Developmental Psychology Copyright 2005 by the American Psychological Association
2005, Vol. 41, No. 3, 506 –516 0012-1649/05/$12.00 DOI: 10.1037/0012-1649.41.3.506
Conscience in Childhood: Old Questions, New Answers
Nazan Aksan and Grazyna Kochanska
University of Iowa
Although conscience has been the focus of reflection for centuries, fundamental questions regarding its
organization have not been fully answered. To address those questions, the authors applied structural
equation modeling techniques to longitudinal data comprising multiple behavioral measures of children’s
conscience, obtained in parallel fashion at 33 and 45 months. The measures encompassed moral emotion
(guilt and empathic distress) and rule-compatible conduct (internalization of maternal prohibitions and
requests and of another adult’s rules). Confirmatory factor analyses supported a differentiated view of
conscience with 2 latent factors at both ages: Moral Emotion and Rule-Compatible Conduct. The
structure of conscience was remarkably stable over time. The coherence between Moral Emotion and
Rule-Compatible Conduct factors increased as children grew older.
The construct of conscience has a long history in many human variation in children’s increasingly complex emotional, behav-
intellectual endeavors, including philosophy, religion, literature, ioral, and cognitive repertoires in morally relevant contexts. This
and the arts. Developmental psychologists have long pondered historical tradition continues in the contemporary work of social-
when and how children come to experience moral emotions and to ization researchers (Eisenberg, 1998; Eisenberg & Fabes, 1998;
behave according to rules and values with no need for external Hoffman, 1970, 1983; Kochanska, 1993, 1995). In some respects,
monitoring (Grusec, 1997; Kagan, in press). As with many devel- one of the goals of an individual differences framework is to
opmental topics (Maccoby, 1984), two prominent approaches to inform developmental psychopathologists concerned with mal-
children’s conscience have emerged over the years. adaptive trajectories. Impoverished moral emotions of guilt (Frick
One approach is rooted in cognitive development and empha- & Ellis, 1999), impaired conduct (Quay, 1988), and deficient
sizes universals in developmental course (Kohlberg, 1969; Piaget, moral cognition (Blair, 1995) have all been linked to future anti-
1932). Researchers who adopt this framework seek to uncover social developmental pathways.
uniform processes and factors that characterize the developmental In this study, we adopt an individual differences perspective to
changes in children’s judgments and reasoning in morally relevant address fundamental questions concerning the structure and sta-
contexts. This historical tradition continues in the work of prom- bility of conscience during the early preschool years. Several
inent contemporary social– cognitive theorists (Nucci & Turiel, emotional and behavioral milestones relevant to children’s con-
1978; Smetana, 1997; Turiel, 1998; Turiel, Killen, & Helwig, science emerge during that period. For example, toddlers and
1997). Though expanded to include observations of children’s young preschoolers acquire the capacity to regulate their conduct
interactions with peers and parents, this approach preserves the and emotion in contextually appropriate ways. Although most of
focus on cognitive representations of moral rules. Acquisition of these skills are acquired and honed during daily parent– child
values in arbitrary conventional versus universal moral domains is interactions in the conventional domain (e.g., “Don’t play with the
among the central questions social– cognitive theorists study. VCR”), some concern the moral domain (e.g., “Don’t hit your
The other approach is rooted in the socialization tradition (Mac- brother”) (Kopp, 1989; Smetana, 1995). Variability in these rudi-
coby, 1984; Sears, Rau, & Alpert, 1965). It emphasizes consis- mentary emotional and behavioral capacities, whether in the con-
tency and stability of observed variation in emotion, behavior, and ventional or moral domain, is both commonly found and critical
cognition across contexts and time as well as variation in chil- for future adjustment in morally relevant contexts (Frick et al.,
dren’s individual trajectories. Researchers who adopt this frame- 2003).
work seek to understand the processes and factors that account for There is a growing consensus that those emerging components
of conscience need to be studied more comprehensively, integrat-
ing emotion, conduct, and cognition (Blasi, 1980; Burton, 1984;
Nazan Aksan and Grazyna Kochanska, Department of Psychology, Dienstbier, Hillman, Lehnhoff, Hillman, & Valkenaar, 1975;
University of Iowa. Grusec & Goodnow, 1994; Grusec & Kuczynski, 1997; Hoffman,
This research was supported by National Science Foundation Grant 1970, 1983; Kagan, in press; Radke-Yarrow, Zahn-Waxler, &
SBR-9510863, National Institute of Mental Health (NIMH) Grant RO1 Chapman, 1983; Thompson, 1998; Turiel, 1998). However, re-
MH63096, and NIMH Independent Scientist Award KO2 MH01446 to searchers’ understanding of the extent of coherence among early
Grazyna Kochanska. We appreciate the excellent contributions of multiple
manifestations of conscience remains limited.
students and staff who conducted the sessions and coded the data. We
thank the families in the parent– child study for their outstanding and
Two classic studies were the first to address questions concern-
enduring commitment to this research. ing the organization of conscience by examining coherence among
Correspondence concerning this article should be addressed to Nazan its components (Hartshorne & May, 1928 –1930; Sears et al.,
Aksan, Department of Psychology, University of Iowa, Seashore Hall, 1965). They asked the following: Are different manifestations of
Iowa City, IA 52242. E-mail: nazan-aksan@uiowa.edu conscience, or character, interrelated? Is there an inner core of
506
, CONSCIENCE IN CHILDHOOD 507
one’s character that is relatively stable and that accounts for one’s exception of maternally reported temperament (Pedlow, San-
relatively coherent behavior and affect in morally relevant con- son, Prior, & Oberklaid, 1993), measurement equivalence has
texts? Hartshorne and May (1928 –1930) concluded that various very rarely been examined in observational studies with young
behaviors and affects were basically unrelated and thus unpredict- children. The second goal of this study was to examine struc-
able from one context to another in a large group of school-age tural stability in young preschoolers’ conscience.
