile
- The world is a place where things are fragile and easily
- It takes a lot for things to fall apart.
- Most things and situations are delicate and easily
- Most situations are delicate. Though they may be fine now,
things could easily unravel.
Understandable vs. too hard to understand
- Most everything is easy enough to understand.
- The world is a confusing place where many skills and subjects are too
hard to figure out.
- Lots of things in the world are too confusing and difficult to
- The world is easy enough to understand.
Worth Exploring vs. not worth exploring
- I feel everything is worth trying, learning about, or exploring further.
- To be honest, though some things are worth trying and exploring, most
things aren’t.
- Everything deserves to be explored.
- Unfamiliar things and places are usually worth trying or checking
Enticing additional items
- No matter where we are or what the topic might be, the world is
- No matter where we are, incredible beauty is always around us.
Good additional item
- On the whole, the world is an uncomfortable and unpleasant place.
Note. Copyright 2018 by Jeremy D. W. Clifton.
71 Good items.
29 Safe items.
28 Enticing items.
14 Alive items.
39 reverse-scored items.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
the PANAS-first condition, these relationships dropped by .15
and .14, respectively, though changes were not significant
(Good:z1.26, p.21; Needs Me:z1.21, p.22). We
observed significant differences for Worth Exploring,r.26,
z1.98, p.048 and Intentional,r.25, z1.99, p.047
and marginally significant changes in Stable,r.25, z 1.85,
Results suggest brief reflection on one’s primals may influence
emotion without determining it, suggesting part of the relationship
observed in Study 1 is due to a sequence effect. However, before
conclusions can be drawn, this study should be replicated in a
larger, non-mTurk sample (specifically, one less used to taking
surveys). By examining stability, Study 6 also sheds light on the
relationship between primals and states, including affect.
Study 5: Convergent, Concurrent, and Discriminant
Our literature review noted that nomological overlap with re-
lated constructs (including two previously studied primals) should
be explored once primals are measurable. Study 5 examines these
points of overlap to further validate our interpretation of PI-99
Method and Participants
Two hundred thirty-one Americans completed the PI-99 and 10
demographic questions. 191 passed validation, received $0.10, and
were invited two weeks later to take a survey which included the
PI-99 and measures of Justice, Luck, Controllability, Randomness,
and Benevolence (32-item WAS; Janoff-Bulman, 1989); belief in
dangerous and competitive worlds (20-item SWS-R; Perry et al.,
2013); incremental theories of intelligence, morality, self, and
world (12-items; Dweck et al., 1995); general self-efficacy (8-item
NGSES; Chen et al., 2001); mastery (7-item Pearlin Mastery
Scale; Pearlin & Schooler, 1978); and purpose, comprehension,
and mattering (15-item MEMS; George & Park, 2017). Of 144
completing the second survey, 122 participants age 18 to 74 (M
35.4, SD 12.4) passed validation, receiving $2. Of these, 44%
were female, 52% white, 52% Christian, 22% nonreligious, 39%
Democrat, 23% Republican, 64% college graduates, and 57% with
household incomes $40,000.
To assess concurrent, convergent, and discriminant validity, we
examined if pairwise correlations between primals and 19 addi-
tional variables violated directional assumptions or were much
higher or lower than expected. Results are organized by construct.
Results and Discussion
As in Study 2, no correlations violated directional assumptions
and effect sizes showed discriminant validity. In what follows, p
.001 unless stated otherwise.
Belief in a Just World. The expectation that BJW and Just
are the same latent variable was supported. Among tertiary pri-
mals, Just best predicted WAS’s Justice subscale at .75. (Lipkus’s
(1991) BJW scale correlated with PI-94’s 4-item Just scale at .74
in Study 2).
Belief in a Dangerous World. Though Safe and BDW osten-
sibly concern the same latent variable, we expected Perry et al.’s
(2013) BDW scale to target Progressing and somewhat neglect
Safe and Safe-related tertiary primals. As expected, BDW pre-
dicted Progressing at .72, which was more than Safe (.68) and
Safe-related tertiary primals: Pleasurable (49), Regenerative
(.53), Harmless (.67), Cooperative (.59), Stable (.39), and
Just (.34).
