Linking Temperament and Personality Traits from Late Childhood to Adulthood by Examining Continuity, Stability, and Change

Abstract

Theories of personality development emphasize the continuity between who we are as a child and who we are as an adult. The conceptual overlap in influential trait taxonomies designed for children (Rothbart’s temperament model) and adults (the Big Five personality), have reinforced theories about developmental continuity, but key hypotheses remain untested because no studies have linked these trait models longitudinally. To bridge this divide, the present study used longitudinal data from a sample of 674 Mexican-origin youth who completed assessments of Rothbart’s temperament traits (i.e., Negative Emotionality, Surgency, Affiliation, Effortful Control) from ages 10-16 and assessments of Big Five personality traits from ages 14-26. Leveraging two waves of overlapping temperament/personality trait assessments at ages 14 and 16, we found: (1) continuity between childhood/adolescent temperament and age 26 personality, with the strongest associations between conceptually similar traits, and Effortful Control predicting all Big Five traits (except Extraversion), suggesting self-regulation broadly promotes maturation; (2) temperament starts predicting adult personality traits by age 12-14, consistent with theory positing the temperamental foundations of adult personality crystallize in adolescence; (3) conceptually similar temperament/personality traits reflect different expressions of the same underlying trait from age 10-26, established by latent growth models of joint temperament/personality factors; (4) mean-level personality development across late childhood to adulthood showing that all joint traits maintain consistent rank-order stability and youth increase in Effortful Control/Conscientiousness, decrease in Negative Emotionality/Neuroticism and Surgency/Extraversion, and do not change in Affiliation/Agreeableness. Findings add novel support for widely accepted–yet largely untested–theories, although some unexpected results undermine prevailing assumptions about personality trait development.

Does how a child thinks, feels, and acts as they’re growing up foretell the kind of person they become as an adult? In other words, is it true that "The child is father of the man"? Anyone who has had the opportunity to watch a child grow would likely recognize that there is continuity in who they are in adulthood. For example, an energetic child develops into an assertive community organizer or an impulsive child who forgets their homework is due becomes an adult who struggles with getting to work on time and meeting deadlines. Such continuity in individual differences from late childhood into adulthood is a crux of developmental theory. However, taxonomies of individual differences in thoughts, feelings, and behaviors in children and adults have largely been developed and studied in parallel, with traits conceived as temperament in children and personality in adults. The present study addresses the question of whether who someone is as a child bears on who they grow up to be by characterizing relations of temperament and personality traits from late childhood (age 10) into early adulthood (age 26).

Temperamental Origins of Personality Traits: Homotypic and Heterotypic Continuity

Temperament is generally defined as biologically-based individual differences in reactivity and self-regulation that are present early in life and relatively stable over time (Rothbart, 2007). Most developmental theories assume that basic temperament traits present from infancy set the stage for personality in adulthood (Cloninger, 1994; McCrae et al., 2000; Rothbart & Ahadi, 1994). In other words, personality develops from the “building blocks” of temperament through elaboration and differentiation (Caspi & Shiner, 2008; Shiner & Caspi, 2003; Wachs, 1994). Elaboration involves processes that shift genetic predispositions to stable phenotypic patterns of thoughts, feelings, and behaviors via mechanisms like classical and operant conditioning and environmental elicitation, selection, and manipulation. For example, elaboration could look like a child with high sensitivity to noise and extreme fear of strangers who selects into quiet environments and fails to engage with their socially peers, leading them to have a more shy personality in adolescence. Differentiation, also known as the orthogenetic principle (Werner, 1940), involves processes that make traits more complex as people develop new abilities and interests and have greater control over their environmental contexts. For example, differentiation could look like a reward-seeking child who gets involved in team sports, becoming competitive, assertive, and taking on leadership roles among their peers. Through elaboration and differentiation, biological and environmental forces produce consistency and change in how traits are expressed. Stabilizing forces include genetic factors and selecting or eliciting contexts that “fit” one’s individual characteristics (e.g., genetic predispositions toward a more active behavioral inhibition or activation systems; Takahashi et al., 2007), whereas transitional forces include cognitive maturation and behavioral adaptations to meet the demands of normative, age-graded changes in social roles of status (e.g., at work, in peer relationships) and belongingness (e.g., among friends and family, in community; Roberts & Wood, 2006). Aligned with the neo-socioanalytic cumulative continuity principle (Roberts and Nickel, 2021), these developmental processes are expected to result in continuity, including homotypic continuity–traits expressed in similar ways across developmental periods–and heterotypic continuity–traits expressed in different, but developmentally appropriate, ways over time.

Notably, these developmental theories concentrate on traits. This focus stems from the fact that most definitions of temperament do not encompass other, non-trait components of personality, such as motivational constructs and narrative identity (Robins et al., 2025). The conception of traits adopted from a developmental perspective tends to align with trait theory, which views traits as enduring dispositions that influence behavior across the lifespan (Costa & McCrae, 1995). Even from a more neutral position that views traits as descriptive summaries rather than endogenous causes, it is assumed that thoughts, feelings, and behavior associated with a trait will change across development (John, 2021). Accordingly, the present study aims to understand continuity by focusing on temperament and personality traits.

Major Trait Taxonomies of Temperament and Personality

Despite theoretical continuity of individual differences in thoughts, feelings, and behavior across the lifespan, separate taxonomies have been developed to describe traits seen in children (called temperament) and traits in adults (called personality). For children, the most widely accepted framework is arguably Rothbart’s model of temperament (Rothbart, 2000), which includes the traits of Effortful Control, Negative Emotionality, Surgency, and Affiliation. For adults, the prevailing trait model is the Big Five.

