Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Mar 1.
Published in final edited form as: Behav Genet. 2019 Dec 9;50(2):73–83. doi: 10.1007/s10519-019-09983-6

Midlife Study of the Louisville Twins: Connecting Cognitive Development to Biological and Cognitive Aging

Christopher R Beam 1, Eric Turkheimer 2, Deborah Finkel 3,4, Morgan Levine 5, Ebrahim Zandi 1, Thomas M Guterbock 2, Evan J Giangrande 2, Lesa Ryan 6, Natalie Pasquenza 6, Deborah W Davis 6
PMCID: PMC7033012  NIHMSID: NIHMS1546042  PMID: 31820295

Abstract

The Louisville Twin Study (LTS) began in 1958 and became a premier longitudinal twin study of cognitive development. The LTS continuously collected data from twins through 2000 after which the study closed indefinitely due to lack of funding. Now that the majority of the sample is age 40 or older (61.36%, N =1,770), the LTS childhood data can be linked to midlife cognitive functioning, among other physical, biological, social, and psychiatric outcomes. We report results from two pilot studies in anticipation of beginning the midlife phase of the LTS. The first pilot study was a participant tracking study, in which we showed that approximately 90% of the Louisville families randomly sampled (N = 203) for the study could be found. The second pilot study consisted of 40 in-person interviews in which twins completed cognitive, memory, biometric, and functional ability measures. The main purpose of the second study was to correlate midlife measures of cognitive functioning to a measure of biological age, which is an alternative index to chronological age that quantifies age as a function of the breakdown of structural and functional physiological systems, and then to relate both of these measures to twins’ cognitive developmental trajectories. Midlife IQ was uncorrelated with biological age (−.01) while better scores on episodic memory more strongly correlated with lower biological age (−.19 – −.31). As expected, midlife IQ positively correlated with IQ measures collected throughout childhood and adolescence. Additionally, positive linear rates of change in FSIQ scores in childhood significantly correlated with biological age (−.68), physical functioning (.71), and functional ability (−.55), suggesting that cognitive development predicts lower biological age, better physical functioning, and better functional ability. In sum, the Louisville twins can be relocated to investigate whether and how early and midlife cognitive and physical health factors contribute to cognitive aging.

The Louisville Twin Study (LTS) began in 1957 under the directorship of Frank Falkner to study factors contributing to child physical development (Davis et al. 2019). The LTS subsequently was directed by Steven Vandenberg, Ronald Wilson, and Adam Matheny to study a range of developmental processes that mainly encompassed cognitive and personality development from infancy through adolescence (Rhea 2015). Study activity of the Louisville twins was continuous until 2000 when the study closed indefinitely due to lack of funding. Under the current direction of Deborah Winders Davis, childhood data were recovered in collaboration with Eric Turkheimer between 2008 and 2016, a portion of which were published in a special section of Behavior Genetics in 2015 (Beam et al. 2015; Davis et al. 2015; Finkel et al. 2015; Turkheimer et al. 2015). For interested readers, the history of the LTS and its significance recently has been described elsewhere (Davis et al. 2019). Currently, we have repurposed the LTS as a lifespan study of cognition in twins to understand the genetic and environmental factors through which childhood cognitive development correlates with cognitive aging and impairment. While continuing to analyze and publish findings using the recovered data (Giangrande et al. 2019), we have concurrently conducted two pilot studies in preparation for the midlife phase of the Louisville Twin Study.

Childhood Phase of the Louisville Twin Study

The LTS is nationally recognized as one of the oldest, largest, and most comprehensive studies of child development related to multiple birth status (Falkner 1957; Vandenberg et al. 1968; Wilson 1972; Wilson 1978; Wilson 1983; Matheny et al. 1984; Wilson 1986; Wilson and Matheny 1986) and considered to be a "premier study in the field of human developmental behavioral genetics" (Plomin et al. 1988, p. 10). The LTS is unique because of the extensive, longitudinal assessments and because of the relatively low rate of relocation out of Jefferson County, Kentucky.

The study began as a behavior genetic investigation of physical growth (Falkner 1963). Soon after its inception, a longitudinal study of mental development was added to the protocol (Wilson 1978). Twins were tested at 3, 6, 9, 12, 18, 24, 30, and 36 months of age in the LTS research lab, then on an annual basis until age 9 or 10 years. Follow-up assessments were conducted at ages 12 and 15 years. The efforts of various LTS teams effectively recruited twins at birth and ensured representation of the Louisville area population. Even to present day, the demographics of the twins represent the demographics of the greater Louisville area population, with nearly 75% of the sample Caucasian and nearly 20% African American (https://www.census.gov/quickfacts/fact/table/louisvillejeffersoncountybalancekentucky,ky/PST045218).

Childhood assessments of Louisville twins were comprehensive and included measures of cognition, personality, achievement, socioemotional development, and parenting factors. In addition, demographic, cognitive, and personality data were collected from the parents of the twins. The LTS closed in 2000 following a loss of funding and the passing of Dr. Matheny, but an R03 grant in 2014 awarded to Turkheimer funded data-recovery efforts that have converted old paper records and VHS recordings into computerized databases (R03AG048850-01).

Midlife Phase of the Louisville Twin Study

Birth years of the original Louisville twins range from 1950 to 1997, so all possible data from the childhood phase of the LTS have been collected. Of the 1,770 individual twins, 1,086 (61.36%) participated in at least one wave of data collection and will be between the ages of 40 and 64 in 2019. Given the middle-aged nature of the Louisville twins, lifespan questions about cognitive aging can be studied, including: How have differences in twins’ cognitive and social development influenced differences in physical and cognitive aging? To what extent might differences in the twins’ early and attained socioeconomic status influenced successful cognitive aging? What are the genetic and environmental processes underlying the association between cognitive development and cognitive aging? And to what extent is cognitive development causally related to differences in cognitive aging and decline? As twin studies allow for examination of within-family differences, adjusting for all unmeasured genetic and environmental factors that contribute to twins’ similarity, the temporal measurement of childhood, adolescent, and middle-age cognitive, biological, psychosocial, and economic factors permit the most stringent investigation into effects of within-family differences in childhood characteristics (e.g., achievement) on within-family differences in midlife cognitive ability outcomes (e.g., memory functioning). In a lifespan study where random assignment to childhood conditions is unethical if not impossible, within-family longitudinal effects from one developmental epoch to another provide the strongest support for a causal association.