World Assumptions Scale. Janoff-Bulman’s (1989) WAS
includes five relevant scales. We thought Benevolence would
concern the same latent variable as Safe (not Good) and that
Justice, Controllability, Luck, and Randomness would be largely
(but not completely) explained by Just. After all, a just world is by
definition a controllable, nonrandom place where one is not per-
petually unlucky. As expected, Just was the tertiary primal that
best predicted all four scales: Justice (.75), Controllability (.53),
Luck, (.49), and Randomness (.49). Further studies might ex-
plore the unique contributions of WAS scales in more detail.
Competitive-Jungle Worldview Scale. Because CJWS items
include many value statements, we expected it would be related to
but distinct from Cooperative, which was supported. Cooperative
correlated with the CJWS at .58.
Improvable-related measures. Numerous constructs may
seem related to Improvable, including mastery, global self-efficacy,
and four forms of incremental theory. However, since a belief about
the malleability of an environment could differ from a self belief about
one’s own ability to change it, we expected Improvable would be
related to yet distinct from these measures, which was supported.
Improvable was the primal best predicting mastery (.64) and general
self-efficacy (.59). Incremental theories were much less related (in-
telligence: r.23, p.01; personality: r.24, p.008; morality:
r.31, p.005; world: r.23, p.01).
In addition to further evidence of convergent and discriminant
validity, Study 5 indicates the gap in the nomological network
mirrors the gap our literature review suggested. In sum, two
primals have been studied—though BDW may be mislabeled—
and several related constructs correlate with (but are distinct from)
primals. It is also worth noting that whereas Safe is related to the
two previously studied primals, Enticing and associated tertiary
primals are particularly virgin territory.
Study 6: Test-Retest Reliability and Stability
Study 6 examined test-retest correlations across 2 weeks, 9
months, and 19 months.
Method and Participants
The 2-week test-retest sample was the same sample used in
Study 5. We created a 9-month sample (March to Dec. 2016) by
randomly inviting 250 participants from Study 2 (N524). Of
136 who completed the second survey, 134 people, age 20 to 72
(M38.6, SD 12.5), passed validation and received $1.25.
They were 54% female, 85% white, 44% Christian, 45% secular,
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
46% Democrat, 21% Republican, 58% college graduates, and 63%
with household incomes $40,000. We created a 19-month sam-
ple (Oct. 2015 to May 2017) by inviting all Study 1 participants
(N930). Of 406 who completed the second survey, 398 people,
age 21 to 75 (M41.3, SD 12.2) passed validation and received
$2. They were 55% female, 79% white, 49% Christian, 45%
secular, 46% Democrat, 24% Republican, 59% college graduates,
and 63% with household incomes $40,000.
For each primal, we examined rs between time 1 and time 2 in
Table 5.
PI-99 scores showed test–retest reliability and stability. The
2-week, 9-month, and 19-month test-retest coefficients were, re-
spectively, .90, .76, and .78 for Good; averaged .88, .75, and .77
for secondary primals; and averaged .71, .62, and .65 for tertiary
primals. In 19-month retest voluntary comment boxes, several
participants also noted the impact of major recent events. For
example: (a) If Hillary Clinton had won, I’d be answering totally
differently. (b) The world is a lot scarier since your last survey
what with Trump and his vile cohort. (c) This was terrible timing.
The terror attack in Manchester last night is weighing heavily on
Turkers’ minds today as reflected in our message boards and
makes us think the world is unsafe. Keep that in mind when you
wonder why we have all become negative and depressed.
PI-99 scores were remarkably stable over time. In addition to item
randomization, which should lower test-retest coefficients by altering
item context each time, minor study design flaws likely suppressed
results in two of three retests. First, seeking retention in the 2-week
retest, we underpaid at Time 1, overpaid at Time 2, and data quality
suffered. Compared to Studies 1, 2, and 4, it took 16 times longer to
recruit fewer participants who were 3 times more likely to fail vali-
dation; mean fell .06 across PI-99 scales. Second, we suspect that
many of the 19 subscales that appeared more stable over 19 months
than 9 months only did so because the 9-month retest compared
alternate versions of the inventory (differences between Study 2’s
initial version of the PI and the finalized PI-99 affected 19 of 26
subscales). Thus, the 19-month retest is likely the most accurate. It
also involved the largest sample.
General Discussion
This article supports two strong conclusions. First, the Primals
Inventory had excellent psychometrics for these online samples.