Although specific content of the two trait taxonomies differs, there is also clear, conceptual overlap between them (De Pauw & Mervielde, 2010; Shiner & DeYoung, 2013). Effortful Control reflects the capacity to plan, suppress impulses, and focus attention and is akin to Conscientiousness. Negative Emotionality involves tendencies to feel distress and frustration, similar to Neuroticism. Surgency involves variation in pleasure derived from high intensity or novel activities, and is most like Extraversion, Finally, Affiliation reflects a desire for close, warm interpersonal connections, resembling Agreeableness. There is no temperamental analogue of Openness is Rothbart’s model, but some Surgency scale content is related to exploratory tendencies (e.g., “You like exploring new places”) making it arguably the temperament trait most akin to Openness. Non-analogous traits are also expected to show continuity. In particular, it is expected that Effortful Control may contribute to development of all the Big Five, as a superordinate trait thought to regulate expression of individual differences across personality domains (Ahadi & Rothbart, 1994; Denissen et al., 2013; Posner & Rothbart, 2000). From a trait theory perspective, this alignment between trait taxonomies of temperament and personality have fueled the idea temperament and personality are best understood as varying expressions of the same underlying traits (Clark & Watson, 2008; Shiner, 2015; Soto & Tackett, 2015). However, no longitudinal studies to date have comprehensively mapped Rothbart’s temperament traits and the Big Five from childhood through adulthood to empirically test these hypotheses about trait (dis)continuity.

Research on the Development of Temperament and Personality Traits

Several lines of research offer indirect insights into the developmental continuity of Rothbart’s temperament traits and the Big Five. To begin with, there is evidence that even though these taxonomies were designed for different age groups, they each include trait content appearing in children and adults. In studies extending personality “backwards,” it has been shown that the Big Five traits emerge in the structure of parent’s ratings of youth personality using items not even designed to measure the Big Five (John et al., 1994), have been identified in children as early as age 3 (Tackett et al., 2012). Likewise, studies extending temperament “forwards” show traits akin to Effortful Control (also referred to as Constraint in adulthood), Negative Emotionality, and Positive Emotionality emerge in the hierarchical structure of personality in adults (Markon, 2009; Zuckerman et al., 1993). Cross-sectional studies further show convergence between temperament and personality traits in children and adults (Angleitner & Ostendorf, 1994; De Fruyt et al., 2000; De Pauw & Mervielde, 2010; Grist & McCord, 2010; Tellegen & Waller, 2008).

There is also evidence that the Big Five content and structure do not fully capture traits in childhood and adolescence, raising the possibility Rothbart’s model could contain unique trait content covering early manifestations of adult personality. For example, prior work has shown that children’s Big Five traits in elementary school have only small or null associations with those same traits 40 years later (Hampson & Goldberg, 2006). Research comparing the structure of Big Five traits in different age groups generally shows that Big Five traits increasingly coherent (consistent with elaboration) and more distinct from one another (aligned with differentiation) from age 3 to 20 (Soto, 2016; Soto et al., 2008; Tackett et al., 2008, 2012, but see John et al., 1994). Differences in the factor loadings of traits across various age groups reveal how the expression of analogous traits change with age (Soto & John, 2014). For example, factor loadings indicate that Extraversion manifests more physically in middle childhood (e.g., motor activity) and then becomes more psychological in nature by adolescence (e.g., competitive; Soto & John, 2014). The strength of factor loadings also speak to the coherence of traits. Although all Big Five traits are found to increase in coherence from age 10 to 20, coherence in Openness increases the most, suggesting this trait is not fully “formed” until adolescence (Soto et al., 2008). Among the most replicated findings of structural shifts across development that Agreeableness and Conscientiousness are more correlated in childhood and adolescence compared to adulthood (Soto & Tackett, 2015). This change in structure suggests there is a general self-regulation trait akin to Rothbart’s Effortful Control during childhood that is refined into separate domains regulating interpersonal and task-related impulses by adulthood (Soto & Tackett, 2015). These trends of continuity, increasing coherence, and differentiation support the idea that traits in childhood and adolescence, like those in Rothbart’s model, provide the building blocks of adult personality traits to be elaborated, combined, and refined throughout development.

Temperament trait models and measures may these capture expressions of personality in children and adolescents that differ from the Big Five, but which presumably develop into those traits at some point. A few longitudinal studies have linked childhood traits assessed by measures designed to cover child-specific temperament content and adult personality traits like the Big Five (Caspi & Silva, 1995; Laursen et al., 2002; Pesonen et al., 2003; Shiner et al., 2002; Shiner et al., 2017). Overall, this research finds evidence of continuity in analogous and non-analogous traits, with multiple temperament domains in childhood predicting a specific personality trait in adulthood (e.g., Compliance, Self-Control, and low Aggression at age 8 all predict Agreeableness at age 33 with mean r = ∣.27∣; Laursen et al., 2002) or, conversely, a single childhood temperament trait predicting multiple adult personality traits (e.g., Surgent Engagement at age 10 predicting high Positive Emotionality and low Negative Emotionality at age 20 with mean r = ∣.23∣; Shiner, 2002). While this literature reinforces how traits conceptualized as temperament could be precursors to adult personality, our understanding of these processes is limited for two reasons. First, all studies only included two time points, with a long interval between assessments (ranging from 10 to 25 years apart), and none had waves with overlapping measures of childhood temperament and adult personality. Second, because these studies focused only on one or two traits rather than comprehensive taxonomies of temperament and personality, sources of heterotypic continuity may have been missed.

Key Unaddressed Questions about Personality Trait Development from Childhood to Adulthood

Our literature review identified four key questions about personality trait development that the present study aimed to address. We focused on Rothbart’s model of temperament and the Big Five because they are the leading trait taxonomies of traits in children and adults, respectively. Because these questions align with core principles of personality development theories, their empirical examination not only speaks to long-speculated connections between Rothbart’s model of temperament and the Big Five, but this work can also inform our understanding of temperament, personality, and their interrelations writ large.

Research Question 1: Which of Rothbart’s temperament traits from late childhood through adolescence predict the Big Five in adulthood?

Although many conceptual reviews have drawn connections between Rothbart’s temperament model and the Big Five, there have been no studies of their prospective relations necessary to test these hypotheses. In contrast to the few studies noted previously linking subsets of child temperament to adult personality traits, a comprehensive mapping of analogous and non-analogous traits between these broad taxonomies would provide a more complete picture of which temperament traits develop into which personality traits.

Research Question 2: When do Rothbart’s temperament traits first predict the Big Five?

Cross-sectional studies and longitudinal studies with two time points measured many years apart suggest that analogous temperament and personality traits in Rothbart’s model and the Big Five eventually converge (e.g., Effortful Control and Conscientiousness), but they do not provide insight into when temperament traits start to predict later personality or whether the timing of convergence varies across trait domains. Identifying the ages at which child or adolescent temperament begins to resemble adult personality can shed light on developmental processes that alter how individual differences are expressed. For instance, if Surgency first predicts adult Extraversion in adolescence, there might be a basic tendency toward energy and excitement-seeking that becomes directed into assertive behavior when peer relationships and social motives gain importance.