Early life development can influence cognitive aging, particularly cognitive development (Deary et al. 2012a). Lifespan studies can be used to understand how early life factors serve as both risk and protective factors of later life outcomes. With the intractability of preventing or curing Alzheimer’s disease, for example, lifespan approaches are deployed to identify when in the lifespan pathological and cognitive symptoms of decline begin (Livingston et al. 2017; Jack et al. 2018). Cognitive development, early life socioeconomic status, and psychosocial development may have downstream effects on cognitive functioning in midlife and beyond. Yet, these same developmental factors may predict biological, physical, psychosocial, and socioeconomic factors in middle age that more proximally influence cognitive aging and therefore likely have larger effects. Thus, there are two issues related to cognitive aging that we have prioritized in restarting data collection on the middle-aged Louisville twins: 1) specifying causal effects of early-life developmental trajectories on cognitive functioning at midlife, and 2) testing whether and how aging mediates effects of cognitive development on cognitive functioning at midlife.

Given the longstanding focus on child cognitive development in Louisville twins, a natural objective of this midlife study is to model the genetic and environmental processes that connect cognitive development to midlife cognitive functioning. In support of this objective, lower cognitive developmental outcomes in childhood correlate with lower adult levels of cognitive functioning, including memory functioning, executive functioning, and risk of cognitive impairment (Stern 2006; Deary et al. 2012b; Russ et al. 2013). Novel data collection in middle-aged Louisville twins will facilitate exploring genetic and environmental etiologies of cognitive maintenance as people enter the second half of the lifespan. For example, do persons with greater growth in their cognitive developmental trajectories have better cognitive functioning in middle age? When in childhood does cognitive development begin to have predictive utility of midlife cognitive functioning? To what extent does greater growth in cognition across development offset the negative effects of being born into low socio-economic status [SES] families?

Effects of child and adult SES on cognitive aging also are of particular interest to us, as research suggests that heritability of cognitive functioning is sensitive to childhood SES (Turkheimer et al. 2003), at least in the United States (Tucker-Drob and Bates 2016). Childhood SES moderates heritability of full-scale IQ scores in the LTS, both cross-sectionally at age 7 years (Turkheimer et al. 2015) and longitudinally across childhood and adolescence (Giangrande et al. 2019). Analysis of early life socioeconomic advantages may facilitate understanding for how and when actualization of genetic potential for cognitive development in midlife occurs (Bronfenbrenner and Ceci 1994). Questions remain, however, about whether better childhood SES levels continue to serve as a basis for high heritability of cognitive functioning in middle age. We will test such a hypothesis, as only one study has examined whether childhood SES moderates heritability of adult cognitive function (Bates et al. 2013). The study in which these results were observed, the Midlife in the United States Study (Brim et al. 2004), unfortunately relied on retrospective reports of childhood SES. The LTS however, has contemporaneous measurements of childhood SES and cognitive ability across 16 measurement occasions, as well as SES measures for twins’ parents and grandparents. Models that consist of both mean-level family wealth and change in wealth as moderators of adult cognitive functioning will provide better insight into the effects childhood SES has on actualizing cognitive potential, be it via heritable or environmental mechanisms.

A midlife study of the Louisville twins also makes adult twins’ SES available to study as a moderator of the heritability of cognitive ability. To our knowledge, only one twin study, the IGEMS consortium (Zavala et al. 2018), has investigated whether adult SES moderates genetic and environmental variance underlying cognition. Higher attained SES augments heritability of some cognitive domains (e.g., symbol digit) while attenuates others (e.g., block design), findings that were not influenced by effects of rearing SES. Relative differences in attained SES within families are expected to lead to differences in heritability of cognitive ability in adulthood, but this remains an open question that might provide insight into the social and economic causes of cognitive functioning at midlife. As an example, wealthy-born twins who become high-paid physicians while their co-twins become low-paid artists might have the same genetic advantage underlying their cognitive functioning as impoverished-born twins who become modest-earning small business owners while their co-twins live hand-to-mouth as gig economy workers. Relative within-family differences in adult SES, in other words, might be the same irrespective of twins’ childhood SES positioning, leading to the possibility that adult SES may actualize genetic potential underlying cognition irrespective of childhood SES effects.

Correlating cognitive developmental trajectories and other early life factors with cognitive outcomes at midlife is an obvious strength of the LTS, but early life cognition is not the primary predictor of cognitive aging. Rather, age is the greatest risk factor of cognitive aging (Deary et al. 2009; Salthouse 2009). Yet measuring age is not as straightforward as one might think. Age-related biological change, known as biological age, is a more efficient and reliable predictor of age-related health outcomes than chronological age (Levine 2013). Biological age is quantified using clinical outcomes to estimate whether some persons’ biological age is older or younger than their chronological age. Measures of biological age circumvent problems of using chronological age to predict cognitive age and cognitive impairment (e.g., mild cognitive impairment and Alzheimer’s disease) by quantifying age as a function of the breakdown of structural and functional physiological systems (Horvath 2013; Levine 2013; Chen et al. 2016). Biological age tends to be more strongly correlated with global cognitive functioning (Belsky et al. 2015) and cognitive decline (Belsky et al. 2017) than is chronological age. No twin studies of biological age and cognitive functioning, however, have been conducted up to now. Yet, twin studies have the potential to clarify whether pair differences in biological age might predict cognitive functioning at midlife. If twins with greater biological age perform worse on cognitive tests relative to their co-twins, stronger causal conclusions can be drawn about the effects of biological age on cognitive aging; neither shared genetic nor shared environmental factors (whether experienced in childhood or adulthood) can explain the effect. In our pilot study of the LTS (below), we used clinical biological markers to quantify biological age and show that it slightly outperforms chronological age as a predictor of cognitive functioning outcomes.