Second, primals met critical benchmarks that indicate a broader
research phase is justified.
The Primals Inventory
Because previous research has not systematically defined or
identified primals, we suspected that many remained unspecified.
To establish content validity, we used various methods to identify
candidate primals across groups, eras, and media. For example, our
textual analyses involved examining 358 highly influential histor-
ical texts, analyzing 80,677 tweets, and sorting the 840 most-used
adjectives in American English. Study 1 revealed a three-level
model— one primary, three secondary, and 22 tertiary primals—
that was replicated in one mTurk and one nonmTurk sample
(Study 2 and 3). Each level trades parsimony for variance ex-
plained (35% vs. 55% vs. 90%). We retained all levels for
psychometric and incremental validity reasons.
Across nearly 3,000 study participants, PI-99 subscales showed
remarkably high internal reliability (see Table 1). Mean was .97
for Good, .96 for Safe, .95 for Enticing, and .89 for Alive. Though
each tertiary scale had only 4 –5 items (including at least one
opposite-scored item) and items were fairly vague and randomized
across 95 other items measuring 25 other constructs, they per-
formed very well (M.86). At a respectable .77, the four Worth
Exploring items were least reliable.
Though PI-99 items might appear subject to whim, they were
not. Across three retests, we were surprised—shocked even— by
the stability we observed (see Table 5). For example, the 19-month
test-retest spanned the 2016 U.S.A. election season, the first 125
days of the Trump presidency, the naming of a special counsel on
Russia, and the 2017 Manchester terror attack (the retest came the
next day). Unprompted, several participants noted these events
strongly shaped their responses. Item randomization also meant
sequence effects weakened results. Furthermore, participants were
not ever aware of their Time 1 scores, what particular constructs
were measured, nor how many. Yet, among nearly 400 partici-
Table 5
PI-99 Test-Retest Pearson Correlations Across 2 Weeks, 9
Months, and 19 Months
2 Weeks
9 Months
19 Months
Good .90 .76 .78
Safe .90 .74 .75
Enticing .85 .77 .77
Alive .89 .75 .79
About Me .66 .31 .48
Abundant .84 .66 .67
Acceptable .68 .63 .46
Beautiful .64 .58 .67
Changing .52 .50 .59
Cooperative .74 .59 .66
Funny .67 .64 .70
Harmless .77 .62 .69
Hierarchical .60 .64 .58
Improvable .60 .71 .65
Intentional .86 .80 .82
Interconnected .73 .70 .65
Interesting .70 .70 .62
Just .80 .70 .71
Meaningful .81 .66 .68
Needs Me .78 .68 .74
Pleasurable .86 .71 .71
Progressing .77 .68 .67
Regenerative .72 .62 .61
Stable .63 .55 .59
Understandable .61 .48 .61
Worth Exploring .68 .58 .65
Note. All relationships were significant at p.001.
Irregular payment structure in the 2-week sample and the use of different
versions of the PI in the 9-month sample likely suppressed coefficients.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
pants, 22 short intermixed tertiary scales were stable 19 months
later (M.65). Most notably, Good (.78), Safe (.75), Enticing
(.77), and Alive (.79) scores were approximately as stable as BIG
5 traits (e.g., John et al., 1991) which, just behind IQ, are among
the most stable variables psychologists measure (Conley, 1984).
Across studies, PI-99 subscales showed high convergent, diver-
gent, and discriminant validity. We examined well over 2,000
relationships with other variables (Studies 2 and 5) as well as the
26 26 correlation matrix of PI-99 subscales (325 relationships)
that was largely unchanged across all eight samples. None of these
relationships violated our directional expectations (,, unre-
lated). Despite many small and some moderate relationships be-
tween affect and primals found in Study 1, Studies 4 and 6
suggested ambient emotion unlikely exerts a substantial influence
on PI-99 scores. PI-99 scores were also not driven by social
PI-99 subscales showed incremental validity and predictive util-
ity, even compared to BIG 5 scores. Table 2 provides 35 examples
of moderate and large effects (see Supplement 4 for many others).