children. Sears et al. (1965) attempted to integrate the psychoan-
alytic focus on emotions of guilt with social-learning focus on
moral conduct in a small group of preschool-aged children, ob- The Current Study
served in the context of transgressions. They concluded that chil-
We examined the latent structure of conscience and its stability
dren’s emotion and behavior were only modestly related.
in a group of preschoolers at 33 and 45 months, using observa-
Two consecutive investigations have revisited Hartshorne and
tional measures. At 33 and 45 months, we conducted two labora-
May’s (1928 –1930) original data to address questions of the
tory sessions that encompassed naturalistic yet carefully scripted
structure of conscience (Burton, 1963; Rushton, Brainerd, & Press-
paradigms. At both ages, we assessed the children’s guilt follow-
ley, 1983). Burton (1963) used principal-components analysis, and
Rushton et al. (1983) applied rules of aggregation. Findings sug- ing transgressions and their empathy to an adult who simulated
gested that the early claims of the relative lack of coherence among distress. In addition, we assessed children’s rule-compatible con-
different aspects of character had been overstated and that in duct in the absence of surveillance, including internalization of
similar contexts, individuals did show a modest degree of consis- maternal prohibitions, maternal requests, and another adult’s rules.
tency. In our own research, we have also often found that moral Theoretically driven questions pertinent to the factor structure of
emotion and conduct are interrelated (Kochanska, Padavich, & fallible, observed measures and stability in factor structure over
Koenig, 1996). Several other studies have yielded variable results, time are best addressed with structural equations modeling (SEM)
from absent or low consistency to substantial (for reviews, see techniques (Bollen, 1989; Marsh & Grayson, 1994). Unlike the
Eisenberg & Fabes, 1998; Grusec & Lytton, 1988; Radke-Yarrow traditional correlational strategies that rely heavily on arbitrary
et al., 1983). The first goal of this study was to examine the factor criteria, SEM techniques can formally test the viability of different
structure of preschoolers’ emotionality and behavior relevant to models of conscience organization as well as its stability over time.
conscience. We applied modern SEM techniques to test two alternative
Another classic question of individual differences concerns the views of conscience organization and structure. The simplest
stability of young children’s conscience. Several empirical studies, model is consistent with the psychoanalytic tradition in that it
including our own, suggest significant stability for the specific advocates a monolithic internal structure like superego that orga-
components of conscience (Grusec & Lytton, 1988; Kochanska, nizes emotion and conduct. We refer to this model as the uniform
Aksan, & Nichols, 2003; Kochanska, Coy, & Murray, 2001; model. The alternative model is consistent with the social-learning
Kochanska, Gross, Lin, & Nichols, 2002). Moreover, research on tradition in that it advocates contextual specificity (Mischel, 1968)
origins of conduct problems has convincingly shown that deficient or the distinctiveness of emotion and conduct. We refer to this
conscience in childhood— callousness, impaired moral emotions, model as the differentiated model. Those alternative views are
and disrupted internalization of rules of behavior—launches a depicted in the panels of Figure 1. Figure 1A depicts the uniform
clear trajectory leading to antisocial personality well into adoles- model, and Figure 1B depicts the differentiated model of con-
cence and adulthood (Frick et al., 2003; Frick & Ellis, 1999; science. Those two models were tested separately at 33 and 45
Lykken, 1995; Shaw & Winslow, 1997). Some prosocial disposi- months.
tions also have been found to be relatively stable from the pre- We also applied SEM techniques to examine the extent of
school age to young adulthood (Eisenberg et al., 1999, 2003). stability in the factor structure of conscience from 33 to 45 months.
Collectively, those studies speak to rank-order stability, on a When the configuration of various components of conscience is
variable-by-variable basis, either in specific components of con- similar over time, we can proceed to even more interesting devel-
science that are measured similarly over short periods of time (e.g., opmental questions. For example, we can ask whether the relations
Kochanska et al., 2001) or in functional continuity of early differ- of various component scales to their respective underlying latent
ences over long periods of time (e.g., Frick et al., 2003). As Caspi construct remain invariant over time. Longitudinal confirmatory
(1998) has argued, however, stability in individual differences can factor analyses are optimally suited to address this question of
be studied at the level of latent structures. equality in factor loadings (Caspi, 1998). Equality in factor load-
Structural stability refers to the degree of invariance that ings over time increases our confidence that the purported latent
governs the relation between the observable, fallible measures construct(s) retains its meaning over time (Kaplan, 2000; Marsh &
and the unobservable latent constructs over time. For example, Grayson, 1994; Meredith, 1964). When the hypothesis of equality
scores on various facets of personality inventories typically in factor loadings cannot be rejected, we can further proceed to ask
reflect the familiar latent Big Five factors from late adolescence questions of invariance in unique variances of component scales.
into adulthood (Costa & McCrae, 1992). Stability in factor This strategy essentially involves conducting graded tests of sta-
structure, sometimes referred to as measurement equivalence bility in factor structure, often referred to as metric invariance. In
(Baltes, Reese, & Nesselroade, 1977), is routinely incorporated summary, we addressed two questions. First, we examined the
into research on adolescent and adult personality as well as organization of two aspects of conscience, emotion and conduct,
other aging research (e.g., Byrne, Shavelson, & Muthen, 1989; by testing its factor structure at 33 and 45 months. Second, we
Hertzog & Schaie, 1986; Marsh & Grayson, 1994). With the examined the stability of that factor structure over time.