Childhood cognitive developmental outcomes predict midlife biological age, too (Schaefer et al. 2015). For this reason, our revitalization efforts of the LTS are ideal for exploring whether individual differences in childhood cognitive developmental trajectories predict midlife biological age. The effects of early-life cognitive ability on biological age also have not been studied using genetically informed longitudinal data spanning childhood through middle age. Any effects that cognitive ability in childhood might have on biological age may be attributed to gene-environment correlative processes, that is, the hypothesis that nonrandom exposure to environments (e.g., those supportive of cognitive development) depends on genetically influenced characteristics (e.g., high achieving parents’ genotype) (Scarr and McCartney 1983). Twins with slight initial genetic and environmental advantages for cognitive ability will select into and be reinforced by their social environments over their co-twins, causing pairs to drift apart in ability level with time (Dickens and Flynn 2001). As we have modeled this process for cognitive ability in LTS childhood data (Beam et al. 2015), it is likely that twins set on positive cognitive trajectories will have younger biological ages compared to their co-twins. The dense measurement of cognitive ability in the childhood phase of the LTS not only will clarify critical periods in development that affect age-related biological changes that speed up or slow down aging, but also will adjust for gene-environment correlation that confounds the temporal association between early-life cognition and biological age.

The LTS is a rare resource for studying antecedent factors of cognitive aging and establishing a life-span twin study of biological and cognitive aging. As noted at the outset of this report, we conducted two separate pilot studies in anticipation of new data collection of the Louisville twins at midlife (R01 AG063949-01). The first study took place in 2016 and focused on locating Louisville Twins. The second study occurred in 2018 and was conducted to demonstrate that the Louisville twins could not only be relocated but were willing to participate in additional follow-up testing. With this pilot study data, we will present stability of cognitive functioning from infancy through midlife and will demonstrate that biological age outperforms chronological age as a correlate of cognitive functioning in midlife. Following results of each pilot study, we discuss the broader aims of the midlife study of Louisville twins.

Method

Participants

In 2016, 203 original LTS families were randomly selected from the total population sample for the purpose of establishing the feasibility of relocating Louisville twins. In 2018, we conducted a second pilot study to establish in-person data collection procedures for gathering cognitive, physical, psychological, functional, and clinical biomarker measures from Louisville twins. Over two weeks in July and August 2018, 40 individuals twins (16 complete pairs (11 MZ pairs, 5 DZ pairs) and 8 incomplete pairs) from the total subsample of middle-aged LTS twins age 40 years or older visited the Department of Pediatrics at the University of Louisville to complete in-person cognitive, physical, and psychiatric assessments. Convenience sampling was used to recruit twins into the pilot study. One month prior to initiating data collection, advertisements were posted to the Louisville twins Facebook group inviting twins to participate in the two-week pilot study. Twins who responded and were available to participate in our time-frame were scheduled for in-person testing and blood collection. The eight twins whose co-twins did not participate reflect this limited time-frame, as there was insufficient time and personnel to schedule them. The pilot study sample was capped at 40 individual twins due to the limited time frame. Twins who participated in the in-person assessments had participated in 3-16 childhood data collections (M = 9.13, SD = 4.76). Top rows of Table 1 present basic demographic summaries of the 40 individuals.

Table 1.

Descriptive Statistics of Midlife Measures (N = 40)

Variable N M SD Minimum Maximum
Gender (Male) 40 38% - - -
Chronological Age 40 51.13 5.61 40.00 59.00
Biological Age 40 51.13 5.91 38.95 61.44
Household Income 34 139,831.80 115,354.20 0.00 650,000.00
WAIS-IV FSIQ 40 108.73 13.74 75.00 142.00
Total Immediate Recall (T) 40 53.55 10.83 20.00 78.00
Short Delay Free Recall (z) 40 0.24 1.06 −2.00 2.00
Long Delay Free Recall (z) 40 0.25 1.14 −3.00 2.00
Functional Ability (normal gait speed) 40 3.05 0.42 2.30 4.52
Physical Functioning 39 88.97 14.20 30 100
Open in a new tab

Note. Median household income = $131,000. T score is standardized (M = 50, SD = 10). z scores are standardized (M = 0, SD = 1). WAIS-IV FSIQ is the full-scale IQ score from the Wechsler Adult Intelligence Scale and is standardized (M = 100, SD = 15). Functional ability was measured by gait speed; physical functioning was measured with 10 items from the SF-36 short form health survey.

Measures

Global Cognitive Functioning.

The Wechsler Adult Intelligence Scale-IV (WAIS-IV) was used to assess global cognitive functioning, verbal comprehension, perceptual reasoning, working memory, and processing speed (Wechsler 2008). The WAIS-IV generates a Full Scale Intelligence Quotient (FSIQ), which is an overall estimate of an individual’s current level of global cognitive functioning. The FSIQ scores reported here are age-scaled (M = 100, SD = 15)

Episodic Memory.

We used the California Verbal Learning Test – Second Edition (CVLT-II), a 16-item list-learning task, to assess episodic memory (Delis et al. 2000). The CVLT-II measures learning, free recall (both short- and long-delayed), cued recall (both short- and long-delayed), and recognition memory. The CVLT-II consists of five trials in which participants are required to pay attention and immediately recall as many target words as they can from a 16-word target list composed of words from four different categories. The total number of words immediately recalled across the five trials was converted into a T score (M = 50, SD = 10). Participants then are administered a 16-word distractor list. Immediately following the distractor word list, participants are asked to freely recall as many words as they can from the target list. The raw number of words recalled was converted into gender- and age-scaled short-delay free recall (SDFR) z scores. Approximately 20 minutes after administration of the short-delay free recall test, participants are asked to again recall as many words as they can from the target word list. The raw number of words recalled was converted into gender- and age-scaled long-delay free recall (LDFR) z scores.

Biological Age.

Estimation of biological age consists of seven biomarkers and chronological age. The biomarkers included are: serum albumin, serum alkaline phosphatase, C-reactive protein, serum creatinine, serum total cholesterol, hemoglobin, and systolic blood pressure. These biomarkers were selected based on the recommendations of Levine (2013) for coverage across multiple physiological systems and costs of the clinical biomarkers, given funding constraints of the pilot study. The estimation procedure combines information from seven regression lines of chronological age regressed on seven biomarkers for calculating biological age:

BAEC=∑j=1m(xj−qj)kjsj2+CAsBA2∑j=1m(kjsj)2+1sBA2,

The term xj represents person j’s measured value for biomarker m; kj and qj represent the slope and intercept, respectively, for chronological age regressed on each biomarker m for person j; sj represents the root mean squared error of chronological age regressed on person j’s biomarker m, and CA represents chronological age. s2 BA is the variance estimate of the random biological age variable. Higher scores indicate greater age in years. In Table 2, we report means and standard deviations (2nd column) of each biomarker used in Equation 1 to estimate biological age, as well as individual correlations between each biomarker and FSIQ and short-delay free recall (SDFR) scores.