For example, controlling for age, sex, education, and income,
Good predicted optimism at ␤⫽.66; Safe predicted neuroticism at
␤⫽⫺.43; Enticing predicted VIA curiosity at ␤⫽.61; Hierar-
chical predicted conservatism at ␤⫽.48; Pleasurable predicted
depression at ␤⫽⫺.52; and Alive predicted meaning in life at ␤⫽
.53. No pairwise relationship with a DV was large enough to
suggest construct redundancy. The strongest was .71 between
Enticing and trait gratitude. When predicting 16 DVs in Study 2,
98% of 48 models that included primals performed better than
demographics-only models; 90% of BIG 5 models were signifi-
cantly improved by adding primals (49% by large margins); and
23% of primals models outperformed BIG 5 models outright.
Good alone and Safe,Enticing, and Alive alone out-predicted the
BIG 5 when it came to trait gratitude, interpersonal trust, growth
mindset, and life satisfaction by average margins 5 times larger
than Raftery’s (1995) convention for strong evidence of a superior
model. If nothing else, the PI-99 can help psychologists make
Though the PI-99 showed excellent psychometrics, it also has
serious limitations. It is not yet validated outside limited online
American samples. Since some primals identified in the qualitative
process were not measured, other primals may exist at the tertiary
level (considered fairly likely) or primary or secondary levels
(considered less likely). A larger sample is needed for the 22-factor
exploratory structural equation model to converge. The 2-week
and 9-month test-retests suffered from methodological irregulari-
ties, including comparing different versions of the PI. Further
studies should more deeply explore the relationship between pri-
mals and affect, employ alternative measurement methods such as
experience sampling, and use postsurvey cognitive interviews to
determine how participants interpret their own PI scores.
Pervasive Influence Is Plausible
Our second conclusion concerns next steps. After decades of
research confirming Beck’s (e.g., 1979) insight that beliefs shape
behavior, Dweck (2008,2017) suggested that as-yet unidentified
beliefs likely shape major personality traits and that “the most
important next step” (2008, p. 394) is to identify these beliefs and
conduct initial correlational research to see if a broader research
phase is justified. This article has introduced a category of beliefs,
identified 26 of them, and found that, in the samples we gathered,
these beliefs met three critical correlational benchmarks. (a) Pri-
mals were stable; people appear to spend years—perhaps
decades— holding the same primals. (b) Primals vary consid-
erably from person to person; PI scores vary on 26 fairly
normal, unimodal distributions, suggesting individuals often
profoundly disagree, perhaps without realizing the extent of
disagreement. (c) Primals are highly predictive, often above and
beyond BIG 5 traits; we found expected patterns of relation-
ships, many large, between primals and over 100 personality,
clinical, wellbeing, political, religious, and demographic vari-
ables. In short, primals behave in the nomological net as if they
play a pervasive role in shaping behavior and wellbeing. Thus,
a broader research phase is justified.
In the meantime, what should we make of widespread dis-
agreement about the nature of a situation we all share (i.e., the
world)? Perhaps primals are an expression of who a person
is— different dispositions lead to different behaviors and be-
liefs. Another option is to extend Beck’s (e.g. 1979) insight that
beliefs shape behavior. Indeed, both primals and disposition are
theoretically capable of explaining considerable variance in
behaviors like neuroticism and curiosity that vary across per-
sons yet are consistent across circumstances (i.e., Kelley’s
(1967) low-consensus, low-distinctiveness, and high-consistency
behaviors). Though not purely exclusive, these options are genuine
alternatives because beliefs differ from disposition in crucial ways.
For example, beliefs are usually more malleable than disposition
(e.g., Dweck, 2008), may influence wellbeing independent of
disposition, may arise for various nondispositional reasons (includ-
ing the situation itself), and may be adopted despite disposition
(e.g., dispositional optimists often see situations as bad). Thus, at
this early stage, agnosticism is prudent. After all, if Jack discov-
ered that Jill profoundly disagrees with him about the nature of a
situation, and Jill’s actions—actions Jack formerly attributed to
Jill’s character—are fully consistent with her perspective, Jack
would be arrogant to assume Jill’s perceptions are mere projec-
tions. Furthermore, discounting the role of beliefs about a situation
may be a predictable and preventable mistake. In the article coin-
ing the term, Ross (1977) famously warns psychologists of com-
mitting the fundamental attribution error by assuming disposi-
tional causes when situational constraints will do— or in this case
perceived constraints.