Research Question 3: Are Rothbart’s temperament traits in children and the Big Five in adults different manifestations of the same underlying traits?

Although some tendencies may form distinct trait domains in children but not adults (e.g., Activity; Soto, 2016) and some trait domains manifest quite differently before versus during adolescence (e.g., Openness), an assumption of continuity (be it from a trait theory or descriptive perspective) is that the content in Rothbart’s model and the Big Five reflect age-graded manifestations of the same core traits present throughout life. This hypothesis can be tested statistically, but, as far as we are aware, it never has been. By using multiple waves of temperament assessments in early years and personality assessments in later years, with multiple years of overlapping assessments, it becomes possible to model the changing measures as indicators of the same latent trait. If this model fits the data well, it would support the idea of a single underlying trait persisting from childhood to adulthood, with the behavioral expression evolving over time.

Research Question 4: What is the rank-order stability and mean-level change of joint temperament-personality traits from late childhood to adulthood?

If continuity of Rothbart’s temperament traits and the Big Five can be established at the latent level, it would be possible to estimate developmental trait trajectories spanning the pivotal time from late childhood to adulthood. To date, longitudinal studies of trait stability and change have been of just temperament or personality measures, and only for relatively short segments of the childhood to adulthood transition. Meta-analyses and individual studies with long follow-up periods show that from age 10-20, all traits become more rank-order stable (Bleidorn et al., 2022; Damnian et al., 2019). Additionally, traits related to Agreeableness, Conscientiousness, Extraversion, and Openness show mean-level increases, whereas traits related to Neuroticism show mean-level decreases. Then after age 20, rates of rank-order stability and mean-level trait change slows except for continued increases in Conscientiousness and decreases in Neuroticism (Bleidorn et al., 2022; Roberts et al., 2006; Van Dijk et al., 2020). Together, this literature has been interpreted to support a trend towards personality maturation (Roberts & Nickel, 2021; Specht et al., 2014).¹ Indirect, piecemeal linking of temperament and personality trajectories across studies is all that is possible with the current literature, but it is not ideal, as variation in trajectories could be due to differences in the measures used to assess these constructs. By modeling trajectories of joint factors reflecting temperament measured early in development and the analogous personality trait later on, we can obtain a truer representation of trait stability and change across the entire period from late childhood to adulthood.

The Present Study

Our study explored the temperamental origins of the Big Five and examined how these individual differences develop from late childhood to adulthood. We used data from a longitudinal sample of 674 Mexican-origin youth who completed four assessments of Rothbart’s temperament traits from ages 10 to 16 and six Big Five personality assessments from ages 14 to 26, including two waves of overlapping data at ages 14 and 16. These long-term, densely sampled longitudinal data on the two dominant trait taxonomies provide a unique opportunity to address our four research questions about the developmental dynamics of temperament and personality traits.

Methods

Transparency and Openness

We report how we determined our sample size, all data exclusions, and all measures used in the present study.² Supplementary materials and complete output files are available on the Open Science Framework (OSF) at https://osf.io/e3a2z/. Analyses for this study were pre-registered, but we deviated from the original plan. All deviations, including our rational for each of them, are reported in the supplementary materials. Due to heightened risk of re-identification in this long-term, longitudinal study of Mexican-origin families, data cannot be provided publicly. Researchers interested in reproducing our results can use the complete covariances matrices in the output files. Data pre-processing, construction of figures, and other non-focal analyses were conducted in R Version 4.3.0 (R Core Team, 2024) with the packages ggplot2 (Wickham, 2016) and Mplus Automation (Hallquist & Wiley, 2018). Primary analytic models were estimated using Mplus Version 8.9 (Muthén & Muthén, 2023).

Participants and Procedures

Data for the present study were drawn from the California Families Project, an ongoing longitudinal study of 674 Mexican-origin youth and their parents. Children were selected at random from rosters of students from the Sacramento and Woodland, CA school districts in 2006-2007. To participate in the study, the focal child had to be in the 5th grade, of Mexican origin, and living with their biological mother. Approximately 72.6% of the eligible families agreed to participate in the study, which was granted approval by the University of California, Davis Institutional Review Board (Protocol # 217484-21). The children were assessed annually from age 10 to 19, and then at ages 21, 23, and 26. This study used data from the waves in which temperament and/or personality traits were assessed – ages 10, 12, 14, 16, 19, 21, 23, and 26. All 674 participants were included in the present analyses because each individual had at least one wave of relevant data. Retention rates compared to the full sample at the first assessment are: 86% (age 12), 90% (age 14), 89% (age 16), 87% (age 19), 80% (age 21), 81% (age 23), 75% (age 26).

Participants were interviewed in their homes in Spanish or English, depending on their preference. Most youth were born in the U.S. (71%) and most of their parents (86%) were born in Mexico. The parents had a median level of education of 9^th grade; 37% graduated from high school and 2% had a 4-year college degree. For context, this level of educational attainment is substantially lower, on average, compared to adults over 25 in Sacramento, where 64% had a high school degree (California Demographic Research Unit, 2005). The median total household income ($32,500) is substantially lower than median income (~$52,000) in Sacramento in 2005-2006 (U.S. Census Bureau Small Area Income and Poverty Estimates program).

Measures

Rothbart’s temperament traits

Temperament traits were self-reported at ages 10, 12, 14, and 16 using the short version of the Early Adolescent Temperament Questionnaire – Revised (EATQ-R; Ellis & Rothbart, 2001). The EATQ-R assesses multiple facets of temperament within three trait domains according to Rothbart’s taxonomy. We used items from the domain-level scales of Effortful Control (16 items) and Negative Emotionality (26 items), and facet-level scales of Surgency (15 items) and Affiliation (4 items). The Aggression facet items were omitted from Negative Emotionality for the present study because it is more conceptually similar to Agreeableness than the target personality analogue of Neuroticism. Ratings were made for each item on a 4-point scale ranging from 1 (not at all true of you) to 4 (very true of you).