Table 2.

Means, Standard Deviations, and Correlations with FSIQ and Short-Delay Free Recall (SDFR) of Biomarkers Collected in the Midlife LTS Pilot Study

Measure Mean (SD) Range r (FSIQ) r (SDFR)
Serum albumin (g/dL) 4.34 (0.21) 3.90-4.80 .18 −.03
Serum alkaline phosphatase SI (U/L) 70.83 (23.17) 22.98-35.00 −.06 −.26
C-Reative Protein (mg/L) 3.93 (6.21) 0.20-30.80 −.01 −.11
Serum creatinine (mg/dL) 0.88 (0.16) 0.58-1.44 −.03 −.12
Serum total cholesterol (mg/dL) 197.00 (34.68) 125-284 .21 −.05
Hemoglobin (g/dL) 14.21 (1.52) 10.00-17.70 .24 −.11
Systolic Blood Pressure (mmHg) 128.78 (22.05) 81-179 −.06 .10
Open in a new tab

Notes. FSIQ = full-scale intelligence quotient; SDFR = CVLT-II short-delay free recall. None of the correlations was statistically significant (p < .05).

Functional Ability.

Functional ability was assessed using gait speed, a simple, valid and reliable measure of the ability to walk comfortably. Three trials were administered at “comfortable velocity” for 4 meters. Participants’ elapsed times to cover a fixed distance were averaged across the 3-trial sets. Higher scores indicate slower walking speeds (i.e., lower functional ability).

Physical Functioning.

We assessed physical functioning using the Short Form Health Survey (SF-36; Ware et al. 1993), which is a self-reported set of quality-of-life measures used in the RAND Corporation’s Medical Outcomes Study. Among several domains the SF-36 assesses (general health, physical health problems, emotional health problems, pain, energy level, social functioning, and emotional well-being), 10 items assess severe and minor physical limitations across all physical activities, including general physical activity, walking, climbing stairs, carrying groceries, lifting, bending, and stooping. Item responses include “Yes, limited a lot” (scored 0), “Yes, limited a little” (scored 50), and “No, not limited at all” (scored 100). Mean scores across the 10 items are calculated so that higher scores indicate better physical functioning.

Procedure

For the first pilot study, two separate attempts to locate LTS twins after a 16-year hiatus were conducted: one by LTS staff at the University of Louisville and one by staff at the University of Virginia Center for Survey Research (CSR) in Charlottesville, Virginia. LTS staff used publically available resources (i.e., on-line social media) to locate twins. The CSR staff developed a state-of-the-art tracking protocol (Stone et al. 2014) that uses the proprietary Accurint database (https://www.accurint.com/) and other on-line tools to find current contact information. The protocol develops a tracking database, training tools, and procedures for tracking individuals and rating the quality of matches to other data sources. For each pilot, 100 (UofL) and 103 (CSR) different families were randomly drawn from the total number of LTS families (N = 885 families). Randomly selected families did not overlap between the LTS and CSR groups.

For the second pilot study, in-person data collection was conducted over two weeks in July and August 2018. Each twin was administered the WAIS-IV, the CVLT-II, functional ability assessments (e.g., gait speed), and the SF-36. All twins consented to blood draws, which were processed to extract a set of biomarkers for estimation of Levine’s (2013) biological age. The University of Louisville Institutional Review Board (18.0576) and the University of Southern California Institutional Review Board (HS-18-00860) approved the study.

Results

Pilot 1: Participant Tracking

Success rates differed significantly between the UofL and CSR participant tracking, with the UofL locating 86% families and UVA locating 95% (χ2 = 5.00, df = 1, p < .050). Without directly contacting any individual twins, we were able to update the records for 95% of the twin families and find clear matches for over 90% of the individual twins in the CSR pilot test. Thus, a total 184 (91%) of the 203 families randomly sampled in the relocation study were found through both methods combined.

We conducted logistic regressions to predict which families could be found from relevant demographic variables: sex, race, birth year, number of waves of participation, mother’s age, and SES. In the UofL sample, only birth year (OR = 1.09, p < .050) and number of waves of participation in the childhood phase (OR = 1.19, p < .050) predicted relocation success for families. In the CSR sample, none of the demographic variables predicted relocation success. Children from families found in the pilot participated in significantly more waves of testing (M = 9.67; SD = 5.20) than those who were not found (M = 6.05, SD = 5.90).

Pilot 2: In-Person Cognitive, Physical, and Biomarker Assessment

Table 1 presents the descriptive results of the 40 individual twins who participated in the 2018 pilot study. Thirty-eight percent of the sample was composed of men. Mean chronological age of the sample was 51.13 (SD = 5.61) while mean biological age also was 51.13 (SD = 5.91). The standard deviation of biological age was greater than chronological age, indicating that biological age estimates are broader than chronological age. The mean household income was higher ($139,831.80, SD = 115,354.20) than the median household income ($131,000.00), suggesting that distribution was right skewed, which is attributed to two outliers. Mean FSIQ in the sample was higher than average (M = 108.73, SD = 13.74). Participants’ immediate recall scores were a quarter unit above the test norm on average, as were short- and long-delay free recall. Functional ability scores were low (i.e., faster gait speed), suggesting that participants’ ability was within the normal range. Finally, mean physical functioning was near the maximum score of 100 (M = 88.97, SD = 14.20), suggesting few limitations. Overall, the descriptive results suggest the sample was high functioning, most likely a reflection that the twins who participated in the pilot study functioned at a higher level than the mean person.