Whatever future research brings, considering the adage to ‘never
judge someone until you’ve walked a mile in their shoes,’ upon
discovering deep disagreement about the basic character of our
world, perhaps a first step toward better understanding others is to
visit their perceived worlds. The next section briefly describes how
the world may seem to those scoring low and high on the three key
primals: Safe,Enticing, and Alive.
A Brief Tour of Implicit Worlds
Those low on Safe see a Hobbesian world defined by misery,
decay, scarcity, brutality and dangers of all sorts. Base rates for
hazards—from germs to terrorism to getting stabbed in the back—
are generally higher. In response to chronic external threats, they
remain on high alert, often viewing the nonvigilant as irresponsi-
ble. Those high on Safe see a world of cooperation, comfort,
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
stability, and few threats. To them, things are safe until proven
otherwise, vigilance appears neurotic, risk is not that risky, and, in
general, people should calm down.
Those low on Enticing inhabit dull and ugly worlds where
exploration offers low return on investment. They know real
treasure—truly beautiful and fascinating things—is rare and
treasure-hunting appropriate only when it’s a sure bet. Those high
on Enticing inhabit an irresistibly fascinating reality. They know
treasure is around every corner, in every person, under every rock,
and beauty permeates all. Thus, life is a gift, boredom a misin-
formed lifestyle choice, and exploration and appreciation is the
only rational way to live.
Those low on Alive inhabit inanimate, mechanical worlds with-
out awareness or intent. Since the universe never sends messages,
it makes no sense to try to hear any. Those high on Alive sense that
everything happens for a purpose and are thus sensitive to those
purposes. To them, life is a relationship with an active universe
that animates events, works via synchronicity, communicates, and
wants help on important tasks.
Likewise, we expect other primals and combinations of primals
suggest particular life approaches. For convenience, Figure 2 lists
all primals the PI-99 measures, presents the structure we found
across studies, shows where Safe,Enticing, and Alive fit, and
displays their histograms.
Future Directions
With primals reasonably well-measured and their influence
plausible but undemonstrated, many important and fascinating
questions follow. For example, cognitive therapy aims to equip
clients to argue against automatic thoughts such as my colleague is
out to get me (Beck, 1979). However, if primals like the world is
competitive feed such thoughts, it may be more effective to target
the underling primal than the symptomatic belief. For wellbeing
and basic research purposes, we see eight areas as most urgent:
1. Clinical Psychology—Which primals contribute to which
disorders? Can altering primals help? Can cognitive ther-
apies be improved by addressing primals more directly?
2. Personality Psychology—When primals change, does
personality change? Do traits like optimism drive PI-99
scores or does Good drive optimism?
3. Developmental Psychology—Where do primals come
from? When do they emerge? How do they vary over a
lifetime? How do environmental conditions shape them?
4. Social Psychology—How do primals impact familial and
work relationships? Are primals self-sustaining? Are pri-
mals self-fulfilling? Are primals contagious?
5. Positive Psychology—Are certain primals preconditions
for certain character strengths? Can we increase wellbe-
ing aspects like meaning, and trust by altering primals?
6. Organizations—Do some work contexts, like policing,
alter primals over time? Which primals engender success
or failure? Do teams with diverse primals perform better?
7. Politics—To what extent are various political views
grounded in primals? Is political persuasion more likely
when underlying primals are addressed directly?
8. Culture and Groups—Do primals differ across cultures
and explain any group differences? Could understanding
the primals of out-groups facilitate conflict resolution?
Figure 2. Primals’ basic structure with Study 1 (N930) relationships and select histograms. Note five
tertiary primals are largely independent; this is not a strictly hierarchical model.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Primals have been historically understudied. In this article, we
sought to chart all major primals and produce a psychometrically
strong measure. Rather than assuming those who share our planet
share our primals, we can use the Primals Inventory to see the
world from the perspective of others in order to better understand
their actions. Thus far, our use of the PI-99 suggests primals vary
from person to person, are stable, and are highly predictive of
numerous behaviors. One explanation that deserves further inves-
tigation is that, broadly speaking, human action may not express
who we are so much as where we think we are and much of what
we become in life—much joy and suffering—may depend on the
sort of world we think this is.
American Film Institute. (1998). AFI’s 100 greatest American movies of all
time. Retrieved from
American Movie Classics. (2015). All time box office top 100. Retrieved
Antony, M. M., Bieling, P. J., Cox, B. J., Enns, M. W., & Swinson, R. P.