Big Five personality traits

The Big Five traits were self-reported using the 44-item Big Five Inventory (BFI; John et al., 2008) at ages 14 and 16 and the 60-item Big Five Inventory-2 (BFI-2; Soto & John, 2016) at ages 19, 21, 23, and 26. The BFI and BFI-2 both used a rating scale ranging from 1 (strongly disagree) to 4 (strongly agree). For the BFI, number of items ranged from 8-10 items per trait, and there were 12 items per trait for the BFI-2.

Analytic Plan

All analyses were structural equation models with robust maximum likelihood estimation. Missing data was accommodated by full information maximum likelihood techniques. We modeled temperament and personality by estimating the traits as latent variables using item parcels as indicators as described in previous studies of this sample (Lawson et al., 2023; Ringwald et al., 2024).³ In this prior work, at least partial strong longitudinal measurement invariance was supported for all traits.⁴ We conducted statistical models that corresponded to each of our four research questions.

Research Question 1: Which of Rothbart’s temperament traits from late childhood through adolescence predict the Big Five in adulthood?

First, we established which temperament traits during late childhood/adolescence predicted adult personality traits. The goal of these analyses was to create a complete mapping of Rothbart’s temperament traits to the Big Five, allowing for tests of continuity in analogous and non-analogous traits. We took a “birds eye view” for these analyses by predicting age 26 personality from temperament trait levels averaged across ages 10-16. To maximize precision, and remain consistent with subsequent analyses conducted within a latent growth curve framework, we estimated average temperament and age 26 personality as latent intercepts. Associations between the intercepts indexed how much each temperament trait predicts each adulthood personality trait. Our pre-registered hypotheses were that the strongest associations would be between analogous traits of Affiliation and Agreeableness, Effortful Control and Conscientiousness, Negative Emotionality and Neuroticism, Surgency and Extraversion, and Surgency and Openness.

Research Question 2: When do Rothbart’s temperament traits first predict the Big Five?

After examining prospective associations between temperament and personality at a “birds-eye” level, we next capitalized on the density of assessments to estimate more precisely the age at which temperament traits begin to resemble personality. To accomplish this task, we estimated bivariate correlations between temperament and personality traits across all ages. Our focus was on the convergence of analogous trait pairs. These correlations allowed us to evaluate when temperament first predicts personality and whether the timing of convergence differs across traits.

Research Question 3: Are Rothbart’s temperament traits in children and the Big Five in adults different manifestations of the same underlying traits?

We addressed the extent to which traits in Rothbart’s temperament traits in childhood/adolescence and the Big Five in adults are different manifestations of the same underlying traits by estimating latent growth curve models (LGCMs) of joint personality-temperament factors, focusing again on homotypic trait pairs. Because only two waves had both temperament and personality trait assessments (i.e., ages 14 and 16), we used a modeling approach that combines several methods to accommodate incomplete data (due to the study design) and link measurement scales over time (McArdle et al., 2009). Figure 1 provides a visual depiction of the model. Essentially, these models use information from the waves with complete data to estimate the joint factor at waves with incomplete data. This estimation is done by assuming measurement invariance of the latent trait and estimating missing temperament/personality scores with full information maximum likelihood techniques. Our models were set up with all parcels for a temperament trait and personality trait loading onto a joint factor, and loadings and intercepts for each parcel were fixed to equality across waves. The global fit of these models provides a formal test of whether traits in Rothbart’s taxonomy and the analog Big Five domain reflect the same latent trait across development.

Figure 1. Model used to test if homotypic temperament/personality reflect the same trait from age 10 to 26.

Note. Assumption of measurement invariance in latent joint traits tested by fixing intercepts and factor loadings for the same item parcel indicators to equality across waves. Double-headed arrow = correlation, single headed arrows = regression paths.

Research Question 4: What is the rank-order stability and mean-level change of joint temperament-personality traits from late childhood to adulthood?

Finally, we evaluated the rank-order stability and mean-level change of any joint traits with adequate fitting models. Rank-order stability was modeled by estimating correlations between the latent joint traits at adjacent waves. Mean-level change was assessed using LGCMs. To find the best-fitting growth trajectory, we compared the fit of three LGCMs: (a) no growth model, where only an intercept (no slope) was estimated, (b) linear growth model where age 10 (first timepoint) is fixed at 0 and the slope linearly increases by units representing the number of years since age 10, and (c) a latent basis model where the first and last time points of the slope are fixed (at 0 and 1, respectively) and the middle time points are freely estimated to detect potentially nonlinear changes. Model selection was guided by fit, parsimony, and interpretability. Because the latent basis model is the most flexible of these models and will always fit the data better, we preregistered that we would retain the linear model for any joint trait in which the fit was acceptable.

Given evidence for sex differences in temperament/personality trait development, and differences in the environmental systems (e.g., neighborhood, school, peer group, cultural context) in which youth born in Mexico versus the U.S. grow up, we also tested whether the levels and slopes in the joint trait models varied by sex and nativity. Specifically, we conducted multigroup models of the joint trait LGCMs, and comparing the global fit of the models freely estimated in each group to models with the intercept and slope constrained to be equal across groups. We considered a ΔCFI difference of <.01 between nested models to support measurement invariance (Cheung & Rensvold, 2002), per our pre-registered plan.

For tests of associations (i.e., regressions, correlations), we considered a p-value < .01 to indicate statistical significance per our pre-registration. Global model fit for the joint trait was evaluated holistically. We considered fit to be acceptable if conventional thresholds for at least two alternative fit indices were met (root mean square error of approximation [RMSEA] < .06, comparative fit index [CFI] >.90, and/or standardized root mean residual [SRMR] < .08) and we also considered changes in absolute fit indices (i.e., Bayesian information criterion [BIC] and Akaike information criterion [AIC]).

Results

Research Question 1: Which of Rothbart’s temperament traits from late childhood through adolescence predict the Big Five in adulthood?

Table 1 shows results from models of average late childhood/early adolescence temperament traits predicting personality traits at age 26. Global fit was adequate for all models (see Supplementary Table S2). There was evidence for continuity in analogous and non-analogous traits. Focusing on the strongest temperament predictors of personality (i.e., standardized β ≥ ∣.30∣, the mid-range for related constructs; Lawson & Robins, 2021), we found Affiliation and Effortful Control predicted Agreeableness; Effortful Control predicted Conscientiousness; Negative Emotionality and low Effortful Control predicted Neuroticism; Surgency and low Negative Emotionality predicted Extraversion; and Affiliation and Effortful Control predicted Openness. Consistent with our hypotheses, the strongest associations were between analogous traits, except for Surgency and Openness (r = .14). However, Extraversion had notably low discriminative validity, associating with all temperament traits between β = ∣.20∣ to ∣.30∣.