Next, we estimated the correlations between midlife FSIQ and all available global cognitive function scores from the childhood phase of the LTS, which includes 16 total measures ranging from 3 months of age to 180 months (15 years) of age. Figure 1 suggests that midlife FSIQ correlates highly with FSIQ measures throughout childhood and adolescence. The mean correlation coefficient was .48 (SD = .17), with a range from .07-.77. The low correlation at 7 years (84 months) of age is .08 and is attributed to three outliers: two with exceptionally high midlife FSIQ scores compared to their 7-year scores (differences of 47 and 48 units) and one with a markedly lower midlife FSIQ score relative to the 7-year score (difference of 31 units). The dotted line in (dotted line in Figure 1) depicts correlations between midlife FSIQ and global cognitive function scores with these three cases removed. As is clear in Figure 1, the correlation between midlife FSIQ and global cognitive function at 7 years was greater (r = .17). The mean correlation coefficient when these cases were removed was .40 (SD = .16), with a range from .14-.74.

Figure 1.

Figure 1.

Open in a new tab

Longitudinal correlations among midlife full-scale IQ score and childhood full-scale IQ scores from age 3 months to 180 months (15 years). Sample sizes (n) varied at each childhood age: n3 months = 11; n6 months = 13; n9 months = 13; n12 months = 13; n18 months = 20; n24 months = 20; n30 months = 16; n36 months = 22; n48 months = 28; n60 months = 30; n72 months = 31; n84 months = 28; n96 months = 40; n108 months = 28; n144 months = 8; n180 months = 40. The solid line (“All”) is based on all available sample sizes at each age of measurement. The dotted line (“3 Outliers Removed”) removes two cases with high midlife FSIQ scores compared to their 7-year scores (differences of 47 and 48 units) and one with a low midlife FSIQ score compared to the 7-year score (difference of 31 units).

Table 3 presents correlations between midlife measures collected in the pilot study and cognitive developmental trajectory parameters (intercept and linear slope of childhood and adolescent FSIQ scores derived from multilevel models). Given the nature of the small sample size, we focus on the magnitude and direction of the correlations only (bolded values indicate results where p < .05). FSIQ positively correlated with short-delay free recall (SDFR) and physical functioning, but coefficients were small (.19, and .20, respectively). As expected, midlife FSIQ positively correlated with the linear slope of childhood FSIQ measurements, albeit not very strongly (.06), while the correlation with childhood intercept was much stronger (.51). FSIQ essentially was uncorrelated with biological age (−.01) whereas measures of episodic memory more strongly correlated with biological age (−.19 – −.31), suggesting that greater biological age correlated with worse performance on episodic memory measures. Correlations between biological age and measures of episodic memory (short-delay free recall = −.31; long-delay free recall = −.19) slightly outperformed chronological age (short-delay free recall = −.28; long-delay free recall = −.18). Biological age positively correlated with functional ability (.23), such that greater biological age corresponded with works functioning; and negatively correlated with physical functioning (−.36), suggesting that greater biological age correlates with greater physical limitations. Lower linear slopes of childhood FSIQ scores correlated with greater biological age, as indicated by the large negative correlation (−.68). Lower FSIQ linear slopes also positively correlated with physical functioning (.73). Both functional ability and physical functioning strongly correlated with linear slopes of childhood FSIQ (−.55 and .71, respectively), suggesting that higher positive slopes correlated with greater functional ability and better physical functioning.

Table 3.

Correlations Among Midlife Measures and Cognitive Developmental Trajectories

FSIQ IR SDFR LDFR BA FA PF Intercept Linear Slope
FSIQ 1 - - - - - - - -
IR .50 1 - - - - - - -
SDFR .19 .54 1 - - - - - -
LDFR .16 .65 .77 1 - - - - -
BA −.01 −.26 −.31 −.19 1 - - - -
FA −.40 −.30 .−.12 −.16 .23 1 - - -
PF .20 .23 .12 .15 −.36 −.62 1 - -
Intercept .51 .28 .13 .14 .29 .13 −.39 1 -
Linear Slope .06 .01 .06 −.03 −.68 −.55 .71 −.65 1
Open in a new tab

Notes. Bolded values indicate statistically significant correlations (p < .05). FSIQ = full-scale IQ; IR = CVLT-II immediate recall; SDFR = short-delay free recall; LDFR = CVLT-II long-delay free recall; BA = biological age; FA = functional ability; PF = physical functioning; intercept = random intercept of child cognitive ability scores from multilevel model for change; linear slope = random slope of child cognitive ability scores from multilevel model for change.

Multi-level regression models were estimated to examine differential effects of biological age on different measures of memory. Although none of the effects was statistically significant, the direction of the effects was in the predicted direction. Greater biological aging scores predicted worse episodic memory: immediate recall (−0.48, SE = 0.59, p = .415); short-delay free recall (−0.12, SE = 0.11, p = .283); and long-delay free recall (−0.07, SE = 0.07, p = .309).

Finally, we report twin correlations for key midlife variables. As there were only 11 complete MZ twin pairs and 5 complete DZ twin pairs, correlations are presented for completeness of this report, but are not interpreted in any serious manner. In general, MZ correlations that are greater than DZ correlations suggest genetic etiology whereas DZ correlations equal to or greater than MZ correlations suggest only environmental etiologies. MZ twin correlations for FSIQ (.92), CVLT-II immediate recall (.65), and functional ability (.44) were greater than DZ twin correlations (.05, .61, .33, respectively). Conversely, DZ twin correlations for CVLT-II short-delay free recall (.70), CVLT-II long-delay free recall (.67), biological age (.97), and physical functioning (.92) were greater than MZ twin correlations (.49, .60, .93, and −.08, respectively).

Discussion

Both LTS pilot studies were designed to reconnect with middle-aged Louisville twins after 16-18 years of study hiatus and establish feasibility of conducting in-person follow-up assessments of all twins in midlife (40-64 years of age). The findings of the first pilot study confirm that the majority of Louisville twins can be relocated without direct contact with the twins. Relocation efforts show that twins can be relocated up to 54 years since the last time they participated in an assessment and that, given the high rate of success in finding the twins, relocated families differ only modestly from the families who were not relocated.

The closest United States twin study comparable to the Louisville Twin Study is the Colorado Adoption/Twin Study of Lifespan Behavioral Development and Cognitive Aging (CATSLife; Reynolds et al. 2017), which has a good success rate of relocating twins (Wadsworth et al. 2019). The LTS, however, differs from CATSLife in two important ways. First, the majority of twins are middle aged (CATSLife was initiated in 1987; Reynolds et al. 2017), which means that the LTS midlife pilot study spans 40-60 years. In this way, the current pilot study constitutes the longest running twin study in the United States to date. Second, the LTS sample size is nearly double that of CATSLife, which implies that power is greater to detect true effects of early life development on midlife outcomes.