(1998). Psychometric properties of the 42-item and 21-item versions of
the Depression Anxiety Stress Scales in clinical groups and a community
sample. Psychological Assessment, 10, 176 –181.
Baird, F. E., & Kaufmann, W. A. (Eds.). (2000). Philosophic classics:
From Plato to Derrida (4th ed.). Upper Saddle River, NJ: Prentice Hall.
Beck, A. T. (1979). Cognitive therapy and the emotional disorders. New
York, NY: Penguin.
Blei, D. M., Ng, A. Y., & Jordan, M. I. (2003). Latent dirichlet allocation.
Journal of Machine Learning Research, 3, 993–1022.
Butler, J., & Kern, M. L. (2016). The PERMA-Profiler: A brief multidi-
mensional measure of flourishing. International Journal of Wellbeing, 6,
1– 48.
Chen, G., Gully, S. M., & Eden, D. (2001). Validation of a new general
self-efficacy scale. Organizational Research Methods, 4, 62– 83. http://
Chen, S. X., Lam, B. C. P., Wu, W. C. H., Ng, J. C., Buchtel, E. E., Guan,
Y., & Deng, H. (2016). Do people’s world views matter? The why and
how. Journal of Personality and Social Psychology, 110, 743–765.
Conley, J. (1984). The hierarchy of consistency. Personality and Individual
Differences, 5, 11–25.
Cuddy, A. J., Fiske, S. T., & Glick, P. (2008). Warmth and competence as
universal dimensions of social perception: The stereotype content model
and the BIAS map. Advances in Experimental Social Psychology, 40,
Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The
Satisfaction With Life Scale. Journal of Personality Assessment, 49,
Duckitt, J. (2001). A dual-process cognitive-motivational theory of ideol-
ogy and prejudice. Advances in Experimental Social Psychology, 33,
Duckitt, J., Wagner, C., du Plessis, I., & Birum, I. (2002). The psycholog-
ical bases of ideology and prejudice: Testing a dual process model.
Journal of Personality and Social Psychology, 83, 75–93. http://dx.doi
Duckworth, A. L., & Quinn, P. D. (2009). Development and validation of
the short grit scale (grit-s). Journal of Personality Assessment, 91,
166 –174.
Dweck, C. S. (2008). Can personality be changed? The role of beliefs in
personality and change. Current Directions in Psychological Science,
17, 391–394.
Dweck, C. S. (2017). From needs to goals and representations: Foundations
for a unified theory of motivation, personality, and development. Psy-
chological Review, 124, 689 –719.
Dweck, C. S., Chiu, C. Y., & Hong, Y. Y. (1995). Implicit theories and
their role in judgments and reactions: A world from two perspectives.
Psychological Inquiry, 6, 267–285.
Eberhardt, J. L., Goff, P. A., Purdie, V. J., & Davies, P. G. (2004). Seeing
black: Race, crime, and visual processing. Journal of Personality and
Social Psychology, 87, 876 893.
Eysenck, H. J. (1963). Biological basis of personality. Nature, 199, 1031–
Fetzer Institute. (1999). Multidimensional measurement of religiousness/
spirituality for use in health research. Kalamazoo, MI: John E. Fetzer
Floyd, F. J., & Widaman, K. F. (1995). Factor analysis in the development
and refinement of clinical assessment instruments. Psychological As-
sessment, 7, 286 –299.
Friedman, J., Hastie, T., & Tibshirani, R. (2010). Regularization paths for
generalized linear models via coordinate descent. Journal of Statistical
Software, 33, 1–22.
George, L. S., & Park, C. L. (2017). The Multidimensional Existential
Meaning Scale: A tripartite approach to measuring meaning. The Jour-
nal of Positive Psychology, 12, 613– 627.
Goldberg, L. R. (2006). Doing it all bass-ackwards: The development of
hierarchical factor structures from the top down. Journal of Research in
Personality, 40, 347–358.
Hardy, G. (2014). Sacred texts of the world. Retrieved from www
Harvey, P., Madison, K., Martinko, M., Crook, T. R., & Crook, T. A.
(2014). Attribution theory in the organizational sciences: The road
traveled and the path ahead. The Academy of Management Perspectives,
28, 128 –146.