Table 1.

Associations between temperament traits averaged across age 10-16 predicting age 26 personality traits.

Temperament trait	Personality Trait
	Agreeableness	Conscientiousness	Neuroticism	Extraversion	Openness
Affiliation	.52	.19	−.10	.21	.40
Effortful Control	.50	.42	−.31	.26	.36
Negative Emotionality	−.11	−.24	.40	−.30	−.18
Surgency	.05	.10	−.18	.29	.14

Note. Effects are standardized regression coefficients. Bolded values = p-value < .01; bolded and underlined = associations ≥ .30; grey cells = analogous trait associatio

Research Question 2: When do Rothbart’s temperament traits first predict the Big Five?

Bivariate correlations of analogous temperament-personality trait pairs across all waves are shown in Table 2. Results for heterotypic trait pairs are in Supplementary Table S3. We considered temperament traits at a given age to predict age 26 personality if the correlation was statistically significant (p-value < .01). We found analogous temperament traits first start predicting adult (age 26) personality in early adolescence, with slight differences between traits in timing. Affiliation and Negative Emotionality predicted Agreeableness and Neuroticism, respectively, by age 12, whereas Effortful Control and Surgency predicted Conscientiousness and Extraversion, respectively, by at age 14. Although Surgency averaged across late childhood/early adolescence was only weakly associated with Openness at age 26 (Research Question 1), these correlations clarified that Surgency does predict Openness, but only once participants reached age 16. Of note, concurrent associations between analogous traits at age 14 and 16 were only moderate in some cases (r = .56 for age 14 Negative Emotionality and Neuroticism, rs =.52/.56 for age 14/16 Surgency and Extraversion), reinforcing these taxonomies cover distinct content.

Table 2.

Bivariate correlations between homotypic temperament and personality traits across all waves.

		Agreeableness
	Age	14	16	19	21	23	26
Affiliation	10	.22	.22	.15	.13	.14	.35
	12	.40	.26	.21	.20	.23	.22
	14	.65	.42	.31	.37	.36	.25
	16	.57	.69	.49	.40	.54	.37
		Conscientiousness
	Age	14	16	19	21	23	26
Effortful Control	10	.23	.20	.16	.08	.18	.17
	12	.49	.39	.22	.20	.16	.09
	14	.74	.51	.33	.32	.35	.30
	16	.50	.73	.48	.42	.49	.43
		Neuroticism
	Age	14	16	19	21	23	26
Negative Emotionality	10	.19	.18	.18	.20	.16	.10
	12	.29	.30	.24	.22	.20	.20
	14	.56	.43	.34	.30	.33	.24
	16	.50	.72	.41	.41	.46	.41
		Extraversion
	Age	14	16	19	21	23	26
Surgency	10	.14	.11	.06	.14	.10	.04
	12	.28	.24	.11	.10	.16	.07
	14	.56	.42	.36	.27	.31	.16
	16	.45	.52	.47	.35	.39	.24
		Openness
Surgency	Age	14	16	19	21	23	26
	10	−.01	−.14	−.02	−.05	−.08	−.08
	12	.17	.07	.04	−.02	.06	.00
	14	.42	.26	.22	.08	.23	.12
	16	.31	.40	.32	.17	.31	.19

Note. Cell colors correspond to strength of correlation; grey values are not significant (p-values ≥ .01).

Research Question 3: Are child temperament and adult personality different manifestations of the same underlying traits?

After establishing continuity between child temperament and adult personality analogues in analyses for Research Questions 1 and 2, we next explored whether these trait pairs are in fact different expressions of the same dimension of individual differences from age 10 to 26 rather than distinct constructs. To do this, we fit LGCMs assuming homotypic traits are indicators of a common joint temperament-personality factor present at every wave. If the longitudinal joint trait LGCMs fit the data well, this would indicate analogous traits can be assumed to reflect a unified domain whereas a poor fitting model would indicate they are separate domains.

Table 3 shows the model fit for the joint trait LGCMs. A joint trait model for all homotypic trait pairs except for Surgency/Openness met our criteria for adequate fit. These results support the proposition that Rothbart’s temperament traits and the Big Five reflect different manifestations of the same traits from late childhood onwards.

Table 3.

Global fit for latent growth curve models of joint temperament/personality traits.

	Chi-square	df	RMSEA	SRMR	CFI	BIC	AIC
Affiliation/Agreeableness
No change	983.15	383	.05	.09	.87	23534.42	23028.94
Linear	874.44	379	.04	.09	.89	23443.65	22920.11
Non-linear	822.88	373	.04	.07	.90	23409.82	22859.20
Effortful Control/Conscientiousness
No change	1333.75	509	.05	.10	.89	20035.29	19493.70
Linear	1079.07	506	.04	.09	.92	19784.32	19229.20
Non-linear	1036.02	500	.04	.08	.93	19772.82	19190.61
Negative Emotionality/Neuroticism
No change	2431.08	506	.08	.15	.81	20650.16	20095.04
Linear	2086.06	503	.07	.12	.84	20299.51	19730.84
Non-linear	1435.43	500	.05	.05	.91	19625.24	19043.04
Surgency/Extraversion
No change	1343.48	380	.06	.13	.85	21211.27	20692.25
Linear	1083.46	376	.05	.11	.89	20969.80	20432.73
Non-linear	1025.50	370	.05	.09	.90	20944.42	20380.27
Surgency/Openness
No change	1354.11	379	.06	.11	.83	20445.51	19921.97
Linear	1522.26	382	.07	.11	.80	20602.68	20092.68
Non-linear	1220.91	373	.06	.10	.85	20348.22	19797.60

Note. Bolded rows are the best-fitting models and were considered interpretable. None of the models for Surgency/Openness were considered to be acceptable. df = degrees of freedom; RMSEA = root mean square error of approximation; CFI = comparative fit index; SRMR = standardized root mean residual; BIC = Bayesian information criterion; AIC = Akaike information criterion.