Midlife study of the Louisville twins constitutes a true lifespan longitudinal twin study. The main advantage of the LTS is the density of repeated measurements during infancy, childhood, and adolescence. Obviously, firm substantive conclusions cannot be drawn based on the pilot study findings presented above. The 40 twins who participated were enrolled over a two-week period due to limited funding and project time constraints. Their high functioning scores most likely reflect bias in favor of wealthier and healthier twins who could make time to participate within our short time-frame. Twins who participated, for example, traveled from both United States coasts to Louisville on their own account to participate in the pilot study. We will attempt to correct for sampling bias in the full midlife study of Louisville Twins using information previously collected from the twins. Yet, the findings in the current pilot report suggest that cognitive developmental trajectories, lower biological age, higher functional ability, and higher physical functioning correlate in expected directions with midlife FSIQ and episodic memory, consistent with results found in larger studies (Deary et al. 2000; Singh-Manoux et al. 2005; Emery et al. 2012; Kimhy et al. 2013; Karlamangla et al. 2014; Schaefer et al. 2016).

Broader Study Aims of the Midlife Phase of the Louisville Twin Study

Study of middle-aged Louisville twins will play an important role in clarifying whether early life SES advantage confers greater genetic variance underlying cognitive functioning in midlife. Few studies have investigated whether early life SES moderates heritability of cognitive ability in adulthood (Tucker-Drob and Bates 2016). Further, the LTS will clarify whether adult SES, as well as differences between childhood and adult SES moderate heritability of cognitive ability. To date, longitudinal effects of childhood SES on adult cognition have not been addressed in twins. Similarly, studies have yet to address whether improvement in SES from childhood to adulthood might offset negative effects of childhood SES on adult cognitive functioning. We will address these questions in the midlife LTS study to fill these gaps in the literature.

Our study of middle-aged twins in the LTS will forge pathways for investigating the direct and indirect pathways through which genetic factors influence cognitive and biological aging, among other phenotypes of interest; in essence, addressing questions of how rather than what (Anastasi 1958). One of the greatest difficulties of developmental behavior genetics is specifying the indirect pathways through which genetic and environmental factors contribute to phenotypic development (Briley et al. 2018). Genetically informed longitudinal designs are useful for quantifying gene-environment correlation that lead to “sibling drift” – the expectation that genetically related individuals increasingly differ with age (Beam and Turkheimer 2017). Yet, differences in biological and environmental mechanisms might moderate the strength of gene-environment correlation underlying cognitive development over the lifespan (Plomin and Spinath 2004). Further, these mechanisms are many, subtle, and notoriously difficult to quantify because of small effect sizes (Anastasi 1958; Cole 2009). Combined with the existing childhood cognitive data in the LTS, we will measure genotype, DNA methylation, and biological age in middle-aged Louisville twins to expand analysis of the causal chain between genotype and cognitive functioning in middle age.

Ultimate cognitive aging outcomes of interest are severe cognitive impairment, namely Alzheimer’s disease and related dementias. The lifespan structure of the LTS will allow us to make contributions to whether and how aging processes from childhood through midlife might inform development of preclinical risk factors of Alzheimer’s disease. The preclinical phase of dementia is at the forefront of Alzheimer’s disease research, as effective disease prevention probably needs to start in the first half of the lifespan (Jack et al. 2018; Livingston et al. 2017). Significant questions in the cognitive aging and Alzheimer’s disease literature include identifying when in middle age beta-amyloid (Aβ) levels begin to rise (i.e., the principle pathological marker of Alzheimer’s disease), identifying who is at greatest risk, and what early life factors account for increases in Aβ (Jack et al. 2018). Aβ levels increase across the pre-symptomatic phase of ADRD when neuropathology is present without evidence of clinical symptoms (Toledo et al. 2013). Studies, unfortunately, tend to include mixed samples of both middle-aged (40-64) and older (≥ 65 years) participants, which confound preclinical with clinical levels of Aβ. As the midlife LTS consists only of middle-aged twins, our goal is to track twins as Aβ aggregation occurs with age and correlate these levels with cognitive decline and worse memory recall (Cosentino et al. 2010; Clark et al. 2016; Clark et al. 2018; Jansen et al. 2018). As the causal ordering between Aβ accumulation and cognitive decline cannot be established in nonfamilial studies, repurposing the LTS for studying early and midlife effects on within-pair differences in Aβ accumulation will clarify this research gap.

Conclusion

Although originally envisioned as a study for clarifying factors that contribute to healthy child physical development, the Louisville Twin Study evolved into a tour de force longitudinal twin study of cognitive development. Despite the passage of years since the study went dormant, we have demonstrated that over 90 percent the twins who participated can still be located. Further, variability in childhood cognitive development and midlife factors, like biological age, correlate with cognitive functioning at midlife. Thus, the LTS is positioned to be a true lifespan longitudinal twin study of cognition.

Footnotes

Publisher's Disclaimer: This Author Accepted Manuscript is a PDF file of an unedited peer-reviewed manuscript that has been accepted for publication but has not been copyedited or corrected. The official version of record that is published in the journal is kept up to date and so may therefore differ from this version.