Hicks, L. E. (1970). Some properties of ipsative, normative and forced-
choice normative measures. Psychological Bulletin, 74, 167–184. http://
Hofmann, S. G., Asnaani, A., Vonk, I. J. J., Sawyer, A. T., & Fang, A.
(2012). The efficacy of cognitive behavioral therapy: A review of
meta-analyses. Cognitive Therapy and Research, 36, 427– 440. http://dx
Hood, R. W., Jr., & Morris, R. J. (1983). Toward a theory of death
transcendence. Journal for the Scientific Study of Religion, 22, 353–365.
Hunt, S. M., & Bhopal, R. (2004). Self report in clinical and epidemio-
logical studies with non-English speakers: The challenge of language
and culture. Journal of Epidemiology and Community Health, 58, 618
Li, A. Z., & Sipps, G. J. (1985). Cross-validation of a short form of the
Marlowe-Crowne Social Desirability Scale. Journal of Clinical Psychol-
ogy, 41, 236 –238.
Janoff-Bulman, R. (1989). Assumptive worlds and the stress of traumatic
events: Applications of the schema construct. Social Cognition, 7, 113–
John, O. P., Donahue, E. M., & Kentle, R. L. (1991). The Big Five
Inventory—Versions 4a and 54. Berkeley: University of California,
Institute of Personality and Social Research.
Jonason, P. K., & Webster, G. D. (2010). The dirty dozen: A concise
measure of the dark triad. Psychological Assessment, 22, 420 432.
Kahneman, D., & Frederick, S. (2002). Representativeness revisited: At-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
tribute substitution in intuitive judgment. In T. Gilovich, D. Griffin, &
D. Kahneman (Eds.), Heuristics of Intuitive judgment: Extensions and
applications (pp. 49 81). New York, NY: Cambridge University Press.
Kashdan, T. B., Gallagher, M. W., Silvia, P. J., Winterstein, B. P., Breen,
W. E., Terhar, D., & Steger, M. F. (2009). The Curiosity and Exploration
Inventory-II. Development, factor structure, and psychometrics. Journal
of Research in Personality, 43, 987–998.
Kelley, H. H. (1967). Attribution theory in social psychology. Nebraska
Symposium on Motivation, 15, 192–238.
Kirk, G. S., Raven, J. E., & Schofield, M. (2011). The Presocratic philos-
ophers: A critical history with a selection of texts (2nd ed.). Cambridge,
England: Cambridge Press.
Kishton, J. M., & Widaman, K. F. (1994). Unidimensional versus domain
representative parceling of questionnaire items: An empirical example.
Educational and Psychological Measurement, 54, 757–765. http://dx.doi
Koltko-Rivera, M. E. (2004). Psychology of worldviews. Review of Gen-
eral Psychology, 8, 3–58.
Lerner, M. J. (1980). The belief in a just world. New York, NY: Plenum
Lipkus, I. (1991). The construction and preliminary validation of a global
belief in a just world scale and the exploratory analysis of the
multidimensional belief in a just world scale. Personality and Individual
Differences, 12, 1171–1178.
Lipkusa, I. M., Dalbert, C., & Siegler, I. C. (1996). The importance of
distinguishing the belief in a just world for self versus for others:
Implications for psychological well-being. Personality and Social Psy-
chology Bulletin, 22, 666 677.
Lucas, S. E., & Medhurst, M. J. (2008). Words of a century: The top 100
American speeches, 1900 –1999. Oxford, England: Oxford University
Magill, F. N., & Roth, J. K. (1991). Masterpieces of world philosophy.
New York, NY: Harper Collins.
Mayer, F. S., & Frantz, C. M. (2004). The connectedness to nature scale:
A measure of individuals’ feeling in community with nature. Journal of
Environmental Psychology, 24, 503–515.
McCullough, M. E., Emmons, R. A., & Tsang, J. A. (2002). The grateful
disposition: A conceptual and empirical topography. Journal of Person-
ality and Social Psychology, 82, 112–127.
McNamara, J. M., Green, R. F., & Olsson, O. (2006). Bayes’ theorem and
its applications in animal behaviour. Oikos, 112, 243–251. http://dx.doi
Miller, S. L., Zielaskowski, K., & Plant, E. A. (2012). The basis of shooter
biases: Beyond cultural stereotypes. Personality and Social Psychology
Bulletin, 38, 1358 –1366.