Research Question 4: What is the stability and mean-level change of personality from late childhood to adulthood?

For the joint temperament-personality models that had adequate fit, our final analyses examined the developmental dynamics of these traits from age 10 to 26. Rank-order stabilities are reported in Table 4. All joint traits were moderately stable across the 2–3-year intervals, with median stabilities ranging from .55 (Affiliation/Agreeableness) to .66 (Surgency/Extraversion). Differences in the number of items in the parcel indicators across traits and across waves likely influenced stabilities. Specifically, Affiliation/Agreeableness may have had the lowest stability because it was estimated from the fewest items across waves (5-6 items per parcel). Further, for all traits, stability was descriptively lowest from ages 10 to 12 when they were estimated from the fewest items (due to using the original, shorter BFI version), and highest between ages 14 to 16 when estimated from the most items (due to including temperament and personality trait parcels). We evaluated whether stability varies across development by testing whether constraining the correlations of adjacent waves to be equal significantly reduced fit according to ΔCFI and ΔBIC (reported in Supplementary Table S4). Only Negative Emotionality/Neuroticism showed evidence of worsening fit, but there was no trend of increasing/decreasing stability based on the magnitude of the correlations across waves. Thus, we did not find evidence for increasing personality stability from late childhood to adulthood, despite theoretical expectations and evidence from prior research (Bleidorn et al., 2022).

Table 4.

Rank order stability of joint temperament/personality traits.

Age interval	Rank-order r	Standard Error
Affiliation/Agreeableness
10 to 12	.39	.07
12 to 14	.49	.05
14 to 16	.72	.03
16 to 19	.60	.05
19 to 21	.53	.05
21 to 23	.50	.05
23 to 26	.56	.05
Effortful Control/Conscientiousness
10 to 12	.41	.06
12 to 14	.60	.04
14 to 16	.70	.03
16 to 19	.53	.04
19 to 21	.61	.04
21 to 23	.56	.05
23 to 26	.68	.04
Negative Emotionality/Neuroticism
10 to 12	.34	.04
12 to 14	.59	.04
14 to 16	.80	.02
16 to 19	.50	.04
19 to 21	.61	.04
21 to 23	.70	.03
23 to 26	.63	.04
Surgency/Extraversion
10 to 12	.33	.08
12 to 14	.49	.06
14 to 16	.76	.03
16 to 19	.62	.04
19 to 21	.64	.04
21 to 23	.67	.04
23 to 26	.70	.04

Finally, we examined mean-level change in the joint traits. We determined the best-fitting growth trajectory by comparing fit of models assuming no change, linear change, or nonlinear change (i.e., latent basis). Although the latent basis model had the best fit for Affiliation/Agreeableness, the slope was non-significant suggesting this trait did not increase or decrease overall. Effortful Control/Conscientiousness increased linearly. Negative Emotionality/Neuroticism showed non-linear change, with a steep decline until age 19, followed by a flattening from 19 to 26. Surgency/Extraversion decreased nonlinearly, with declines from age 10 to 19 then relative stability until age 26. These mean-level trajectories and individual differences in the trajectories are plotted in Figure 2. Of note, there were considerable individual differences in these trajectories, both in rate and direction of change. Invariance of levels and slope parameters across sex and nativity was supported for all joint trait models (Supplementary Tables S5 and S6).

Figure 2. Joint temperament/personality trait trajectories.

Note. Bold line is the mean-level growth trajectory, faint lines are individual trajectories.

Discussion

Our study addressed previously unexamined questions about the development of Rothbart’s temperament traits and Big Five personality traits from late childhood to adulthood. We found evidence of continuity in analogous and non-analogous traits, consistent with processes of elaboration and differentiation. Results are consistent with adolescence being a formative time for personality development, with temperament traits beginning to predict adult personality around age 12-14. Despite the unique content of traits in Rothbart’s model and the Big Five, we further showed that analogous traits reflect evolving expressions of a core set of individual differences present from late childhood onward. By linking traits across the long study duration (spanning 17 years), we were able to characterize personality development over the entire critical period from pre-adolescence to early adulthood in the same sample, for the first time in the published literature that we know of. Together, our study adds novel support for widely-accepted–yet untested–theories and calls to question some prevailing assumptions about personality trait development.

Continuity, Stability, and Change in Personality Traits

We found clear evidence for continuity between analogous temperament and personality in the prevailing trait taxonomies for children and adults, respectively. For all traits except Openness, the strongest prospective associations were between analogous temperament traits and adult personality, adding empirical support for what have been largely speculative developmental links between Rothbart’s model and the Big Five. Moreover, we leveraged the density of temperament assessments to better understand the precise ages when temperament traits first foretell a person’s adult personality traits. Broadly, results showed that youth’s temperament begins predicting their adult personality by early adolescence (12-14 years old). Early adolescence is a major transitional period when youth are discovering their identities, acquiring advanced cognitive abilities, adapting to new roles and expectations, and navigating increasingly complex social relationships (Branje, 2022; Dahl, 2004; Larsen & Luna, 2018; Steinberg & Morris, 2001) —and our results reinforce how lasting these changes are. Put another way, patterns established at this relatively young age lay the foundation for who we become as adults. Adolescence plays a key developmental role in forming an identity (e.g., Erikson’s psychosocial stage for adolescence is identity vs. role confusion; Syed & McLean, 2017). Indeed, our findings lend support to the identity development principle from neo-socioanalytic theory (Roberts & Wood, 2006), as the process of developing, committing to, and maintaining an identity that is central to adolescent development (Marcia, 1980) leads to greater personality consistency with age.

Different temperament traits first predicted personality at different ages, suggesting some specificity in the processes influencing continuity. We found that Negative Emotionality predicted adult Neuroticism and Affiliation predicted adult Agreeableness by age 12. The window between 10- and 12-years old marks the peak of puberty in well-nourished populations (Rosenfield et al., 2009), with cascades of neural and biological changes driving increases in sensation-seeking and orientation toward social goals (Sawyer et al., 2018), which may, in turn, form the basis of variation in adult emotional stability and prosociality. The timing for other homotypic traits was later, with Effortful Control predicting adult Conscientiousness and Surgency predicting adult Extraversion by age 14. The period between 12- and 14-years old may be formative for developing these traits because many youth in the United States attend middle school and transition into high school. New academic expectations in middle school demand more responsibility, self-direction, and organizational abilities that may shift Effortful Control to more adult expressions of Conscientious, whereas the increasing salience of social status among peers may drive a shift from broad sensation-seeking to sociability and assertiveness associated with Extraversion. By narrowing down the age ranges that traits in Rothbart’s model begins predicting the Big Five, more specific hypotheses about mechanisms driving continuity in these early and later manifestations of personality can be generated and tested in future work. For example, the present findings suggest that the continuing development of the frontal lobes into early adulthood is not a potent source of discontinuity in the link between early Effortful Control and adult Conscientiousness, highlighting the importance of social processes.