References

  1. Anastasi A (1958) Heredity, environment, and the question “how?” Psychol Rev 65:197–208. [DOI] [PubMed] [Google Scholar]
  2. Bates TC, Lewis GJ, Weiss A (2013) Childhood socioeconomic status amplifies genetic effects on adult intelligence. Psychol Sci 24:2111–2116. doi: 10.1177/0956797613488394 [DOI] [PubMed] [Google Scholar]
  3. Beam CR, Turkheimer E (2017) Gene-environment correlation as a source of stability and diversity in development In: Tolan PH, Leventhal BL (eds) Gene-environment transactions in developmental psychopathology: The role in intervention research. Springer, Cham, Switzerland, pp 111–130 [Google Scholar]
  4. Beam CR, Turkheimer E, Dickens WT, Davis DW (2015) Twin differentiation of cognitive ability through phenotype to environment transmission: The Louisville Twin Study. Behav Genet 45:622–634. doi: 10.1007/s10519-015-9756-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Belsky DW, Caspi A, Houts R, et al. (2015) Quantification of biological aging in young adults. Proc Natl Acad Sci 201506264. doi: 10.1073/pnas.1506264112 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Belsky DW, Moffitt TE, Cohen AA, et al. (2017) Eleven telomere, epigenetic clock, and biomarker-composite quantifications of biological aging: Do they measure the same thing? Am J Epidemiol 187:1220–1230. doi: 10.1093/aje/kwx346 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Briley DA, Livengood J, Derringer J, et al. (2018) Interpreting Behavior Genetic Models: Seven Developmental Processes to Understand. Behav Genet 0:0. doi: 10.1007/s10519-018-9939-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Brim OG, Ryff CD, Kessler RC (2004) The MIDUS national survey: An overview. How Heal are we? A Natl study well-being midlife 1–36. [Google Scholar]
  9. Bronfenbrenner U, Ceci SJ (1994) Nature-nurture reconceptualized in developmental perspective: A bioecological model. Psychol Rev 101:568–586. doi: 10.1037/0033-295X.101.4.568 [DOI] [PubMed] [Google Scholar]
  10. Chen BH, Marioni RE, Colicino E, et al. (2016) DNA methylation-based measures of biological age: Meta-analysis predicting time to death. Aging (Albany NY) 8:1844–1865. doi: 10.18632/aging.101020 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Clark LR, Berman SE, Norton D, et al. (2018) Age-accelerated cognitive decline in asymptomatic adults with CSF β-amyloid. Neurology 90:e1306–e1315. doi: 10.1212/WNL.0000000000005291 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Clark LR, Racine AM, Koscik RL, et al. (2016) Beta-amyloid and cognitive decline in late middle age: Findings from the Wisconsin Registry for Alzheimer’s Prevention study. Alzheimer’s Dement 12:805–814. doi: 10.1016/j.jalz.2015.12.009 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Cole SW (2009) Social regulation of human gene expression. Curr Dir Psychol Sci 18:132–137. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Cosentino SA, Stern Y, Sokolov E, et al. (2010) Plasma β-amyloid and cognitive decline. Arch Neurol 67:1485–1490. doi: 10.1001/archneurol.2010.189 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Davis DW, Finkel D, Turkheimer E, Dickens W (2015) Genetic and Environmental Contributions to Behavioral Stability and Change in Children 6–36 months of Age Using Louisville Twin Study Data. Behav Genet 45:610–621. doi: 10.1007/s10519-015-9759-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Davis DW, Turkheimer E, Finkel D, et al. (2019) The Louisville Twin Study: Past, present and future. Twin Res Hum Genet 1–6. doi: 10.1017/thg.2019.37 [DOI] [PubMed] [Google Scholar]
  17. Deary IJ, Corley J, Gow AJ, et al. (2009) Age-associated cognitive decline. Br Med Bull 92:135–152. doi: 10.1093/bmb/ldp033 [DOI] [PubMed] [Google Scholar]
  18. Deary IJ, Whalley LJ, Lemmon H, et al. (2000) The stability of individual differences in mental ability from childhood to old age: Follow-up of the 1932 Scottish mental survey. Intelligence 28:49–55. doi: 10.1016/S0160-2896(99)00031-8 [DOI] [Google Scholar]
  19. Deary IJ, Yang J, Davies G, et al. (2012a) Genetic contributions to stability and change in intelligence from childhood to old age. Nature 482:212–215. doi: 10.1038/nature10781 [DOI] [PubMed] [Google Scholar]
  20. Deary IJ, Yang J, Davies G, et al. (2012b) Genetic contributions to stability and change in intelligence from childhood to old age. Nature 482:212–215. doi: 10.1038/nature10781 [DOI] [PubMed] [Google Scholar]
  21. Delis DC, Kramer JH, Kaplan E, Ober BA (2000) California Verbal Learning Test–Second Edition (CVLT-II). [Google Scholar]
  22. Dickens WT, Flynn JR (2001) Heritability estimates versus large environmental effects: The IQ paradox resolved. Psychol Rev 108:346–369. [DOI] [PubMed] [Google Scholar]
  23. Emery CF, Finkel D, Pedersen NL (2012) Pulmonary function as a cause of cognitive aging. Psychol Sci 23:1024–1032. doi: 10.1177/0956797612439422 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Falkner F (1957) An appraisal ofthe potential contribution of longitudinal twin studies In: The nature and transmission of the genetic and cultural characteristics of human populations. Milbank Memorial Fund, Philadelphia, pp 122–141 [Google Scholar]
  25. Falkner F (1963) An air displacement method of measuring body volume in babies: A preliminary communication. Ann N Y Acad Sci 110:75–79. doi: 10.1111/j.1749-6632.1963.tb17074.x [DOI] [PubMed] [Google Scholar]
  26. Finkel D, Davis DW, Turkheimer E, Dickens WT (2015) Applying Biometric Growth Curve Models to Developmental Synchronies in Cognitive Development: The Louisville Twin Study. Behav Genet 45:600–609. doi: 10.1007/s10519-015-9747-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Giangrande EJ, Beam CR, Carroll S, et al. (2019) Interaction of SES and heritability of cognitive ability across middle childhood and early adolescence. [Google Scholar]