An alternative explanation for differential convergence in temperament and personality traits is that the measures simply vary in the amount of overlapping content they have. The divergent content in Rothbart’s model and the Big Five likely not only reflect developmental differences in the manifestation of personality traits but also differences in the emphasis placed on features of the “same” constructs. To illustrate, Surgency items in the EATQ-R refer more to sensation seeking and activity and less on sociability compared to the BFI Extraversion items—but is this a true representation of how the underlying Extraversion trait manifests in children versus adults or just the product of two research traditions prioritizing different aspects of Extraversion? Further, these substantive and methodological interpretations are not mutually exclusive. Given there is no straightforward way to empirically adjudicate between these explanations, caution is warranted in overinterpreting our results on the timing of trait convergence.

We took a step further in evaluating continuity by showing that Rothbart’s temperament traits and the Big Five are evolving expressions of the same traits present from late childhood onwards. This conclusion was supported by evidence for the longitudinal measurement invariance of joint temperament/personality traits. These joint traits, representing aspects of personality that persist amidst the myriad biological, social, and identity-based changes happening during those 17 years of development, were more stable than meta-analytic estimates of Big Five personality stability at these same ages (Bleidorn et al., 2022). Comparing across waves within this sample shows the relatively high stability is due to the joint traits per se rather than other aspects of the study design, as their stability was highest between ages 14 and 16 when both measures were administered. Statistically, higher rank-order stability in the latent traits can be explained by having more items at these waves, and thus more content coverage and more “true score” variance. More reliable variance means less measurement error, and therefore less attenuation of stability estimates. Conceptually, this result implies that the core traits that cut across these taxonomies of temperament and personality (e.g., Effortful Control/Conscientiousness) capture more stable aspects of who we are across development than do traits from one model or the other (e.g., just Effortful Control or Conscientiousness). These results reinforce that Rothbart’s temperament traits and the Big Five personality traits are indeed more similar than they are different, but are most powerful when combined together.

By modeling traits present from 10 to 26 years old, we were able to chart the course of personality trait development across an age range that has, until now, only been examined in a piecemeal fashion or with very sparse assessments (e.g., two time points). We found mean-level change in the joint traits generally followed trajectories of maturation; that is, youth became more conscientious and emotionally stable, with notably more dramatic changes in Negative Emotionality/Neuroticism than Effortful Control/Conscientiousness. However, they did not become more agreeable, and extraversion levels declined. Although these patterns are generally consistent with the neo-socioanalytic maturity principle (Roberts & Wood, 2006) and prior empirical work, our results extend and enhance the literature in two significant ways. First, estimating change across the entire interval from late childhood to adulthood allowed us to more accurately describe developmental trajectories than extrapolating from studies on segments of this time period. Second, because the joint traits reflect attributes cutting across Rothbart’s temperament model and the Big Five, their trajectories are more generalizable and reflect a broader universe of individual differences than prior work focused on one taxonomy or the other.

Continuity of Dissimilar Temperament and Personality Traits

We also found evidence of continuity in non-analogous traits. Perhaps most notably, we found Effortful Control across late childhood and early adolescence predicted every Big Five personality trait in adulthood. This multifinality of Effortful Control contrasts with prevailing ideas about personality development, and shows that Effortful Control is not isomorphic with Conscientiousness (Inzlicht & Roberts, 2024). Instead, this finding aligns with theories positing Effortful Control is a superordinate domain that regulates the expression of all personality traits (Ahadi & Rothbart, 1994; Posner & Rothbart, 2000). The ability to plan, focus attention, and inhibit impulses may facilitate the social learning necessary to shape temperament into personality traits adapted to the complexities of modern society and the roles and responsibilities of adulthood (Eisenberg et al., 2004). For example, an affectionate child may only become a compassionate and respectful adult if they are able to internalize social norms and values, and channel their affection into broadly prosocial behavior. Effortful Control could also be viewed as having a buffering effect; for example, a child predisposed to emotional reactivity may become a relatively emotionally stable adult if they have sufficient abilities to modulate their responses. All told, our results add unique support for the role of self-regulation processes in personality maturation.

An unexpected finding was the continuity between Affiliation and adult Openness. Although unexpected, this association has precedence. According to some conceptualizations of Openness, the trait entails open-mindedness to other people’s ideas and experiences (McCrae, 1996). Indeed, numerous studies have shown Openness is related to liberal political attitudes and social activism (Brandstätter & Opp, 2014; Carney et al., 2008; Gerber et al., 2011; Lee et al., 2010). Our results add a developmental perspective to the possibly overlooked, interpersonal aspects of Openness in suggesting such justice-driven, action-oriented tendencies emerge from an affiliative disposition in childhood. We caution against overinterpreting this finding though as there are likely other temperamental precursors more consistent with the canonical, intrapsychic aspects of Openness (e.g., intellect, imagination, aesthetic interests), but which are not captured by most temperament scales, including the ones used in the present study. Nonetheless, given the lack of evidence or even theoretical consensus about the origins of Openness (Schwaba & Thalmayer, 2024), these possible roots in Affiliation suggest new avenues of inquiry for future work.

Role of Sociocultural Context on Generalizability

When interpreting these results, it is important to remember that our sample of predominantly low-income Mexican-origin youth grew up in a sociocultural context that is different than most samples in personality development research, which could account for some discrepancies with previous findings. As discussed in our prior work with this sample, such contextual factors could shape the developmental trajectories of temperament (Lawson et al., 2023) and personality traits (Ringwald et al., 2024). Most germane to the aims of the present study, we might expect our sample’s sociocultural context to affect linkages between temperament and personality because of the specific wording of some items on the EATQ-R. For example, the item “You worry about your family when you’re not with them” may represent maladaptively high fear in a sample of white, high-income adolescents, whereas it may be more normative in a sample of Mexican-origin adolescents whose family members are concerned about severe economic hardship, deportation, discrimination, and other forms of systemic racism. This item, therefore, could impact the association between temperamental Negative Emotionality and Big Five Neuroticism in our sample compared to samples of wealthier and/or ethnically majoritized youth. As another example, three of the Surgency items relate to high-cost activities that participants in our sample likely have little to no experience with given the average socioeconomic status (i.e., “Skiing fast down a steep slope sounds scary to you”, “You would not be afraid to try a risky sport, like deep-sea diving”, “You wouldn’t be afraid to try something like mountain climbing.”) These items may capture a different construct in our sample than in a more affluent sample, potentially limiting generalizability of the link between Surgency and Openness we found. Such possibilities underline that we must attend to how systems, structures, and contexts affect our understanding of development, from how personality is expressed to how it is measured (Klimstra & McLean, 2024).

Implications for Theory, Research, and Applied Work

Our findings advance theory, research, and efforts to improve developmental outcomes. On theory, we provide new evidence to bear on the ontology of individual differences. It has long been argued that temperament and personality are simply different names for the same constructs on rational grounds and indirect empirical support (Caspi & Shiner, 2008; Clark & Watson, 2008; Denissen et al., 2013; Shiner & DeYoung, 2013). Here, we directly tested this idea and showed analogous traits from prevailing taxonomies developed for children and adults reflect the same constructs across the years personality is thought to crystallize. Because Rothbart’s temperament model and the Big Five personality traits contain non-overlapping content (e.g., inhibitory control is only in Rothbart’s model, creativity is only in the Big Five), our findings also show how combining these frameworks provides more comprehensive construct coverage and thus a more complete picture of trait development (De Pauw et al., 2009). By dissolving largely arbitrary distinctions between temperament and personality traits, a unified taxonomy can help integrate findings from disparate research traditions with distinct methods towards a more cumulative science of personality development.

On the research front, our study raises questions about existing knowledge of development based on piecing together findings from different studies, each extrapolating trajectories from relatively short intervals (or two waves of data over longer periods), and each using only one measure of personality traits. One discrepancy between our study and prior work is that we found no changes in Affiliation/Agreeableness in contrast to the oft observed increases in Agreeableness during adolescence into early adulthood (Roberts et al., 2006). Perhaps studies extrapolating change from shorter periods of time have distorted the magnitude of change in Agreeableness, and these upticks are actually minimal when considering the longer arc of development. Likewise, several longitudinal studies show temporary dips in personality maturity in early adolescence (Brandes et al., 2021; Denissen et al., 2013; Soto, 2016; Van den Akker et al., 2014), but we did not find this pattern. Although these findings are often explained by the disruption hypothesis (Soto & Tackett, 2015), which proposes the maturity dip results from biological and social transitions happening during that time, our results suggest that methodological explanations may be at play. Namely, in addition to being a possible overextrapolation from trajectories in short time intervals or different samples, the dips in maturity found in prior work may be an artifact of using a single measure. Perhaps decreases in endorsement are due to the rapidly changing meaning of outward behavior and how personality manifests in adolescence, but when these developmentally different expressions are accounted for (as we did in the present study) there really isn’t a deviation from maturation at the latent level. For example, observed decreases in Conscientiousness may be because behavior seen as developmentally normative messiness in a 10-year-old is seen as disorganization at age 14 when adolescents are independently responsible for time management and homework completion for the first time. Finally, we did not find increasing rank-order stability in the joint traits as expected based on meta-analytic evidence (Bleidorn et al., 2022). One possibility for this discrepancy is that the relatively lower stability at younger ages is due to personality measures being age-inappropriate whereas the latent traits in our model capture multiple, age-appropriate individual difference measures. Of course, it is also possible that these divergences are due to idiosyncrasies of our sample or the measures we used. Thus, although our results need to be replicated in other samples, they call into question some widely accepted beliefs about development, and evoke hypotheses about the role of time frame and changing measures to be tested in future work.

Our findings have implications for the timing and targets of efforts to improve developmental outcomes. The need for intervention is evident in our results: within the mean-level trend towards maturity, we found wide individual differences, with some youth not maturing as rapidly or fully as their peers. Moreover, given the high rank-order stability of traits from age 10 onwards, children with less adaptive traits will continue lagging behind their peers into adulthood, even if they mature at the same rate. Our results, aligned with integrative developmental research on the early antecedents of personality (De Fruyt & De Clercq, 2014), suggest that interventions should optimally occur before age 12, when stable patterns of adult personality begin to consolidate. Additionally, the predictive power of Effortful Control on the formation of almost all adult traits suggests that targeting improvements in this trait early on would maximize an intervention’s impact.

Limitations

There are limitations of our study that can inform future research. First, as noted, findings from our sample of Mexican-origin youth growing up in the U.S. may not generalize to other populations because of unique sociocultural experiences that can shape development (Garcia Coll et al., 1996; Marks et al., 2014; Perez-Brena et al., 2018). Thus, although this study adds much needed racial/ethnic and socioeconomic diversity to the personality development literature, results should be replicated in samples of youth from other sociocultural backgrounds and geographic locations. Second, our results may not generalize beyond the specific temperament and personality traits measures we used. In particular, although Rothbart’s model is arguably the most widely accepted, there is considerable variability in temperament models and measures compared to the much greater agreement about the Big Five as a model of adult personality traits. Future research could investigate continuity of individual differences by assessing a broad range of trait measures and models across ages to ensure generalizability across conceptualizations and operationalizations. Third, we focused on personality traits in the present study, but there are many important, non-trait aspects of adult personality (e.g., personal concerns, life goals, motives, narrative identity). Although traits overlap with many of these other levels of personality, our results cannot directly speak to how non-traits are influenced by childhood temperament. An important future direction is to investigate developmental links between non-trait measures of temperament and non-trait aspects of personality, especially during adolescence and young adulthood when motivational and identity processes have heightened salience.

Conclusion

Overall, our findings make it clear that who we are as a child determines, in part, who we become as an adult. Although such throughlines in personality appear obvious when watching others grow up or looking back on our own life story, this study adds empirical support for these processes and new insights into how this development unfolds.