  28. Horvath S (2013) DNA methylation age of human tissues and cell types. Genome Biol 14:1–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Jack CR, Bennett DA, Blennow K, et al. (2018) NIA-AA Research Framework: Toward a biological definition of Alzheimer’s disease. Alzheimer’s Dement 14:535–562. doi: 10.1016/j.jalz.2018.02.018 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Jansen WJ, Ossenkoppele R, Tijms BM, et al. (2018) Association of cerebral amyloid-β Aggregation with cognitive functioning in persons without dementia. JAMA Psychiatry 75:84–95. doi: 10.1001/jamapsychiatry.2017.3391 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Karlamangla AS, Miller-martinez D, Lachman ME, et al. (2014) Biological correlates of adult cognition: Midlife in the United States ( MIDUS ). Neurobiol Aging 35:387–394. doi: 10.1016/j.neurobiolaging.2013.07.028 [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Kimhy D, Crowley OV, McKinley PS, et al. (2013) The association of cardiac vagal control and executive functioning - Findings from the MIDUS study. J Psychiatr Res 47:628–635. doi: 10.1016/j.jpsychires.2013.01.018 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Levine ME (2013) Modeling the rate of senescence: Can estimated biological age predict mortality more accurately than chronological age? Journals Gerontol - Ser A Biol Sci Med Sci 68:667–674. doi: 10.1093/gerona/gls233 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Livingston G, Sommerlad A, Orgeta V, et al. (2017) Dementia prevention, intervention, and care. Lancet 390:2673–2734. doi: 10.1016/S0140-6736(17)31363-6 [DOI] [PubMed] [Google Scholar]
  35. Matheny AP, Wilson RS, Nuss SM (1984) Toddler temperament: Stability across settings and over. Child Dev 55:1200–1211. [Google Scholar]
  36. Plomin R, DeFries JC, Fulker DW (1988) Nature and nurture during infancy and early childhood. Cambridge University Press, New York [Google Scholar]
  37. Plomin R, Spinath FM (2004) Intelligence: genetics, genes, and genomics. J Pers Soc Psychol 86:112–29. doi: 10.1037/0022-3514.86.1.112 [DOI] [PubMed] [Google Scholar]
  38. Reynolds CA, Corley RP, DeFries JC, et al. (2017) The Colorado Adoption/Twin Study of Lifespan Development and Cognitive Aging (CATSLife): An introduction. Behav Genet 47:656–656. doi: 10.1007/s10519-008-9251-y [DOI] [Google Scholar]
  39. Rhea SA (2015) Reviving the Louisville Twin Study: An introduction. Behav Genet 45:597–599. doi: 10.1007/s10519-015-9763-1 [DOI] [PubMed] [Google Scholar]
  40. Russ TC, Stamatakis E, Hamer M, et al. (2013) Socioeconomic status as a risk factor for dementia death: Individual participant meta-analysis of 86 508 men and women from the UK. Br J Psychiatry 203:10–17. doi: 10.1192/bjp.bp.112.119479 [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Salthouse TA (2009) When does age-related cognitive decline begin? Neurobiol Aging 30:507–514. doi: 10.1016/j.neurobiolaging.2008.09.023 [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Scarr S, McCartney K (1983) How people make their own environments: A theory of genotype-environment effects. Child Dev 54:424–435. [DOI] [PubMed] [Google Scholar]
  43. Schaefer JD, Caspi A, Belsky DW, et al. (2015) Early-Life Intelligence Predicts Midlife Biological Age. Journals Gerontol Ser B Psychol Sci Soc Sci 71:1–8. doi: 10.1093/geronb/gbv035 [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Schaefer JD, Caspi A, Belsky DW, et al. (2016) Early-Life Intelligence Predicts Midlife Biological Age. 71:968–977. doi: 10.1093/geronb/gbv035 [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Singh-Manoux A, Hillsdon M, Brunner E, Marmot M (2005) Effects of physical activity on cognitive functioning in middle age: Evidence from the Whitehall II prospective cohort study. Am J Public Health 95:2252–2258. doi: 10.2105/AJPH.2004.055574 [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Stern Y (2006) Cognitive reserve and Alzheimer disease. Alzheimer Dis Assoc Disord 20:112–117. [DOI] [PubMed] [Google Scholar]
  47. Stone C, Scott L, Battle D, Maher P (2014) Locating longitudinal respondents after a 50-year hiatus. J Off Stat 30:311–334. doi: 10.2478/jos-2014-0019 [DOI] [Google Scholar]
  48. Toledo JB, Shaw LM, Trojanowski JQ, et al. (2013) Plasma amyloid beta measurements - a desired but elusive Alzheimer’s disease biomarker. Alzheimers Res Ther 5:8. doi: 10.1186/alzrt162 [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Tucker-Drob EM, Bates TC (2016) Large cross-national differences in gene × socioeconomic status interaction on intelligence. Psychol Sci 27:138–149. doi: 10.1177/0956797615612727 [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Turkheimer E, Beam CE, Davis DW (2015) The Scarr-Rowe interaction in complete seven-year WISC data from the Louisville Twin Study: Preliminary report. Behav Genet 45:635–639. doi: 10.1007/s10519-015-9760-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Turkheimer E, Haley A, Waldron M, et al. (2003) Socioeconomic status modified heritability of IQ in young children. Psychol Sci 14:623–628. doi: 10.1046/j.0956-7976.2003.psci_1475.x [DOI] [PubMed] [Google Scholar]
  52. Vandenberg SG, Stafford RE, Brown AM (1968) The Louisville Twin Study In: Progress in human behavior genetics. The Johns Hopkins Press, Baltimore, MD, pp 154–204 [Google Scholar]
  53. Wadsworth SJ, Corley RP, Munoz E, et al. (2019) A study of lifespan behavioral development and cognitive functioning. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Ware JE, Snow KK, Kosinski M, Gandek B (1993) SF-36 Health Survey Manual and Interpretation Guide. Boston, MA [Google Scholar]
  55. Wechsler D (2008) Wechsler adult intelligence scale–Fourth Edition. 498. [Google Scholar]
  56. Wilson RS (1972) Twins: Early mental development. Science (80-) 175:914–917. [DOI] [PubMed] [Google Scholar]
  57. Wilson RS (1978) Synchronies in mental development: An epigenetic perspective. Science (80-) 202:939–948. [DOI] [PubMed] [Google Scholar]
  58. Wilson RS (1983) The Louisville Twin Study: Developmental synchronies in behavior. Child Dev 54:298–316. [PubMed] [Google Scholar]
  59. Wilson RS (1986) Continuity and change in cognitive ability profile. Behav Genet 16:45–60. [DOI] [PubMed] [Google Scholar]
  60. Wilson RS, Matheny AP (1986) Behavior-Genetic Research in Infant Temperament: The Louisville Twin Study In: Plomin R, Dunn J (eds) The study of temperament: Changes, continuities, and challenges. Lawrence Erlbaum Associates, Hillsdale, NJ, pp 81–97 [Google Scholar]
  61. Zavala C, Beam CR, Finch BK, et al. (2018) Attained SES as a moderator of adult cognitive performance: Testing gene-environment interaction in various cognitive domains. Dev Psychol 54:2356–2370. doi: 10.1037/dev0000576 [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES