View in EDS HTML Full Text PDF Full Text

The Breadth of Impacts from the Abecedarian Project Early Intervention on Cognitive Skills

Saved in:
Bibliographic Details
Title: The Breadth of Impacts from the Abecedarian Project Early Intervention on Cognitive Skills
Language: English
Authors: Pages, Remy (ORCID 0000-0002-0396-6498), Protzko, John, Bailey, Drew H.
Source: Journal of Research on Educational Effectiveness. 2022 15(2):243-262.
Availability: Routledge. Available from: Taylor & Francis, Ltd. 530 Walnut Street Suite 850, Philadelphia, PA 19106. Tel: 800-354-1420; Tel: 215-625-8900; Fax: 215-207-0050; Web site: http://www.tandf.co.uk/journals
Peer Reviewed: Y
Page Count: 20
Publication Date: 2022
Document Type: Journal Articles
Reports - Research
Education Level: Early Childhood Education
Descriptors: Early Intervention, Cognitive Ability, Intelligence Quotient, Children, Adolescents, Young Adults, Early Childhood Education, Disadvantaged, Program Effectiveness, Thinking Skills
Geographic Terms: North Carolina
DOI: 10.1080/19345747.2021.1969711
ISSN: 1934-5747
1934-5739
Abstract: Early life interventions impacting cognitive abilities are most often followed by post-treatment fadeout. Some have hypothesized that persistence is unlikely when gains are specific to trained skills and distinguishable from impacts on general cognitive ability (classically modeled as a hierarchical factor, so-called psychometric g). Using measurement invariance testing and multiple-indicators multiple-causes models, we investigated impacts on IQ subtests from the Abecedarian early childhood intervention (n = 107). We found that (1) observed impacts on IQ scores from age 5 to age 21 were consistent with persistent positive effects on g; (2) subtest-specific variance that was differentiable from changes on g did fade. Together, these findings indicated that Abecedarian early impact persisted across a range of cognitive skills, providing some evidence for the hypothesis that breadth and persistence of impacts from educational interventions are related.
Abstractor: As Provided
Entry Date: 2022
Accession Number: EJ1349873
Database: ERIC
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
    Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwHEpgvpFIbVA6ECd3ME71KsAAAA4jCB3wYJKoZIhvcNAQcGoIHRMIHOAgEAMIHIBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDNeyIqmvbjzOyeHYQwIBEICBmgooSi1k-NUl_Rk_MWJcntKZwuMiVjO4m0kJToW52N0hq6TT7fVSWxgBMQY2nPLSKfdoWryDAPeAohLu0FU823x6HKF_eZ-rK8BSROH5-txrMHo0ZjWXaUGexmMPrBeoAhQNFBrboqqa8onHjH17AjKweOhejaRdydXTW9529F4hbFxw3cQXI5nsu1BnP3nHLyVYsdfNW8_eUNc=
Text:
  Availability: 1
  Value: <anid>AN0156866495;[5ew9]01apr.22;2022May17.07:46;v2.2.500</anid> <title id="AN0156866495-1">The Breadth of Impacts from the Abecedarian Project Early Intervention on Cognitive Skills </title> <p>Early life interventions impacting cognitive abilities are most often followed by post-treatment fadeout. Some have hypothesized that persistence is unlikely when gains are specific to trained skills and distinguishable from impacts on general cognitive ability (classically modeled as a hierarchical factor, so-called psychometric g). Using measurement invariance testing and multiple-indicators multiple-causes models, we investigated impacts on IQ subtests from the Abecedarian early childhood intervention (n = 107). We found that (<reflink idref="bib1" id="ref1">1</reflink>) observed impacts on IQ scores from age 5 to age 21 were consistent with persistent positive effects on g; (<reflink idref="bib2" id="ref2">2</reflink>) subtest-specific variance that was differentiable from changes on g did fade. Together, these findings indicated that Abecedarian early impact persisted across a range of cognitive skills, providing some evidence for the hypothesis that breadth and persistence of impacts from educational interventions are related.</p> <p>Keywords: Early childhood intervention; RCT; cognitive skills; measurement invariance; long-term impact</p> <p>Early childhood education (ECE) programs like Perry Preschool, Abecedarian, larger-scale programs like Head Start, and others have yielded heterogeneous effects across programs on a panoply of long-run outcomes (e.g., Duncan & Magnuson, [<reflink idref="bib24" id="ref3">24</reflink>]; Elango et al., [<reflink idref="bib25" id="ref4">25</reflink>]; McCoy et al., [<reflink idref="bib61" id="ref5">61</reflink>]; Pages et al., [<reflink idref="bib69" id="ref6">69</reflink>]; Watts et al., [<reflink idref="bib96" id="ref7">96</reflink>]). The nature of these impacts, as well as the mechanisms of their impacts' persistence are important topics of study for investing in the most effective interventions and for understanding mechanisms of child development (e.g., Abenavoli, [<reflink idref="bib1" id="ref8">1</reflink>]; Bailey et al., [<reflink idref="bib3" id="ref9">3</reflink>]; Cunha et al., [<reflink idref="bib19" id="ref10">19</reflink>]; Duncan & Magnuson, [<reflink idref="bib23" id="ref11">23</reflink>]; Heckman, [<reflink idref="bib39" id="ref12">39</reflink>]; Protzko, [<reflink idref="bib72" id="ref13">72</reflink>]).</p> <p>The most important puzzle is why some earlier ECE programs, such as Perry and Head Start, explicitly targeting children's early cognitive skills, generate impacts that fade out while still yielding positive rates of returns on consequential long-run outcomes (e.g., Deming, [<reflink idref="bib20" id="ref14">20</reflink>]; Heckman et al., [<reflink idref="bib40" id="ref15">40</reflink>]; Yoshikawa et al., [<reflink idref="bib103" id="ref16">103</reflink>]). Several hypotheses have been proposed to explain this, in particular pointing to possible early effects mediation via so-called noncognitive skills (e.g., Chetty et al., [<reflink idref="bib15" id="ref17">15</reflink>]; Heckman et al., [<reflink idref="bib41" id="ref18">41</reflink>]). Another important puzzle which has received less attention is why for some ECE interventions, impacts on cognitive test scores seem to persist at higher rates than for others. Some explanations hypothesize that children in the control group "catch up," because later educational experiences influence the same cognitive skills as earlier educational experiences; or that those in the experimental group lose what had been gained (e.g., Protzko, [<reflink idref="bib71" id="ref19">71</reflink>]; for review, see Bailey et al., [<reflink idref="bib2" id="ref20">2</reflink>]). However, another common explanation for the fadeout effect is that early interventions fail to impact a key set of attributes producing positive covariance among standardized cognitive test scores (e.g., Jensen, [<reflink idref="bib46" id="ref21">46</reflink>]; Te Nijenhuis et al., [<reflink idref="bib92" id="ref22">92</reflink>]). Protzko ([<reflink idref="bib72" id="ref23">72</reflink>]) framed this argument as the 'raising IQ/raising <emph>g</emph> distinction': fadeout ensues because general intelligence operationalized as psychometric <emph>g</emph> is unaffected by early interventions that seek to change children's cognitive performance (Jensen, [<reflink idref="bib48" id="ref24">48</reflink>], [<reflink idref="bib47" id="ref25">47</reflink>]; Te Nijenhuis et al., [<reflink idref="bib92" id="ref26">92</reflink>]). In contrast, if a cognitive performance advantage were to persist, there should be persistent impacts across IQ subtests proportional to their <emph>g</emph> loadings, as would be predicted if early interventions had raised <emph>g</emph> (i.e., a raising IQ/raising <emph>g</emph> distinction contrapositive).</p> <p>Intelligence researchers have not reached consensus on the ontological and theoretical status of <emph>g</emph> (for recent overviews on this debate, see Kovacs & Conway, [<reflink idref="bib54" id="ref27">54</reflink>]; along with corresponding commentaries). Some have proposed psychometric <emph>g</emph> as an artifact of overlapping cognitive process sheaves—not as a causal factor producing variation in cognitive test scores but as a formative index stemming from variation in a variety of cognitive processes (Kovacs & Conway, [<reflink idref="bib54" id="ref28">54</reflink>]). By contrast, others have theorized that the empirical positive intercorrelations of test scores on cognitive tasks directly resulted from a network of <emph>mutually</emph> causal factors in development. In this case, higher-order <emph>g</emph> is replaced by an immanent system of multivariate dynamic interactions (Van Der Maas et al., [<reflink idref="bib94" id="ref29">94</reflink>]). These theories depart from more classical reflective second-order modeling of <emph>g</emph> (e.g., Carroll, [<reflink idref="bib14" id="ref30">14</reflink>]; Gustafsson, [<reflink idref="bib37" id="ref31">37</reflink>]; Johnson & Bouchard, [<reflink idref="bib49" id="ref32">49</reflink>]; Vernon, [<reflink idref="bib95" id="ref33">95</reflink>]). Yet, attributing explanatory power to lower level processes that either work together during cognitively complex tasks or are associated via mutually beneficial developmental processes can nonetheless accommodate the raising IQ/raising <emph>g</emph> distinction. And regardless of <emph>g</emph>'s ontological status, if <emph>g</emph> is left unaffected by intervention, what then could explain fleeting IQ impacts?</p> <p>One possibility, which clearly applies to some cognitive gains, is a test-retest effect: taking a given test twice often leads to a higher score on the second testing occasion, with limited or no transfer to other tests (Jensen, [<reflink idref="bib46" id="ref34">46</reflink>]). Across 64 studies in which participants took the same test twice, observed IQ score gains on retest were not attributable to changes in <emph>g</emph> (Te Nijenhuis et al., [<reflink idref="bib93" id="ref35">93</reflink>]). However, simple test-retest effects are unlikely to explain treatment effects in randomized interventions, wherein the treatment and control groups are usually pre- and post-tested the same number of times.</p> <p>A second kind of non-<emph>g</emph> effect likely to result in fadeout is teaching-to-the-test (discussed in Protzko, [<reflink idref="bib71" id="ref36">71</reflink>]; and Jensen, [<reflink idref="bib46" id="ref37">46</reflink>]). This label refers to effects induced by teaching or coaching on cognitive tasks similar in content to IQ tests (for common examples, see Protzko, [<reflink idref="bib73" id="ref38">73</reflink>]). For instance, a preschool setting might potentially intensify both the informational content and might produce familiarity with communicative situations similar to those in which cognitive tests are administered. Consider, for example, a classroom situation where teacher and children are looking at a picture book depicting various animals and a question comes up: "Which of these animals bark?" In such cases, the teacher engages in games with children that closely resemble testing settings and materials, because they are deemed convenient for educational activities with young children.</p> <p>Being situation-bound, such learning would be less likely to transfer to other contexts than would more generic skills or propensities (e.g., consistent interest in or access to books). In all such situations, IQ test score gains would correspond to changes on specific cognitive test, and a higher-order <emph>g</emph> factor would not capture group differences (Estrada et al., [<reflink idref="bib27" id="ref39">27</reflink>]). Consequently, per the raising IQ/raising <emph>g</emph> distinction, fadeout might be expected, because the specific skills (knowing animal sounds or being comfortable interacting with an unrelated adult) may be quickly learned by children in the control group as they transition to formal educational settings. Further detailed in the method section, measurement invariance testing within randomized control trials can in principle permit to detect such teaching-to-the-test effects (see also Bailey et al., [<reflink idref="bib2" id="ref40">2</reflink>]; and Protzko, [<reflink idref="bib72" id="ref41">72</reflink>]). However, to our knowledge, the usefulness of such procedure for informing theories of fadeout and persistence has not yet been convincingly demonstrated.</p> <p>Two kinds of evidence suggest that the raising IQ/raising <emph>g</emph> distinction does not perfectly predict which environmental impacts on cognitive test performance will persist and which will fade, however. First, some non-<emph>g</emph> effects may cause persistent differences between individuals. Educational attainment has been shown to predict working memory related IQ subtests at age 70, controlling for initial IQ score at age 11, with no effect on latent <emph>g</emph> (Ritchie et al., [<reflink idref="bib83" id="ref42">83</reflink>]). Ritchie and Tucker-Drob ([<reflink idref="bib82" id="ref43">82</reflink>]) further presented strong meta-analytic evidence for causal effects of education on a variety of cognitive outcomes, although they were not able to test whether or not impacts were distinguishable from psychometric <emph>g</emph>. Second, sustained and environmentally induced secular gains on IQ test scores, the <emph>Flynn effect</emph>, have been recorded across the twentieth century, with larger gains occurring on a factor of so-called fluid intelligence measures (Baker et al., [<reflink idref="bib4" id="ref44">4</reflink>]; Colom et al., [<reflink idref="bib17" id="ref45">17</reflink>]; Flynn, [<reflink idref="bib30" id="ref46">30</reflink>], [<reflink idref="bib31" id="ref47">31</reflink>]). By contrast, while further evidence from studies on the Flynn effect points to gains on specific cognitive skills, these were separate from <emph>g</emph> (Flynn et al., [<reflink idref="bib32" id="ref48">32</reflink>]; Te Nijenhuis & van der Flier, [<reflink idref="bib91" id="ref49">91</reflink>]; Wicherts et al., [<reflink idref="bib102" id="ref50">102</reflink>]).</p> <p>Based on the logic and evidence above, it is clear that in some cases, interventions not affecting psychometric <emph>g</emph> are likely to produce effects that fade. On the other hand, for the raising IQ/raising <emph>g</emph> distinction to underlie both fadeout and persistence, one would expect gains on <emph>g</emph> to be followed by lasting effects on children's IQ scores. Investigating the latter prediction in the Infant Health and Development Program experiment (IHDP), impacts on IQ that were not discernable from impacts on psychometric <emph>g</emph> (about a.5 sample <emph>SD</emph> increase on <emph>g</emph>) for age-3 children was followed by complete fadeout on IQ and <emph>g</emph> by ages 5 and 8 (Protzko, [<reflink idref="bib72" id="ref51">72</reflink>]). This pattern of findings is inconsistent with the raising IQ/raising <emph>g</emph> distinction. It is possible that cognitive ability is sufficiently undifferentiated at age 3 that non-<emph>g</emph> effects could be mistaken for <emph>g</emph> effects. Nonetheless, these findings raise some doubt about the hypothesis that <emph>g</emph> underlies persistence. The raising IQ/raising <emph>g</emph> distinction deserves further scrutiny.</p> <hd id="AN0156866495-2">Present Study</hd> <p>The goal of the present study is to investigate relations between the classic Carolina Abecedarian Project early intervention (thereafter, Abecedarian), IQ subtest scores, and psychometric <emph>g</emph>. Abecedarian stands apart among interventions geared toward disadvantaged children by its scope and intensity, and its generated effects across various long-term outcomes: educational attainment, level of employment (e.g., skilled jobs), crime reduction and health (e.g., Campbell et al., [<reflink idref="bib11" id="ref52">11</reflink>]; García et al., [<reflink idref="bib33" id="ref53">33</reflink>]). Started in the early 1970s, this 50+ year-longitudinal study initially aimed to demonstrate that developmental delay of "disadvantaged children can be prevented, and to explain how various psychological and biological processes are affected by such preventive attempts" (Ramey et al., 1974/[<reflink idref="bib77" id="ref54">77</reflink>], p. 9, [<reflink idref="bib78" id="ref55">78</reflink>]). Remarkably, Abecedarian produced an early boost on IQ scores (age 5), which persisted across childhood and adolescence (ages 6.5, 8, 12, 15) and into adulthood (age 21).</p> <p>It was also apparent that Abecedarian had generated gains on other cognitive tests like Piaget's conservation tasks, and produced similar gains on scholastic achievement measures (Campbell et al., [<reflink idref="bib12" id="ref56">12</reflink>]; Campbell & Ramey, [<reflink idref="bib9" id="ref57">9</reflink>], [<reflink idref="bib10" id="ref58">10</reflink>]). Jensen—a prominent skeptic of environmental interventions' impact on the raising of intelligence (e.g., Jensen, [<reflink idref="bib45" id="ref59">45</reflink>])—suggested that the Abecedarian intervention might have indeed raised <emph>g</emph> (Jensen, [<reflink idref="bib46" id="ref60">46</reflink>], p. 344). However, up to now, no formal methods have been applied to test this suggestion. Attempting to do so is important for at least two theoretical reasons. First, despite much debate over the malleability of <emph>g</emph>, to our knowledge no one has ever tested, at any time during participants' development, whether Abecedarian influenced scores on psychometric <emph>g</emph>. Second, the raising IQ/raising <emph>g</emph> distinction predicts that the early intervention's persistent observed effects on IQ test scores should parallel those on <emph>g</emph>. If supported, both hypotheses would give credence to the broader assertion that Abecedarian, which appears to show a unique level of persistent impacts on IQ, impacted a broad range of cognitive skills.</p> <hd id="AN0156866495-3">Method</hd> <p></p> <hd id="AN0156866495-4">Abecedarian Sample</hd> <p>Targeting children born to disadvantaged backgrounds, Abecedarian drew its sample across four cohorts of families, from 1972 to 1977. Living in Chapel Hill, North Carolina or its vicinity, they were screened at local social services agencies or prenatal clinics, based on a 13-indicator sociodemographic high-risk index cut-point score of 11 or higher (criteria included level of family income, parental education, maternal IQ; for details on each criterion, see Ramey, [<reflink idref="bib75" id="ref61">75</reflink>]). Ninety-nine percent of the families screened expressed interest in participating and were visited by the project staff who presented the intervention and the random assignment procedure. Ninety-three percent of eligible families remained interested in the study (Ramey et al., [<reflink idref="bib79" id="ref62">79</reflink>]).</p> <p>Randomization was done by matching children in pairs across eligible families based on the child's gender, maternal IQ score, the number of siblings, and the overall score on the high-risk index (for details on the randomization protocol and adjustments, see Ramey et al., 1974/[<reflink idref="bib77" id="ref63">77</reflink>]; and the appendix of García et al., [<reflink idref="bib33" id="ref64">33</reflink>]). The baseline pretreatment sample we used consisted of 107 infants: 54 assigned in early intervention (treatment) group (28 boys; 26 girls) and 53 to control (23 boys; 30 girls); 97 percent of all children were identified as African American. Infants in the treatment group entered the study at an age ranging from 3 to 21 weeks (average: 8.8 weeks). As shown in Table 1, we found no evidence of group imbalance on entry-level family characteristics (Table 1).</p> <p>Table 1. Pretreatment descriptive statistics across treatment and control groups.</p> <p> <ephtml> <table><thead><tr><td /><td>Treatment (T)</td><td>Control (C)</td><td>(T–C)</td><td /></tr><tr><td /><td><italic>M</italic></td><td><italic>SD</italic></td><td><italic>M</italic></td><td><italic>SD</italic></td><td>Diff.</td><td><italic>SE</italic></td></tr></thead><tbody valign="top"><tr><td>Birth Weight</td><td char=".">3.10</td><td char=".">0.59</td><td char=".">3.27</td><td char=".">0.64</td><td char=".">–0.17</td><td char=".">0.12</td></tr><tr><td>Gender</td><td char=".">0.48</td><td char=".">0.50</td><td char=".">0.57</td><td char=".">0.50</td><td char=".">–0.08</td><td char=".">0.10</td></tr><tr><td>Bayley MDI (at 3 mos.)</td><td char=".">95.44</td><td char=".">11.83</td><td char=".">95.42</td><td char=".">12.19</td><td char=".">0.02</td><td char=".">2.36</td></tr><tr><td>Mother's WAIS/WISC score</td><td char=".">85.39</td><td char=".">12.57</td><td char=".">84.25</td><td char=".">10.87</td><td char=".">1.14</td><td char=".">2.27</td></tr><tr><td> Age</td><td char=".">19.65</td><td char=".">3.89</td><td char=".">20.42</td><td char=".">5.74</td><td char=".">–0.77</td><td char=".">0.95</td></tr><tr><td> Education</td><td char=".">10.57</td><td char=".">1.73</td><td char=".">10.06</td><td char=".">1.84</td><td char=".">0.52</td><td char=".">0.35</td></tr><tr><td>Number of siblings</td><td char=".">0.48</td><td char=".">0.88</td><td char=".">0.74</td><td char=".">1.20</td><td char=".">–0.25</td><td char=".">0.20</td></tr><tr><td>Father lives in HH</td><td char=".">0.22</td><td char=".">0.42</td><td char=".">0.34</td><td char=".">0.48</td><td char=".">–0.12</td><td char=".">0.09</td></tr><tr><td>High-risk index</td><td char=".">19.89</td><td char=".">5.71</td><td char=".">21.53</td><td char=".">5.87</td><td char=".">–1.64</td><td char=".">1.12</td></tr><tr><td>Income</td><td char=".">4.50</td><td char=".">4.74</td><td char=".">3.96</td><td char=".">4.75</td><td char=".">0.54</td><td char=".">0.92</td></tr><tr><td>Sample size</td><td char=".">54</td><td char="." /><td char=".">53</td><td char="." /><td char=".">107</td><td char="." /></tr></tbody></table> </ephtml> </p> <p>1 <emph>Notes</emph>. <emph>SD</emph>: standard deviation. Diff.: mean differences between treatment and control group; <emph>p</emph>-values were all > 0.1 (<emph>t</emph> test). SE: standard error. Weight is in kilograms. Gender: Female = 1. Bayley MDI: mental development index. Mother's WAIS/WISC: Wechsler Adult Intelligence Scale/Wechsler Intelligence Scale for Children (for mothers younger than 16) full scale score (population median of 100; standard deviation of 15); age is in years and education stands for the last grade completed. Father lives in HH: father lives in mother's home at child's birth. High-risk index is a composite of 13 pre-intervention demographic weighted variables. Income: natural log of yearly family income at child's birth, in 2016 USD.</p> <p>In particular, there was no difference across cohorts and groups on the 3 months Bayley mental development index (MDI). Moreover, Ramey ([<reflink idref="bib74" id="ref65">74</reflink>]) conducted a multivariate analysis of variance repeated over that same index at 8 time points (from 6 to 54 months) and found no significant interactions between age at entrance in the program, cohorts, and group. Overall, there was no indication of an inadvertent selection into any of the two groups.</p> <hd id="AN0156866495-5">The Abecedarian Project</hd> <p>The overall project consisted of two interventions: (<reflink idref="bib1" id="ref66">1</reflink>) A high-quality multi-component (education, health care, nutrition) early intervention, delivered from birth to age 5; and (<reflink idref="bib2" id="ref67">2</reflink>) A school-age intervention (bi-weekly teacher home-visits) during the first 3 years of public schooling, starting in Kindergarten (Campbell & Burchinal, [<reflink idref="bib8" id="ref68">8</reflink>]). A separate random assignment was done for the school-age intervention, based on a 48 months IQ score pairwise matching. As in other evaluations analyses of the Abecedarian (e.g., Campbell et al., [<reflink idref="bib11" id="ref69">11</reflink>]; Conti et al., [<reflink idref="bib18" id="ref70">18</reflink>]), we tested whether equality on IQ full-scale and subtest scores between those who had received the school-age intervention from those who did not, could be rejected. This was never the case, which was in line with previous research showing that Abecedarian school-age intervention had no effect on a large set of cognitive measures, including IQ scores (Campbell et al., [<reflink idref="bib12" id="ref71">12</reflink>], [<reflink idref="bib11" id="ref72">11</reflink>]; Campbell & Ramey, [<reflink idref="bib10" id="ref73">10</reflink>]; Ramey, [<reflink idref="bib74" id="ref74">74</reflink>]). Thus, while an indicator signaling participation in the follow-up intervention was included along other baseline covariates for each time point estimation of <emph>g</emph>, exogeneity was derived from the early intervention random assignment only.</p> <hd id="AN0156866495-6">Early Intervention</hd> <p>The early intervention consisted of year-round (50 weeks), 6.5–9.75 hours a day, five days a week center-based childcare services up until entry in kindergarten. Transportation to and from the intervention center was offered. The staff, recruited locally, had degrees beyond high school, with an average of 7 years of working experience in early childcare; caregiver-child ratio was 1:3, for infants; 1:4 for toddlers (up to 3 years old); 1:5–6 for children between 3 and 5 years old (Elango et al., [<reflink idref="bib25" id="ref75">25</reflink>]). Developmentally appropriate curricula were developed emphasizing play around language and emotional development at all time of the intervention (Sparling & Lewis, [<reflink idref="bib86" id="ref76">86</reflink>], [<reflink idref="bib87" id="ref77">87</reflink>]). Frequent one-on-one or two-on-one child-adult interactions intensified by age 3 and 4 while children engaged with similar objects of attention in small group settings. Through detailed formative assessments, individualized caregiving by teachers ensured progressive internalizations of new objects of learning—e.g., repetition of structured verbal interactions during joint and shared activity—aiming to promote the child's acquisition and extension of higher-order thinking in a social context (Sparling et al., [<reflink idref="bib89" id="ref78">89</reflink>]; Sparling & Meunier, [<reflink idref="bib88" id="ref79">88</reflink>]). Of note, control group children also attended some form of early childhood education: 75% of them had attended some alternative center-based childcare by age 5 (Elango et al., [<reflink idref="bib25" id="ref80">25</reflink>]).</p> <p>Pediatric care (routine screening; immunizations) was systematically provided, along with daily meals (breakfast, lunch, and snack; iron-fortified formula; for a thorough description of the health care and nutritional components, see Campbell et al., [<reflink idref="bib11" id="ref81">11</reflink>]). Control group families also received iron-fortified formula up to age 15 months. Control group's first cohort also received pediatric care at the center up to age 1; after which, and for the later three cohorts, no medical care was provided. However, in 1970s' North Carolina, eligible family could have access to Medicaid. Finally, regarding maintenance of control group subjects over time, Abecedarian staff were able to locate them thanks to having an extensive social network among the area African American community. Moreover, control group families received financial compensation when meeting with project staff for interviews and testing (M.R. Burchinal, personal communication, July 31, 2018).</p> <hd id="AN0156866495-7">Measures</hd> <p></p> <hd id="AN0156866495-8">IQ Tests</hd> <p>Abecedarian personnel administered full-scale IQ test batteries (comprising Verbal and Performance subscales) of the Wechsler Preschool and Primary Scale of Intelligence at age 5 (WPPSI; Wechsler, [<reflink idref="bib97" id="ref82">97</reflink>]); followed by the Wechsler Intelligence Scale for Children at ages 6.5 (WISC-R; Wechsler, [<reflink idref="bib98" id="ref83">98</reflink>]). Then, examiners with no knowledge of participants' treatment history administered the WISC-R at age 8, 12, and 15; and the Wechsler Adult Intelligence Scale at age 21 (WAIS-R, Wechsler, [<reflink idref="bib99" id="ref84">99</reflink>]). These instruments' reliability and validity have been established (Flanagan et al., [<reflink idref="bib29" id="ref85">29</reflink>]). All testing sessions took place at the research center (Campbell et al., [<reflink idref="bib12" id="ref86">12</reflink>]; Campbell & Ramey, [<reflink idref="bib10" id="ref87">10</reflink>]). Descriptive statistics for each age wave full-scale and subscale IQ scores are displayed in Table 2 (IQ scores kernel density distributions by groups can be found in the Supplementary Online Material, thereafter SOM, Figure S.1).</p> <p>Table 2. Wechsler IQ Test Scores: full scale, verbal, & performance by age wave.</p> <p> <ephtml> <table><thead><tr><td /><td>Full Sample</td><td>Control (C)</td><td>Treatment (T)</td><td>(T–C)</td></tr><tr><td /><td><italic>M</italic></td><td><italic>SD</italic></td><td>N</td><td><italic>M</italic></td><td><italic>SD</italic></td><td>N</td><td><italic>M</italic></td><td><italic>SD</italic></td><td>N</td><td>Diff.</td><td><italic>SE</italic></td><td>ES</td></tr></thead><tbody valign="top"><tr><td>WPPSI age 5</td><td char=".">97.8</td><td char=".">12.8</td><td char=".">95</td><td char=".">94.0</td><td char=".">13.7</td><td char=".">46</td><td char=".">101.4</td><td char=".">11.0</td><td char=".">49</td><td char="."><bold>7.5***</bold></td><td char=".">(2.5)</td><td char=".">0.55</td></tr><tr><td>WPPSI age 5 Verbal</td><td char=".">98.7</td><td char=".">13.1</td><td char=".">95</td><td char=".">94.5</td><td char=".">13.6</td><td char=".">46</td><td char=".">102.7</td><td char=".">11.4</td><td char=".">49</td><td char="."><bold>8.2***</bold></td><td char=".">(2.6)</td><td char=".">0.61</td></tr><tr><td>WPPSI age 5 Performance</td><td char=".">97.3</td><td char=".">12.4</td><td char=".">95</td><td char=".">94.7</td><td char=".">13.3</td><td char=".">46</td><td char=".">99.7</td><td char=".">11.1</td><td char=".">49</td><td char="."><bold>5.1**</bold></td><td char=".">(2.5)</td><td char=".">0.38</td></tr><tr><td>WISC-R age 6.5</td><td char=".">95.5</td><td char=".">12.4</td><td char=".">90</td><td char=".">92.3</td><td char=".">11.8</td><td char=".">43</td><td char=".">98.5</td><td char=".">12.3</td><td char=".">47</td><td char="."><bold>6.2**</bold></td><td char=".">(2.5)</td><td char=".">0.53</td></tr><tr><td>WISC-R age 6.5 Verbal</td><td char=".">96.2</td><td char=".">13.2</td><td char=".">90</td><td char=".">93.6</td><td char=".">10.9</td><td char=".">43</td><td char=".">98.5</td><td char=".">14.8</td><td char=".">47</td><td char=".">4.9</td><td char=".">(2.8)</td><td char=".">0.45</td></tr><tr><td>WISC-R age 6.5 Performance</td><td char=".">93.4</td><td char=".">14.8</td><td char=".">90</td><td char=".">92.4</td><td char=".">13.9</td><td char=".">43</td><td char=".">94.3</td><td char=".">15.6</td><td char=".">47</td><td char=".">1.9</td><td char=".">(3.1)</td><td char=".">0.13</td></tr><tr><td>WISC-R age 8</td><td char=".">95.7</td><td char=".">12.2</td><td char=".">90</td><td char=".">93.3</td><td char=".">12.4</td><td char=".">42</td><td char=".">97.8</td><td char=".">11.8</td><td char=".">48</td><td char=".">4.5</td><td char=".">(2.6)</td><td char=".">0.36</td></tr><tr><td>WISC-R age 8 Verbal</td><td char=".">95.4</td><td char=".">12.6</td><td char=".">90</td><td char=".">92.6</td><td char=".">13.0</td><td char=".">42</td><td char=".">97.9</td><td char=".">11.9</td><td char=".">48</td><td char="."><bold>5.3**</bold></td><td char=".">(2.6)</td><td char=".">0.41</td></tr><tr><td>WISC-R age 8 Performance</td><td char=".">97.0</td><td char=".">12.9</td><td char=".">90</td><td char=".">95.2</td><td char=".">13.3</td><td char=".">42</td><td char=".">98.6</td><td char=".">12.6</td><td char=".">48</td><td char=".">3.4</td><td char=".">(2.7)</td><td char=".">0.26</td></tr><tr><td>WISC-R age 12</td><td char=".">91.6</td><td char=".">11.1</td><td char=".">98</td><td char=".">88.2</td><td char=".">11.6</td><td char=".">47</td><td char=".">94.8</td><td char=".">9.6</td><td char=".">51</td><td char="."><bold>6.6***</bold></td><td char=".">(2.1)</td><td char=".">0.57</td></tr><tr><td>WISC-R age 12 Verbal</td><td char=".">90.7</td><td char=".">11.3</td><td char=".">98</td><td char=".">86.9</td><td char=".">10.9</td><td char=".">47</td><td char=".">94.2</td><td char=".">10.6</td><td char=".">51</td><td char="."><bold>7.4***</bold></td><td char=".">(2.2)</td><td char=".">0.68</td></tr><tr><td>WISC-R age 12 Performance</td><td char=".">94.6</td><td char=".">12.6</td><td char=".">98</td><td char=".">92.1</td><td char=".">14.8</td><td char=".">47</td><td char=".">96.8</td><td char=".">9.8</td><td char=".">51</td><td char=".">4.7</td><td char=".">(2.5)</td><td char=".">0.32</td></tr><tr><td>WISC-R age 15</td><td char=".">92.9</td><td char=".">12.2</td><td char=".">101</td><td char=".">89.8</td><td char=".">12.6</td><td char=".">49</td><td char=".">95.8</td><td char=".">11.2</td><td char=".">52</td><td char="."><bold>6.1**</bold></td><td char=".">(2.4)</td><td char=".">0.48</td></tr><tr><td>WISC-R age 15 Verbal</td><td char=".">89.4</td><td char=".">12.3</td><td char=".">101</td><td char=".">85.7</td><td char=".">11.7</td><td char=".">49</td><td char=".">92.8</td><td char=".">12.0</td><td char=".">52</td><td char="."><bold>7.1***</bold></td><td char=".">(2.4)</td><td char=".">0.61</td></tr><tr><td>WISC-R age 15 Performance</td><td char=".">98.4</td><td char=".">13.8</td><td char=".">101</td><td char=".">96.0</td><td char=".">14.9</td><td char=".">49</td><td char=".">100.6</td><td char=".">12.5</td><td char=".">52</td><td char=".">4.6</td><td char=".">(2.7)</td><td char=".">0.31</td></tr><tr><td>WAIS-R age 21</td><td char=".">87.5</td><td char=".">9.7</td><td char=".">103</td><td char=".">85.2</td><td char=".">8.6</td><td char=".">51</td><td char=".">89.8</td><td char=".">10.2</td><td char=".">52</td><td char="."><bold>4.5**</bold></td><td char=".">(1.9)</td><td char=".">0.53</td></tr><tr><td>WAIS-R age 21 Verbal</td><td char=".">86.4</td><td char=".">9.5</td><td char=".">103</td><td char=".">84.2</td><td char=".">8.3</td><td char=".">51</td><td char=".">88.6</td><td char=".">10.3</td><td char=".">52</td><td char="."><bold>4.3**</bold></td><td char=".">(1.8)</td><td char=".">0.52</td></tr><tr><td>WAIS-R age 21 Performance</td><td char=".">91.7</td><td char=".">11.5</td><td char=".">103</td><td char=".">89.9</td><td char=".">11.2</td><td char=".">51</td><td char=".">93.5</td><td char=".">11.6</td><td char=".">52</td><td char=".">3.6</td><td char=".">(2.3)</td><td char=".">0.32</td></tr></tbody></table> </ephtml> </p> <p>2 <emph>Note.</emph> * <emph>p</emph> <.05; ** <emph>p</emph> <.01 (<emph>t</emph> test). Units are in IQ points. Wechsler IQ tests are scaled to have a population median of 100 and standard deviation of 15. ES: effect size calculated as Diff./<emph>SD</emph><subs>control</subs>.</p> <p>Across age waves, we considered IQ subtests <emph>Information</emph> ('retrieving and using broad and factual knowledge'), <emph>Vocabulary</emph> ('naming object or defining words'), <emph>Comprehension</emph> ('understanding of socially situated general principles or concepts'), <emph>Arithmetic</emph> (involving 'carrying some form of mental calculation, applied to situations presented verbally and figurally'), parts of the Weschler Verbal subscales. And <emph>Picture Completion</emph> ('visual perceiving, understanding, and organizing')<emph>, Picture Arrangement</emph> ('chronological sequencing of pictures'), <emph>Geometric Design</emph> (for the WPPSI only; 'reproducing geometric figures'), <emph>Block Design</emph> and <emph>Object Assembly</emph> ('analyzing visual stimuli; part-whole relationship; followed by constructing'), which are parts of the Weschler Performance subscales (Flanagan et al., [<reflink idref="bib29" id="ref88">29</reflink>]).</p> <hd id="AN0156866495-9">Modeling of g</hd> <p>Seven of the subtests described above were used throughout for the modeling of <emph>g</emph> (out of the 10, 12, & 11 subtests of the WPPSI, WISC-R, and WAIS-R, respectively). A wide range of cognitive skills was thus covered, while keeping number of parameters and model fit adequate. For each age wave, we modeled <emph>g</emph> as a second-order common factor, loading on Verbal and Performance first-order factors, themselves loading on individual subtests, following Wechsler subscale structures. Various factor analyses of the Wechsler's' have found <emph>Arithmetic</emph> difficult to pinpoint to a particular group factor, and in some cases yielding better fit when loading directly onto <emph>g</emph> (Gignac, [<reflink idref="bib34" id="ref89">34</reflink>]; Reynolds & Keith, [<reflink idref="bib81" id="ref90">81</reflink>]). Thus, we singled out <emph>Arithmetic</emph> to allow for the identification of a unitary <emph>g</emph> (Figure 1), which differs in this respect from how previous Abecedarian studies measured IQ (e.g., Campbell et al., [<reflink idref="bib12" id="ref91">12</reflink>]).</p> <p>Graph: Figure 1. Second-order g, first-order verbal & performance, measured by WPPSI subtests at age 5. Note. Treatment/Control group standardized loadings λi (p <.001) for the partial strict invariance model (within brackets: configural model).</p> <p>Across each age wave, we included identical or very similar subtests. When baseline fit deteriorated for a given age wave, fit was improved by switching subtest within the same Wechsler subscale structure (SOM, Figure S.2). Overall, factor loadings were consistent with those found in similar models of larger samples (e.g., Keith & Reynolds, [<reflink idref="bib53" id="ref92">53</reflink>]). Some have raised concerns regarding modeling with parcels instead of items when, for example, subtests might risk obscuring item-level multi-dimensionality (e.g., Bandalos & Finney, [<reflink idref="bib5" id="ref93">5</reflink>]). Others have noted substantial advantages of parcels, like higher reliability and lower indicator-to-sample size ratio (Little et al., [<reflink idref="bib56" id="ref94">56</reflink>]). Yet, the potential sort of bias induced by modeling parcels as indicators (e.g., Marsh et al., [<reflink idref="bib59" id="ref95">59</reflink>]) are perhaps more salient with self-report instruments than with performance tasks like those found in the Wechsler IQ subtests (whose reliability and validity are well established). More practically, item-level data not being available, subtests were our default indicators.</p> <p>Because of slight differences in the composition of the test batteries, of the relative geographical, socioeconomic, and demographic homogeneity of the sample, and potential for restriction of range, we might have expected some attenuation in the correlation of the various <emph>g</emph> factors (Johnson et al., [<reflink idref="bib50" id="ref96">50</reflink>], [<reflink idref="bib51" id="ref97">51</reflink>]). The latter were however reasonably highly correlated across time: e.g.,.91 between age 5 and age 8;.83 between age 5 and 21; and.90 between age 8 and age 21. This suggested the measurement of <emph>g</emph> was sufficiently consistent for our purposes (we reiterate that we do not make strong assumptions about the nature of <emph>g</emph>, including that the nature and composition of <emph>g</emph> factors were identical across the various time points).</p> <hd id="AN0156866495-10">Reliability</hd> <p>For each age wave, composite reliability indexes (McDonald, [<reflink idref="bib62" id="ref98">62</reflink>]; Raykov & Marcoulides, [<reflink idref="bib80" id="ref99">80</reflink>]) obtained for <emph>g</emph> stood between <emph>ω</emph> =.87 (99% CI = [.80;.94]) and.91 (99% CI = [.86;.95]), indicating strong internal consistency reliability (SOM, Table S.1).</p> <hd id="AN0156866495-11">Analysis</hd> <p></p> <hd id="AN0156866495-12">Attrition</hd> <p>Sample sizes are slightly different across age waves, ranging from a count of 90–103; compared to baseline sample size of 107. Pretreatment variables and other covariates were compared across groups and for each age wave: all descriptive statistics were stable (SOM, Tables S.2–S.7), indicating that group baseline variable balance was robust to attrition. Using Mplus Version 8 (Muthén & Muthén, [<reflink idref="bib68" id="ref100">68</reflink>]), estimations were obtained via maximum likelihood with bias-corrected bootstrapping (case resampling with replacement; 5000 replications). To handle missing data, full information maximum likelihood (FIML; Enders, [<reflink idref="bib26" id="ref101">26</reflink>]; Graham, [<reflink idref="bib36" id="ref102">36</reflink>]) was computed using the High-Risk index—a key element in sample selection, and subsuming some of baseline data—as an auxiliary variable (Graham, [<reflink idref="bib35" id="ref103">35</reflink>]), while keeping other baseline covariates in the model: weight at birth; gender; maternal IQ; whether father was present in the home at birth; maternal age at birth; number of siblings; HOME score; and family income at birth; as well as an indicator signaling participation in Abecedarian follow-up intervention (for descriptive statistics, see SOM, Tables S.2–S.7).</p> <hd id="AN0156866495-13">Procedure</hd> <p>Here is in general terms the psychometric rationale for conducting measurement invariance testing in a multi-group confirmatory factor analysis (MGCFA) setting (see Millsap, [<reflink idref="bib65" id="ref104">65</reflink>]). If factor loadings were not invariant across groups, the underlying covariance structures would be group-specific. This would result in a <emph>non-uniform</emph> bias in the comparison of group latent means: an <emph>identical</emph> position across groups on the latent factor would map onto <emph>different</emph> observed scores. Moreover, this bias would <emph>vary</emph> along the factor rank. Assuming that factor loading equivalence stands; intercept(s) non-invariance across groups would yield a <emph>uniform</emph> bias: slopes would be the same, but intercept(s) would differ between groups (Lubke et al., [<reflink idref="bib58" id="ref105">58</reflink>]; Wicherts et al., [<reflink idref="bib102" id="ref106">102</reflink>]). In this case, unbiased latent mean estimations might be obtained by letting the non-invariant intercept(s) vary across groups (Lemos et al., [<reflink idref="bib55" id="ref107">55</reflink>]; Wicherts & Dolan, [<reflink idref="bib101" id="ref108">101</reflink>]). Finally, although not strictly necessary to obtain an unbiased latent mean estimation, invariant residual factor variances are desirable, since residuals include random measurement error as well as subtest/lower-order factor specific <emph>unmodeled</emph> sources of variance (DeShon, [<reflink idref="bib21" id="ref109">21</reflink>]; Lubke et al., [<reflink idref="bib58" id="ref110">58</reflink>]; Lubke & Dolan, [<reflink idref="bib57" id="ref111">57</reflink>]; Wicherts & Dolan, [<reflink idref="bib101" id="ref112">101</reflink>]).</p> <p>In the present study, the procedure unfolded like so: for each age tested, the factor structure was configured equivalently across groups and the fit was assessed (<emph>configurational</emph> invariance). Then, we systematically compared the fit of nested models, from least to most constrained: starting with imposing invariance over first-, then second-order factor loadings (<emph>metric</emph> invariance); followed by constraining subtests intercepts, then first-order factors intercepts (<emph>scalar</emph> invariance); finally, by constraining factors and subtests residual variances (<emph>strict</emph> factorial invariance; see Meredith, [<reflink idref="bib64" id="ref113">64</reflink>]).</p> <p>Fit was systematically assessed through testing the significance of <bold><emph>χ<sups>2</sups></emph></bold>differences, Comparative Fit Index (CFI; Bentler, [<reflink idref="bib7" id="ref114">7</reflink>]; Hu & Bentler, [<reflink idref="bib44" id="ref115">44</reflink>]) pairwise difference—where, following Cheung and Rensvold ([<reflink idref="bib16" id="ref116">16</reflink>]), a downward difference greater than.01 suggests non-invariance—as well as changes in Root Mean Square Error of Approximation index (Hu & Bentler, [<reflink idref="bib43" id="ref117">43</reflink>]). When misfit occurred, modification indices (Sörbom, [<reflink idref="bib85" id="ref118">85</reflink>]) were used to locate which of previously set equivalent parameters (e.g., which estimated intercepts) would yield the greatest significant gain in model fit if let unconstrained.</p> <hd id="AN0156866495-14">Power</hd> <p>Based on observed overall IQ group mean-difference (averaged Cohen's d of.50), an average group sizes of 46 and 50 (control and treatment, respectively), and a type I error rate (alpha) of.05, the computed post-hoc achieved power amounted to.79 (obtained with G*Power 3; Faul et al., [<reflink idref="bib28" id="ref119">28</reflink>]), suggesting a viable sample size to detect Abecedarian IQ effect, notwithstanding mentioned attrition. However, for multi-group confirmatory factor analysis (MGCFA), sample sizes were too small to detect all but the most severe lack of invariance and large <emph>g</emph> effects (Meade et al., [<reflink idref="bib63" id="ref120">63</reflink>]). In an attempt to mitigate this issue, a multiple-indicators, multiple-causes model (MIMIC; e.g., Jöreskog & Goldberger, [<reflink idref="bib52" id="ref121">52</reflink>]) was also implemented and results reported for each age wave. Because MIMIC models pool observations from all groups, and assuming that overall <emph>metric</emph> invariance holds, such models are considered more appropriate to assess heterogeneity over latent means and intercepts with small samples (Bauer, [<reflink idref="bib6" id="ref122">6</reflink>]; Hancock, [<reflink idref="bib38" id="ref123">38</reflink>]; Muthén, [<reflink idref="bib67" id="ref124">67</reflink>]): degrees of freedom to sample size ratios are about twice as small for MIMIC models (33–107) compared to factor loading constrained (metric) MGCFA (Table 3; SOM Tables S8–S16).</p> <p>Table 3. WPPSI at age 5—assessment of factorial invariance.</p> <p> <ephtml> <table><thead><tr><td>Models</td><td><italic>χ<sup>2</sup></italic></td><td /><td><italic>df</italic></td><td><italic>p</italic></td><td>CFI</td><td>RMSEA</td><td>Compare</td><td>Δχ<sup>2</sup> (<italic>p</italic>)</td><td>ΔCFI</td></tr></thead><tbody valign="top"><tr><td>1. Invariant Factorial Structure (configural)</td><td char=".">30.41</td><td char="." /><td char=".">24</td><td char=".">.17</td><td char=".">.971</td><td char=".">.07</td><td char="." /><td char="." /><td char="." /></tr><tr><td>2a. Invariant 1<sup>st</sup> Order Factor Loadings (metric)</td><td char=".">35.46</td><td char="." /><td char=".">28</td><td char=".">.16</td><td char=".">.968</td><td char=".">.07</td><td char=".">2a–1</td><td char=".">5.05 (.28)</td><td char=".">–.003</td></tr><tr><td>2b. Invariant 2<sup>nd</sup> Order Factor Loadings (metric)</td><td char=".">39.16</td><td char="." /><td char=".">30</td><td char=".">.12</td><td char=".">.959</td><td char=".">.08</td><td char=".">2b–2a</td><td char=".">4.85 (.09)</td><td char=".">–.009</td></tr><tr><td><bold>3a. Invariant Subtest Intercepts (scalar)</bold></td><td char="."><bold>54.24</bold></td><td char="." /><td char="."><bold>37</bold></td><td char="."><bold>.03</bold></td><td char="."><bold>.922</bold></td><td char="."><bold>.10</bold></td><td char="."><bold>3a–2b</bold></td><td char="."><bold>15.10 (.04)</bold></td><td char="."><bold>–.037</bold></td></tr><tr><td>3b. <italic>Information</italic> intercept free (partial scalar)</td><td char=".">45.29</td><td char="." /><td char=".">36</td><td char=".">.14</td><td char=".">.958</td><td char=".">.07</td><td char=".">3b–2b</td><td char=".">6.13 (.41)</td><td char=".">–.001</td></tr><tr><td>3c. Invariant 1<sup>st</sup> Order Factor Intercepts (scalar)</td><td char=".">46.16</td><td char="." /><td char=".">37</td><td char=".">.14</td><td char=".">.959</td><td char=".">.07</td><td char=".">3c–3b</td><td char=".">.87 (.35)</td><td char=".">.001</td></tr><tr><td>4a. Invariant Factor Residual Variances (partial strict)</td><td char=".">46.74</td><td char="." /><td char=".">38</td><td char=".">.16</td><td char=".">.961</td><td char=".">.07</td><td char=".">4a–3c</td><td char=".">.60 (.45)</td><td char=".">.002</td></tr><tr><td>4b. Invariant Subtest Residual Variances (partial strict)</td><td char=".">49.55</td><td char="." /><td char=".">44</td><td char=".">.26</td><td char=".">.975</td><td char=".">.05</td><td char=".">4b–4a</td><td char=".">2.81 (.83)</td><td char=".">.014</td></tr></tbody></table> </ephtml> </p> <p>3 <emph>Notes.</emph> n (T) = 49; n (C) = 46. "Invariant" means "forced to equivalence across groups." <bold>Bold</bold> = lack of invariance.</p> <hd id="AN0156866495-15">Summary</hd> <p>Using MGCFA for factorial invariance testing and latent mean difference estimation at several time points (along with a check with MIMIC models), the present study asked whether Abecedarian's effect on cognitive skills as measured by IQ scores were broad, consistent with an effect on psychometric <emph>g</emph> and/or on test-specific variance. The raising IQ/raising <emph>g</emph> distinction predicts fadeout of test-specific effects, but also that broad impacts not distinguishable from psychometric <emph>g</emph> might persist.</p> <hd id="AN0156866495-16">Results</hd> <p></p> <hd id="AN0156866495-17">Factorial Invariance Age 5</hd> <p>Imposing model restrictions to investigate measurement invariance showed invariance held for all factor loadings (λ<subs>i</subs>), and all intercepts but for one subtest (<emph>Information</emph>; see Table 3). Model 3a indices signaled a fit deterioration compared to preceding model 2b: significant Δχ<sups>2</sups>; a comparative fit index downward difference (ΔCFI) of −.037, which is larger than recommended invariance threshold of −.01 (Cheung & Rensvold, [<reflink idref="bib16" id="ref125">16</reflink>]). This situation could have occurred because the children in the intervention group were exposed to some of the test-specific information on that subtest. Modification indices inspection suggested setting subtest <emph>Information</emph> intercept free. Doing so yielded back a good fit (Table 3, model 3b). Invariance held as well when residual variances were subsequently constrained (Table 3, model 4a & 4b). The partial strict invariant structure of the scale—composite reliability index <emph>ω</emph> =.90 (99% CI = [.86;.94])— indicated that, after allowing <emph>Information</emph> intercept to vary freely between groups, the group difference was not distinguishable from within-group variation in psychometric <emph>g.</emph></p> <hd id="AN0156866495-18">Observed vs. Estimated Subtest Effects</hd> <p>Comparing the effect of the Abecedarian intervention on <emph>g</emph> with IQ assessed at age 5, the treatment group (T) was higher on <emph>g</emph> by 4.4 IQ points (<emph>p</emph> =.02; 95% bias-corrected CI = [0.9; 8.4]) than the control group (C; see Table 4; using FIML—sample size: <emph>n</emph> (T) = 54; n (C) = 53—a 4.3 IQ points <emph>g</emph> effect was obtained; <emph>p</emph> <.05; 95% CI = [1.0; 7.6]). Although differences recovered by <emph>g</emph> were not perfectly congruent with IQ subtests' observed differences, the discrepancies were not statistically significant, except for subtest <emph>Information</emph> with a discrepancy of about 8 IQ points (<emph>p</emph> <.01). This was expected because subtest <emph>Information</emph> was not invariant across groups.</p> <p>Table 4. Age 5—WPPSI. Observed vs. estimated.</p> <p> <ephtml> <table><thead><tr><td /><td>Info.</td><td>Vocab.</td><td>Comp.</td><td>PC</td><td>GD</td><td>BD</td><td>Arith.</td><td>IQ</td><td><italic>g</italic></td></tr></thead><tbody valign="top"><tr><td>C</td><td /><td /><td /><td /><td /><td /><td /><td /><td /></tr><tr><td> M</td><td>8.57</td><td>9.04</td><td>8.91</td><td>9.83</td><td>8.52</td><td>9.65</td><td>9.37</td><td>94</td><td /></tr><tr><td> SD</td><td>2.93</td><td>2.49</td><td>2.74</td><td>2.71</td><td>2.67</td><td>2.58</td><td>2.02</td><td>13.7</td><td /></tr><tr><td>T</td><td /><td /><td /><td /><td /><td /><td /><td /><td /></tr><tr><td> M</td><td>11.02</td><td>9.65</td><td>9.90</td><td>10.65</td><td>9.49</td><td>10.45</td><td>10.53</td><td>101</td><td /></tr><tr><td> SD</td><td>2.61</td><td>2.57</td><td>2.6</td><td>2.44</td><td>2.43</td><td>2.23</td><td>1.86</td><td>11.0</td><td /></tr><tr><td><italic>d</italic></td><td><bold>.82***</bold>(.19)</td><td>.20 (.17)</td><td>.33 (.18)</td><td>.28 (.18)</td><td>.32 (.17)</td><td>.27 (.17)</td><td><bold>.39**</bold>(.13)</td><td><bold>.50**</bold>(.17)</td><td><bold>.29*</bold>(.13) [.06;.56]</td></tr><tr><td>ΔIQ obs.</td><td>12.3 (2.9)</td><td>3.1 (2.6)</td><td>4.9 (2.7)</td><td>4.2 (2.7)</td><td>4.9 (2.6)</td><td>4.0 (2.5)</td><td>5.8 (2.0)</td><td>7.5 (2.5)</td><td /></tr><tr><td>ΔIQ from <italic>g</italic></td><td>3.9 (1.4)</td><td>3.3 (1.2)</td><td>2.9 (1.1)</td><td>3.4 (1.2)</td><td>2.9 (1.0)</td><td>3.1(1.1)</td><td>3.1 (1.4)</td><td>4.4 (1.9)</td><td /></tr></tbody></table> </ephtml> </p> <p>4 <emph>Notes.</emph> *<emph>p</emph> <.05; **<emph>p</emph> <.01; ***<emph>p</emph> <.001. Standard errors are in parenthesis. 95 % bias-corrected CI are in brackets. n (T) = 49; n (C) = 46. Subtests and IQ means are observed. Effect size <emph>d</emph> is calculated with population <emph>SD</emph> (3 or 15, for subtests and full-scale IQ, respectively). d of <emph>g</emph> is obtained from Model 4b (Table 2). Estimated subtest IQ points differences (ΔIQ) due to <emph>g</emph>: <emph>d</emph> of <emph>g</emph> × corresponding Schmid-Leiman <emph>g</emph> loading × <emph>SD</emph> (= 15); see SOM pp. 14–16.</p> <p>Using the same factor structure as with age 5, MGCFA at age 6.5, 8, 12, 15, and 21 (SOM, Figure S.2) showed no evidence for lack of strict factorial invariance (SOM, Tables S.8–S.16), except at age 12, where the Performance factor residual was set free (partial strict invariance; moreover a Wald test showed that residual parameters were only marginally different across groups; 2.83(<reflink idref="bib1" id="ref126">1</reflink>), <emph>p</emph> =.09). As predicted by the raising IQ/raising <emph>g</emph> distinction, a fadeout of subtest <emph>Information</emph> test-specific gain at age 5 was apparent by age 8 (SOM, Table S.11). Further, equality between IQ effect and <emph>g</emph> effect at age 5 (a difference of about 3 IQ points) could be rejected (<emph>p</emph> =.02; <emph>z</emph> test). This was not the case at any of the later age waves (from age 6.5 to 21) with differences between IQ effect and <emph>g</emph> effect running from about 0 to 2 IQ points, with respective two-sided <emph>p</emph>-values from <emph>z</emph> tests ranging from.11 to.87 (see Table 5, IQ—<emph>g</emph> column). This is also consistent with the raising IQ/raising <emph>g</emph> distinction, persistence in IQ effect was not distinguishable from persistence in <emph>g</emph> effect (Figure 2).</p> <p>Graph: Figure 2. Abecedarian IQ effect & g effect across age time points.</p> <p>Table 5. Abecedarian g effect, IQ effect and IQ/g difference for each age wave.</p> <p> <ephtml> <table><thead><tr><td /><td>Fit indices</td><td><italic>g</italic> effect size (IQ points units)</td><td>IQ effect size</td><td>IQ – <italic>g</italic></td></tr><tr><td>Age</td><td>CFI</td><td>RMSEA</td><td>χ<sup>2</sup></td><td>(1)</td><td>(2)</td><td>(3)</td><td>(4)</td><td>(5)</td><td>(6)</td></tr></thead><tbody valign="top"><tr><td>5</td><td char=".">.98</td><td char=".">.05</td><td char=".">49.6 (.26)</td><td char=".">4.4 (.022)</td><td char=".">4.4 (.024)</td><td char=".">4.7 (.006)</td><td char=".">4.7 (.003)</td><td char=".">7.5 (.004)</td><td char=".">3.1 (.02)</td></tr><tr><td>6.5</td><td char=".">1.0</td><td char=".">.00</td><td char=".">44.8 (.48)</td><td char=".">3.9 (.008)</td><td char=".">3.9 (.007)</td><td char=".">4.6 (.004)</td><td char=".">4.4 (.002)</td><td char=".">6.2 (.017)</td><td char=".">2.3 (.11)</td></tr><tr><td>8</td><td char=".">1.0</td><td char=".">.00</td><td char=".">44.0 (.51)</td><td char=".">4.9 (.039)</td><td char=".">4.8 (.021)</td><td char=".">4.3 (.021)</td><td char=".">4.7 (.013)</td><td char=".">4.5 (.081)</td><td>–0.3 (.80)</td></tr><tr><td>12</td><td char=".">1.0</td><td char=".">.01</td><td char=".">44.4 (.47)</td><td char=".">4.6 (.009)</td><td char=".">4.8 (.003)</td><td char=".">4.4 (.009)</td><td char=".">4.5 (.004)</td><td char=".">6.6 (.003)</td><td char=".">2.0 (.17)</td></tr><tr><td>15</td><td char=".">1.0</td><td char=".">.00</td><td char=".">34.7 (.87)</td><td char=".">3.8 (.050)</td><td char=".">4.0 (.030)</td><td char=".">3.7 (.059)</td><td char=".">3.9 (.031)</td><td char=".">6.1 (.012)</td><td char=".">2.3 (.19)</td></tr><tr><td>21</td><td char=".">.99</td><td char=".">.03</td><td char=".">46.6 (.41)</td><td char=".">4.4 (.003)</td><td char=".">4.4 (.007)</td><td char=".">3.7 (.015)</td><td char=".">4.1 (.002)</td><td char=".">4.5 (.016)</td><td char=".">0.2 (.87)</td></tr></tbody></table> </ephtml> </p> <p>5 <emph>Notes.</emph> Fit indices of MGCFA (partial) strict invariant model at each age time point. Two-sided <emph>p</emph>-values are in parenthesis (<emph>z</emph> test). (<reflink idref="bib1" id="ref127">1</reflink>) MGCFA; (<reflink idref="bib2" id="ref128">2</reflink>) MIMIC; (<reflink idref="bib3" id="ref129">3</reflink>) MIMIC with Covariates: weight at birth, gender, maternal IQ, father present in the home at birth, maternal age at birth, number of siblings, HOME score, family income at birth, High-Risk Index, Abecedarian follow-up intervention (SOM, Tables S.2–S.7); (<reflink idref="bib4" id="ref130">4</reflink>) FIML MIMIC with Covariates. IQ – <emph>g</emph>: (<reflink idref="bib6" id="ref131">6</reflink>) = (<reflink idref="bib5" id="ref132">5</reflink>) – (<reflink idref="bib1" id="ref133">1</reflink>).</p> <p>Table 5 also reports MGCFA model fit indices, the MGCFA <emph>g</emph> effect size in IQ points (<reflink idref="bib1" id="ref134">1</reflink>), and MIMIC <emph>g</emph> effect sizes in IQ points (2–4). For the MIMIC models (described in the method section), second-order factor <emph>g</emph> and all subtests in turn were regressed on Abecedarian. We found consistent <emph>g</emph> effects across age points. For age 5, a subtest <emph>Information</emph> specific effect of 6.3 IQ points (<emph>p</emph> <.01) was obtained, confirming <emph>Information</emph> MGCFA lack of scalar invariance. In line with MGCFA findings, we found no evidence of scalar invariance violation for other subtests at any of the age point with the MIMIC method.</p> <hd id="AN0156866495-19">Discussion</hd> <p>Investigating the relation between cognitive test scores and <emph>g</emph>, our results were consistent in two ways with the raising IQ/raising <emph>g</emph> distinction, whose contrapositive implies that a raising of <emph>g</emph> should be evidenced by the proportionality of <emph>g</emph> loadings with the persistence of impact across IQ subtests. First, the persistent effect of the Abecedarian program on IQ scores (Table 2) was not distinguishable from an impact on psychometric <emph>g</emph> (Figure 2; Table 5). This analysis brought some validity to the position that for an educational intervention effect on cognition to last, impacts on cognitive abilities must be broad, or must eventually become broad via some combination of transfer of learning and skill-environment transactions (Dickens & Flynn, [<reflink idref="bib22" id="ref135">22</reflink>]; Van Der Maas et al., [<reflink idref="bib94" id="ref136">94</reflink>]). Second, 8 out of the 12 IQ points in group difference on the <emph>Information</emph> subtest at age 5 were not captured by changes in <emph>g</emph>. By age 8, the effect on this subtest faded by precisely that much. This trend agreed with the position that effects on skills or knowledge not captured by <emph>g</emph> were more likely to fade. A complementary explanation for the <emph>Information</emph> fadeout effect would suggest that participants in the control condition eventually acquired more of basic factual knowledge, thereby closing-up some of the earlier gap (Bailey et al., [<reflink idref="bib3" id="ref137">3</reflink>]).</p> <hd id="AN0156866495-20">Mechanisms</hd> <p>What could explain the persistent effect of the Abecedarian intervention on psychometric <emph>g</emph> and IQ scores? Abecedarian was conceived with an emphasis on communicative competence and evidence of early verbal development as a potential mediator of later cognitive achievement (Campbell & Burchinal, [<reflink idref="bib8" id="ref138">8</reflink>]), and elementary school reading skills have been hypothesized to affect academic outcomes many years later (Hernandez, [<reflink idref="bib42" id="ref139">42</reflink>]). On the other hand, fadeout on the verbal dimension occurred in another intensive early intervention focusing on language (i.e., IHDP; Protzko, [<reflink idref="bib72" id="ref140">72</reflink>]) and for some other early language or literacy-specific interventions (e.g., Suggate, [<reflink idref="bib90" id="ref141">90</reflink>]; Whitehurst et al., [<reflink idref="bib100" id="ref142">100</reflink>]).</p> <p>Abecedarian impacts on health later in adulthood has also been documented with treatment group participants' body mass index at one years-old statistically mediating half of the intervention's positive effect on mid-30s' hypertension and obesity (Campbell et al., [<reflink idref="bib13" id="ref143">13</reflink>]). That early health factors might have been determinant in changing early general cognitive ability is intuitive. However, it is not clear which causal bundle of health contributors could be linked with cognitive development (e.g., we found no weight difference either at birth or at age 5 between children of the treatment and control groups). Further, in Abecedarian, IQ impacts did seem to mediate program impacts on adult health (Muennig et al., [<reflink idref="bib66" id="ref144">66</reflink>]), suggesting multiple pathways through which the program influenced much later child outcomes.</p> <p>Other prototypical early interventions like Perry or IHDP, which had no persistent IQ or <emph>g</emph> score effects were less immersive: Perry counted 2.5 hours of instruction per day from mid-October through May, starting at age 3 or 4 and for 2 years (Schweinhart et al., [<reflink idref="bib84" id="ref145">84</reflink>]); IHDP ran only from birth to age 3 (Ramey et al., [<reflink idref="bib76" id="ref146">76</reflink>]). In particular, although IHDP implemented the Abecedarian curriculum (Ramey, [<reflink idref="bib75" id="ref147">75</reflink>]), only low-birth-weight babies were eligible to participate making the comparison with Abecedarian difficult (see McCarton et al., [<reflink idref="bib60" id="ref148">60</reflink>]). Yet, as mentioned in the introduction, Protzko ([<reflink idref="bib72" id="ref149">72</reflink>]) found gains on <emph>g</emph> for IHDP experimental group at age 3, followed by fadeout at age 5 and 8, providing evidence <emph>against</emph> the raising IQ/raising <emph>g</emph> distinction, inconsistent with our findings. One possible explanation for this inconsistency is that the raising IQ/raising <emph>g</emph> distinction is not a useful theory for forecasting how durable impacts will be after the end of treatment (as suggested by Protzko, [<reflink idref="bib72" id="ref150">72</reflink>]). This is worth taking seriously and can only be addressed via further similar research. Another possible explanation is that the IQ impacts in IHDP were observed at age 3, when cognitive skills may be insufficiently differentiated and measurement error sufficiently high that violations of measurement invariance may be difficult to detect. Further, the age 3 MGCFA was conducted at the item level, with 34 Stanford-Binet Intelligence Scale items loading on one factor. To the extent that non-<emph>g</emph> effects are concentrated at the subtest level, this approach would make violations more difficult to probe.</p> <p>By and large, there is still far too little research on the raising IQ/raising <emph>g</emph> distinction to allow a firm judgment on whether the breadth of test score impacts is useful for forecasting how durable impacts will be after the end of treatment in the context of evaluations of intensive educational programs. Still, our findings present the strongest evidence thus far for the claim that gains on <emph>g</emph> are indicative of persistent gains on measured cognitive skills. It is nonetheless difficult to pinpoint the actual working elements of Abecedarian impact mechanisms. More research on the contributions of various interventions, contexts, and interactions between them, and ideally replications in other programs, would be useful.</p> <hd id="AN0156866495-21">Limitations</hd> <p>How to best account for the positive manifold across development is an open question, which primarily depends on how cognitive ability development is conceived. Although we used a higher-order reflective model for psychometric purposes, we remained agnostic regarding the ontology of a single entity modeled as a latent psychometric <emph>g</emph> factor, focusing rather on whether the common causes of cognitive skills within the control group appear substantially different from those affected by the treatment, and whether such gains persisted or faded out. Practically, due to small sample size and the incomplete overlap in assessments across waves, our model was limited in the range of cognitive tasks covered and in the flexibility of its structure.</p> <p>Moreover, it is possible that the <emph>g</emph> factors we modeled captured somewhat different commonalities at each age wave (due to lack of power and incomplete overlap of subtests, we did not assess the measurement invariance of <emph>g</emph> across time). In particular, if <emph>g</emph> factors at age 5 and age 8 picked up widely heterogeneous variance then the claim that fadeout was due to age 5 non-<emph>g</emph> effect cannot be easily tested against the alternative claim that fadeout resulted from a difference in what was being measured at the different waves.</p> <p>Finally, with the possibility that some level of less severe non-invariance remained undetected, and given that we selected the Abecedarian data and not IQ impacts randomly selected from a population of early childhood education randomized controlled trials, our estimated <emph>g</emph> effects are best considered as upper bounds. As expected, analyses conducted with a correlated two-factor Verbal-Performance model showed similar measurement invariance patterns; overlaps between IQ subscales and corresponding factors effects (SOM, Table S.24) were consistent with those between full scale IQ and psychometric <emph>g</emph> effects (Table 5).</p> <hd id="AN0156866495-22">Implications</hd> <p>Overall, documented lack of persistent impacts on a broad range of cognitive tests in randomized evaluations of other early childhood educational programs makes it difficult to generate clear practical recommendations for generating persistent impacts on IQ based only on evidence from Abecedarian. The extent to which earlier programs can provide useful benchmarks for ECE interventions and programs is limited for at least two reasons. First, the counterfactual experiences of children not attending preschool full-time has changed since the 1970s, with more children today experiencing some form of early education. Second, the high cost of scaling-up program like Abecedarian makes such programs perhaps less relevant for policy and programmatic purposes (Philips et al., [<reflink idref="bib70" id="ref151">70</reflink>]).</p> <p>For these reasons, it might seem unreasonable to expect current or future ECE programs to replicate the magnitude of impacts on cognitive skills as measured by IQ tests estimated in Abecedarian. However, the question of whether the breadth of intervention impacts across skills might forecast durable impacts should be of interest to designers and evaluators of ECE programs. We hope additional work on the relation between intervention persistence and breadth of impacts could continue to inform theories of the possible mechanisms through which early interventions generate persistent or fading impacts on cognitive skills. This would inform theories of cognitive development and, potentially, the design of educational interventions with persistent effects.</p> <ref id="AN0156866495-23"> <title> Footnotes </title> <blist> <bibl id="bib1" idref="ref1" type="bt">1</bibl> <bibtext> Supplemental data for this article can be accessed at https://doi.org/10.1080/19345747.2021.1969711.</bibtext> </blist> </ref> <ref id="AN0156866495-24"> <title> References </title> <blist> <bibtext> Abenavoli, R. M. (2019). The mechanisms and moderators of "fade-out": Towards understanding why the skills of early childhood program participants converge over time with the skills of other children. Psychological Bulletin, 145 (12), 1103 – 1127. https://doi.org/10.1037/bul0000212</bibtext> </blist> <blist> <bibl id="bib2" idref="ref2" type="bt">2</bibl> <bibtext> Bailey, D. H., Duncan, G. J., Cunha, F., Foorman, B. R., & Yeager, D. S. (2020). Persistence and fade-out of educational-intervention effects: Mechanisms and potential solutions. Psychological Science in the Public Interest, 21 (2), 55 – 97. https://doi.org/10.1177/1529100620915848</bibtext> </blist> <blist> <bibl id="bib3" idref="ref9" type="bt">3</bibl> <bibtext> Bailey, D., Duncan, G. J., Odgers, C. L., & Yu, W. (2017). Persistence and fadeout in the impacts of child and adolescent interventions. Journal of Research on Educational Effectiveness, 10 (1), 7 – 39. https://doi.org/10.1080/19345747.2016.1232459</bibtext> </blist> <blist> <bibl id="bib4" idref="ref44" type="bt">4</bibl> <bibtext> Baker, D. P., Eslinger, P. J., Benavides, M., Peters, E., Dieckmann, N. F., & Leon, J. (2015). The cognitive impact of the education revolution: A possible cause of the Flynn Effect on population IQ. Intelligence, 49, 144 – 158. https://doi.org/10.1016/j.intell.2015.01.003</bibtext> </blist> <blist> <bibl id="bib5" idref="ref93" type="bt">5</bibl> <bibtext> Bandalos, D. L., & Finney, S. J. (2001). Item parceling issues in structural equation modeling. In G. Marcoulides & R. Schumacher (Eds.), New developments and techniques in structural equation modeling (pp. 289 – 316). Psychology Press.</bibtext> </blist> <blist> <bibl id="bib6" idref="ref122" type="bt">6</bibl> <bibtext> Bauer, D. J. (2017). A more general model for testing measurement invariance and differential item functioning. Psychological Methods, 22 (3), 507 – 526. https://doi.org/10.1037/met0000077</bibtext> </blist> <blist> <bibl id="bib7" idref="ref114" type="bt">7</bibl> <bibtext> Bentler, P. M. (1990). Comparative fit indexes in structural models. Psychological Bulletin, 107 (2), 238 – 246. https://doi.org/10.1037/0033-2909.107.2.238</bibtext> </blist> <blist> <bibl id="bib8" idref="ref68" type="bt">8</bibl> <bibtext> Campbell, F. A., & Burchinal, M. R. (2007). Early childhood interventions: The Abecedarian project. In P. C. Kyllonen, R. D. Roberts, & L. Stankov (Eds.), Extending intelligence: Enhancement and New Constructs (pp. 61 – 83). Routledge.</bibtext> </blist> <blist> <bibl id="bib9" idref="ref57" type="bt">9</bibl> <bibtext> Campbell, F. A., & Ramey, C. T. (1990). The relationship between Piagetian cognitive development, mental test performance, and academic achievement in high-risk students with and without early educational experience. Intelligence, 14 (3), 293 – 308. https://doi.org/10.1016/0160-2896(90)90020-T</bibtext> </blist> <blist> <bibtext> Campbell, F. A., & Ramey, C. T. (1995). Cognitive and school outcomes for high-risk African-American students at middle adolescence: Positive effects of early intervention. American Educational Research Journal, 32 (4), 743 – 772. https://doi.org/10.3102/00028312032004743</bibtext> </blist> <blist> <bibtext> Campbell, F. A., Pungello, E. P., Burchinal, M., Kainz, K., Pan, Y., Wasik, B. H., Barbarin, O. A., Sparling, J. J., & Ramey, C. T. (2012). Adult outcomes as a function of an early childhood educational program: An Abecedarian Project follow-up. Developmental Psychology, 48 (4), 1033 – 1043. https://doi.org/10.1037/a0026644</bibtext> </blist> <blist> <bibtext> Campbell, F. A., Ramey, C. T., Pungello, E., Sparling, J., & Miller-Johnson, S. (2002). Early childhood education: Young adult outcomes from the Abecedarian Project. Applied Developmental Science, 6 (1), 42 – 57. https://doi.org/10.1207/S1532480XADS0601_05</bibtext> </blist> <blist> <bibtext> Campbell, F., Conti, G., Heckman, J. J., Moon, S. H., Pinto, R., Pungello, E., & Pan, Y. (2014). Early childhood investments substantially boost adult health. Science, 343 (6178), 1478 – 1485. https://doi.org/10.1126/science.1248429</bibtext> </blist> <blist> <bibtext> Carroll, J. B. (1993). Human cognitive abilities: A survey of factor-analytic studies. Cambridge University Press.</bibtext> </blist> <blist> <bibtext> Chetty, R., Friedman, J. N., Hilger, N., Saez, E., Schanzenbach, D. W., & Yagan, D. (2011). How does your kindergarten classroom affect your earnings? Evidence from Project STAR. The Quarterly Journal of Economics, 126 (4), 1593 – 1660. https://doi.org/10.1093/qje/qjr041</bibtext> </blist> <blist> <bibtext> Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modeling: A Multidisciplinary Journal, 9 (2), 233 – 255. https://doi.org/10.1207/S15328007SEM0902_5</bibtext> </blist> <blist> <bibtext> Colom, R., Juanespinosa, M., & Garcia, L. (2001). The secular increase in test scores is a "Jensen effect. Personality and Individual Differences, 30 (4), 553 – 559. https://doi.org/10.1016/S0191-8869(00)00054-4</bibtext> </blist> <blist> <bibtext> Conti, G., Heckman, J. J., & Pinto, R. (2016). The effects of two influential early childhood interventions on health and healthy behaviour. Economic Journal, 126 (596), F28 – F65. https://doi.org/10.1111/ecoj.12420</bibtext> </blist> <blist> <bibtext> Cunha, F., Heckman, J. J., Lochner, L., & Masterov, D. V. (2006). Interpreting the evidence on life cycle skill formation. Handbook of the Economics of Education, 1, 697 – 812. https://doi.org/10.1016/S1574-0692(06)01012-9</bibtext> </blist> <blist> <bibtext> Deming, D. (2009). Early childhood intervention and life-cycle skill development: Evidence from Head Start. American Economic Journal: Applied Economics, 1 (3), 111 – 134. https://doi.org/10.1257/app.1.3.111</bibtext> </blist> <blist> <bibtext> DeShon, R. P. (2004). Measures are not invariant across groups without error variance homogeneity. Psychology Science, 46, 137 – 149.</bibtext> </blist> <blist> <bibtext> Dickens, W. T., & Flynn, J. R. (2001). Heritability estimates versus large environmental effects: The IQ paradox resolved. Psychological Review, 108 (2), 346 – 369. https://doi.org/10.1037/0033-295x.108.2.346</bibtext> </blist> <blist> <bibtext> Duncan, G. J., & Magnuson, K. (2011). The nature and impact of early achievement skills, attention skills, and behavior problems. In G. J. Duncan & R. Murnane (Eds.), Whither opportunity?: Rising inequality, schools, and children's life chances (pp. 47–70). Russell Sage Foundation. https://<ulink href="http://www.jstor.org/stable/">www.jstor.org/stable/</ulink> 10.7758/9781610447515</bibtext> </blist> <blist> <bibtext> Duncan, G. J., & Magnuson, K. (2013). Investing in preschool programs. The Journal of Economic Perspectives, 27 (2), 109 – 132. https://doi.org/10.1257/jep.27.2.109</bibtext> </blist> <blist> <bibtext> Elango, S., García, J. L., Heckman, J. J., & Hojman, A. (2015). Early childhood education. In Economics of means-tested transfer programs in the United States (Volume 2, pp. 235 – 297). University of Chicago Press.</bibtext> </blist> <blist> <bibtext> Enders, C. K. (2010). Applied missing data analysis. Guilford Press.</bibtext> </blist> <blist> <bibtext> Estrada, E., Ferrer, E., Abad, F. J., Román, F. J., & Colom, R. (2015). A general factor of intelligence fails to account for changes in tests' scores after cognitive practice: A longitudinal multi-group latent-variable study. Intelligence, 50, 93 – 99. https://doi.org/10.1016/j.intell.2015.02.004</bibtext> </blist> <blist> <bibtext> Faul, F., Erdfelder, E., Lang, A. G., & Buchner, A. (2007). G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39 (2), 175 – 191. https://doi.org/10.3758/bf03193146</bibtext> </blist> <blist> <bibtext> Flanagan, D. P., McGrew, K. S., & Ortiz, S. O. (2000). The Wechsler Intelligence Scales and Gf-Gc theory: A contemporary approach to interpretation. Allyn & Bacon.</bibtext> </blist> <blist> <bibtext> Flynn, J. R. (1987). Massive IQ gains in 14 nations: What IQ tests really measure. Psychological Bulletin, 101 (2), 171 – 191. https://doi.org/10.1037/0033-2909.101.2.171</bibtext> </blist> <blist> <bibtext> Flynn, J. R. (2007). What is intelligence?: Beyond the Flynn effect. Cambridge University Press.</bibtext> </blist> <blist> <bibtext> Flynn, J. R., Te Nijenhuis, J., & Metzen, D. (2014). The g beyond Spearman's g : Flynn's paradoxes resolved using four exploratory meta-analyses. Intelligence, 44, 1 – 10. https://doi.org/10.1016/j.intell.2014.01.009</bibtext> </blist> <blist> <bibtext> García, J. L., Heckman, J. J., Leaf, D. E., & Prados, M. J. (2017). Quantifying the life-cycle benefits of a prototypical early childhood program (No. w23479). National Bureau of Economic Research.</bibtext> </blist> <blist> <bibtext> Gignac, G. E. (2005). Revisiting the factor structure of the WAIS-R: Insights through nested factor modeling. Assessment, 12 (3), 320 – 329. https://doi.org/10.1177/1073191105278118</bibtext> </blist> <blist> <bibtext> Graham, J. W. (2003). Adding missing-data-relevant variables to FIML-based structural equation models. Structural Equation Modeling: A Multidisciplinary Journal, 10 (1), 80 – 100. https://doi.org/10.1207/S15328007SEM1001_4</bibtext> </blist> <blist> <bibtext> Graham, J. W. (2009). Missing data analysis: Making it work in the real world. Annual Review of Psychology, 60, 549 – 576. https://doi.org/10.1146/annurev.psych.58.110405.085530</bibtext> </blist> <blist> <bibtext> Gustafsson, J. E. (1984). A unifying model for the structure of intellectual abilities. Intelligence, 8 (3), 179 – 203. https://doi.org/10.1016/0160-2896(84)90008-4</bibtext> </blist> <blist> <bibtext> Hancock, G. R. (2001). Effect size, power, and sample size determination for structured means modeling and MIMIC approaches to between-groups hypothesis testing of means on a single latent construct. Psychometrika, 66 (3), 373 – 388. https://doi.org/10.1007/BF02294440</bibtext> </blist> <blist> <bibtext> Heckman, J. J. (2006). Skill formation and the economics of investing in disadvantaged children. Science, 312 (5782), 1900 – 1902. https://doi.org/10.1126/science.1128898</bibtext> </blist> <blist> <bibtext> Heckman, J. J., Moon, S. H., Pinto, R., Savelyev, P. A., & Yavitz, A. (2010). The rate of return to the HighScope Perry Preschool Program. Journal of Public Economics, 94 (1–2), 114 – 128. https://doi.org/10.1016/j.jpubeco.2009.11.001</bibtext> </blist> <blist> <bibtext> Heckman, J., Pinto, R., & Savelyev, P. (2013). Understanding the mechanisms through which an influential early childhood program boosted adult outcomes. American Economic Review, 103 (6), 2052 – 2086. https://doi.org/10.1257/aer.103.6.2052</bibtext> </blist> <blist> <bibtext> Hernandez, D. J. (2011). Double jeopardy: How third-grade reading skills and poverty influence high school graduation. Annie E. Casey Foundation. https://<ulink href="http://www.aecf.org/resources/doublejeopardy">www.aecf.org/resources/doublejeopardy</ulink></bibtext> </blist> <blist> <bibtext> Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6 (1), 1 – 55. [Database] https://doi.org/10.1080/10705519909540118</bibtext> </blist> <blist> <bibtext> Hu, L. T., & Bentler, P. M. (1995). Estimating model fit. In R. H. Hoyle (Ed.), Structural equation modeling: Concepts, issues, and applications (pp. 76 – 99). Sage Publications, Inc.</bibtext> </blist> <blist> <bibtext> Jensen, A. R. (1981). Raising the IQ: the Ramey and Haskins study. Intelligence, 5 (1), 29 – 40. https://doi.org/10.1016/0160-2896(81)90015-5</bibtext> </blist> <blist> <bibtext> Jensen, A. R. (1998). The g factor: The science of mental ability. Praeger.</bibtext> </blist> <blist> <bibtext> Jensen, A. R. (2002). Psychometric g : Definition and substantiation. In R. J. Sternberg & E. L. Grigorenko (Eds.), The General factor of intelligence: How general is it? (pp. 51 – 66). Psychology Press.</bibtext> </blist> <blist> <bibtext> Jensen, A. R. (1969). How much can we boost IQ and scholastic achievement? Harvard Educational Review, 39 (1), 1 – 123. https://doi.org/10.17763/haer.39.1.l3u15956627424k7</bibtext> </blist> <blist> <bibtext> Johnson, W., & Bouchard, T. J. Jr, (2005). The structure of human intelligence: It is verbal, perceptual, and image rotation (VPR), not fluid and crystallized. Intelligence, 33 (4), 393 – 416. https://doi.org/10.1016/j.intell.2004.12.002</bibtext> </blist> <blist> <bibtext> Johnson, W., Bouchard, T. J., Jr, Krueger, R. F., McGue, M., & Gottesman, I. I. (2004). Just one g : Consistent results from three test batteries. Intelligence, 32 (1), 95 – 107. https://doi.org/10.1016/S0160-2896(03)00062-X</bibtext> </blist> <blist> <bibtext> Johnson, W., Te Nijenhuis, J., & Bouchard, T. J. Jr, (2008). Still just 1 g : Consistent results from five test batteries. Intelligence, 36 (1), 81 – 95. https://doi.org/10.1016/j.intell.2007.06.001</bibtext> </blist> <blist> <bibtext> Jöreskog, K. G., & Goldberger, A. S. (1975). Estimation of a model with multiple indicators and multiple causes of a single latent variable. Journal of the American Statistical Association, 70 (351a), 631 – 639. https://doi.org/10.1080/01621459.1975.10482485</bibtext> </blist> <blist> <bibtext> Keith, T. Z., & Reynolds, M. R. (2018). Using confirmatory factor analysis to aid in understanding the constructs measured by intelligence tests. In D. P. Flanagan & E. M. McDonough (Eds.), Contemporary intellectual assessment: Theories, tests, and issues (pp. 853 – 900). The Guilford Press.</bibtext> </blist> <blist> <bibtext> Kovacs, K., & Conway, A. R. (2016). Process overlap theory: A unified account of the general factor of intelligence. Psychological Inquiry, 27 (3), 151 – 177. https://doi.org/10.1080/1047840X.2016.1153946</bibtext> </blist> <blist> <bibtext> Lemos, G. C., Abad, F. J., Almeida, L. S., & Colom, R. (2013). Sex differences on g and non-g intellectual performance reveal potential sources of STEM discrepancies. Intelligence, 41 (1), 11 – 18. https://doi.org/10.1016/j.intell.2012.10.009</bibtext> </blist> <blist> <bibtext> Little, T. D., Rhemtulla, M., Gibson, K., & Schoemann, A. M. (2013). Why the items versus parcels controversy needn't be one. Psychological Methods, 18 (3), 285 – 300. https://doi.org/10.1037/a0033266</bibtext> </blist> <blist> <bibtext> Lubke, G. H., & Dolan, C. V. (2003). Can unequal residual variances across groups mask differences in residual means in the common factor model? Structural Equation Modeling, 10 (2), 175 – 192. https://doi.org/10.1207/S15328007SEM1002_1</bibtext> </blist> <blist> <bibtext> Lubke, G. H., Dolan, C. V., Kelderman, H., & Mellenbergh, G. J. (2003). On the relationship between sources of within-and between-group differences and measurement invariance in the common factor model. Intelligence, 31 (6), 543 – 566. https://doi.org/10.1016/S0160-2896(03)00051-5</bibtext> </blist> <blist> <bibtext> Marsh, H. W., Lüdtke, O., Nagengast, B., Morin, A. J., & Von Davier, M. (2013). Why item parcels are (almost) never appropriate: Two wrongs do not make a right-camouflaging misspecification with item parcels in CFA models. Psychological Methods, 18 (3), 257 – 284. https://doi.org/10.1037/a0032773</bibtext> </blist> <blist> <bibtext> McCarton, C. M., Brooks-Gunn, J., Wallace, I. F., Bauer, C. R., Bennett, F. C., Bernbaum, J. C., & Tyson, J. (1997). Results at age 8 years of early intervention for low-birth-weight premature infants: The Infant Health and Development Program. JAMA, 277 (2), 126 – 132. https://doi.org/10.1001/jama.1997.03540260040033</bibtext> </blist> <blist> <bibtext> McCoy, D. C., Yoshikawa, H., Ziol-Guest, K. M., Duncan, G. J., Schindler, H. S., Magnuson, K., Yang, R., Koepp, A., & Shonkoff, J. P. (2017). Impacts of early childhood education on medium- and long-term educational outcomes. Educational Researcher, 46 (8), 474 – 487. https://doi.org/10.3102/0013189X17737739</bibtext> </blist> <blist> <bibtext> McDonald, R. P. (1999). Test theory: A unified treatment. Lawrence Erlbaum Associates Publishers.</bibtext> </blist> <blist> <bibtext> Meade, A. W., Johnson, E. C., & Braddy, P. W. (2008). Power and sensitivity of alternative fit indices in tests of measurement invariance. Journal of Applied Psychology, 93 (3), 568 – 592. https://doi.org/10.1037/0021-9010.93.3.568</bibtext> </blist> <blist> <bibtext> Meredith, W. (1993). Measurement invariance, factor analysis and factorial invariance. Psychometrika, 58 (4), 525 – 543. https://doi.org/10.1007/BF02294825</bibtext> </blist> <blist> <bibtext> Millsap, R. E. (2011). Statistical approaches to measurement invariance. Routledge.</bibtext> </blist> <blist> <bibtext> Muennig, P., Robertson, D., Johnson, G., Campbell, F., Pungello, E. P., & Neidell, M. (2011). The effect of an early education program on adult health: The Carolina Abecedarian Project randomized controlled trial. American Journal of Public Health, 101 (3), 512–516. https://doi.org/10.2105/AJPH.2010.200063</bibtext> </blist> <blist> <bibtext> Muthén, B. O. (1989). Latent variable modeling in heterogeneous populations. Psychometrika, 54 (4), 557 – 585. https://doi.org/10.1007/BF02296397</bibtext> </blist> <blist> <bibtext> Muthén, L. K., & Muthén, B. O. (2017). Mplus User's Guide. Eighth Edition. Muthén & Muthén.</bibtext> </blist> <blist> <bibtext> Pages, R., Lukes, D. J., Bailey, D. H., & Duncan, G. J. (2020). Elusive longer-run impacts of head start: Replications within and across cohorts. Educational Evaluation and Policy Analysis, 42 (4), 471 – 492. https://doi.org/10.3102/0162373720948884</bibtext> </blist> <blist> <bibtext> Philips, D. A., Lipsey, M. W., Dodge, K. A., Haskins, R., Bassok, D., Burchinal, M. R., Duncan, G. J., Dynarski, M., Magnuson, K. A., & Weilans, C. (2017). Puzzling it out: The current state of scientific knowledge on pre-kindergarten effects. Brookings Institution.</bibtext> </blist> <blist> <bibtext> Protzko, J. (2015). The environment in raising early intelligence: A meta-analysis of the fadeout effect. Intelligence, 53, 202 – 210. https://doi.org/10.1016/j.intell.2015.10.006</bibtext> </blist> <blist> <bibtext> Protzko, J. (2016). Does the raising IQ-raising g distinction explain the fadeout effect? Intelligence, 56, 65 – 71. https://doi.org/10.1016/j.intell.2016.02.008</bibtext> </blist> <blist> <bibtext> Protzko, J. (2017). Raising IQ among school-aged children: Five meta-analyses and a review of randomized controlled trials. Developmental Review, 46, 81 – 101. https://doi.org/10.1016/j.dr.2017.05.001</bibtext> </blist> <blist> <bibtext> Ramey, C. T. (1992). High-risk children and IQ: Altering intergenerational patterns. Intelligence, 16 (2), 239 – 256. https://doi.org/10.1016/0160-2896(92)90007-E</bibtext> </blist> <blist> <bibtext> Ramey, C. T. (2018). The Abecedarian approach to social, educational, and health disparities. Clinical Child and Family Psychology Review, 21 (4), 527 – 544. https://doi.org/10.1007/s10567-018-0260-y</bibtext> </blist> <blist> <bibtext> Ramey, C. T., Bryant, D. M., Wasik, B. H., Sparling, J. J., Fendt, K. H. L., & Vange, L. M. (1992). Infant Health and Development Program for low birth weight, premature infants: Program elements, family participation, and child intelligence. Pediatrics, 89 (3), 454 – 465.</bibtext> </blist> <blist> <bibtext> Ramey, C. T., Collier, A. M., Sparling, J. J., Loda, F. A., Campbell, F. A., Ingram, D. A., & Finkelstein, N. W. (1976). The Carolina Abecedarian Project: A longitudinal and multidisciplinary approach to the prevention of developmental retardation. In T. Tjossem (Ed.), Intervention strategies for high-risk infants and young children (pp. 629 – 655). University Park Press. (Original work published 1974)</bibtext> </blist> <blist> <bibtext> Ramey, C. T., MacPhee, D., & Yeates, K. O. (1982). Preventing developmental retardation: A general systems model. In L. A. Bond & J. M. Joffee (Eds.), Facilitating infant and early childhood development (pp. 343–401)</bibtext> </blist> <blist> <bibtext> Ramey, C. T., Campbell, F. A., Burchinal, M., Skinner, M. L., Gardner, D. M., & Ramey, S. L. (2000). Persistent effects of early childhood education on high-risk children and their mothers. Applied Developmental Science, 4 (1), 2–14. https://doi.org/10.1207/S1532480XADS0401_1</bibtext> </blist> <blist> <bibtext> Raykov, T., & Marcoulides, G. A. (2016). Scale reliability evaluation under multiple assumption violations. Structural Equation Modeling: A Multidisciplinary Journal, 23 (2), 302 – 313. https://doi.org/10.1080/10705511.2014.938597</bibtext> </blist> <blist> <bibtext> Reynolds, M. R., & Keith, T. Z. (2017). Multi-group and hierarchical confirmatory factor analysis of the Wechsler Intelligence Scale for Children – Fifth Edition: What does it measure ? Intelligence, 62, 31 – 47. https://doi.org/10.1016/j.intell.2017.02.005</bibtext> </blist> <blist> <bibtext> Ritchie, S. J., & Tucker-Drob, E. M. (2018). How much does education improve intelligence? A meta-analysis. Psychological Science, 29 (8), 1358 – 1369. https://doi.org/10.1177/0956797618774253</bibtext> </blist> <blist> <bibtext> Ritchie, S. J., Bates, T. C., & Deary, I. J. (2015). Is education associated with improvements in general cognitive ability, or in specific skills? Developmental Psychology, 51 (5), 573 – 582. https://doi.org/10.1037/a0038981</bibtext> </blist> <blist> <bibtext> Schweinhart, L. J., Barnes, H. V., & Weikart, D. P. (1993). Significant benefits: The High/Scope Perry Preschool Study through Age 27. Monographs of the High/Scope Educational Research Foundation, No. Ten. High/Scope Educational Research Foundation, 600 North River Street, Ypsilanti, MI 48198 – 2898.</bibtext> </blist> <blist> <bibtext> Sörbom, D. (1989). Model modification. Psychometrika, 54 (3), 371 – 384. https://doi.org/10.1007/BF02294623</bibtext> </blist> <blist> <bibtext> Sparling, J., & Lewis, I. (1979). Learning games for the first three years: A guide to parent/child play. Walker and Company.</bibtext> </blist> <blist> <bibtext> Sparling, J., & Lewis, I. (1984). Learning games for threes and fours. Walker and Company.</bibtext> </blist> <blist> <bibtext> Sparling, J., & Meunier, K. (2019). Abecedarian: An early childhood education approach that has a rich history and a vibrant present. International Journal of Early Childhood, 51 (2), 207 – 216. https://doi.org/10.1007/s13158-019-00247-2</bibtext> </blist> <blist> <bibtext> Sparling, J., Ramey, C. T., & Ramey, S. L. (2007). The Abecedarian experience. In M. E. Young &, L. M. Richardson (Eds.), Early child development from measurement to action (pp.103–127). The World Bank.</bibtext> </blist> <blist> <bibtext> Suggate, S. P. (2016). A meta-analysis of the long-term effects of phonemic awareness, phonics, fluency, and reading comprehension interventions. Journal of Learning Disabilities, 49 (1), 77 – 96. https://doi.org/10.1177/0022219414528540</bibtext> </blist> <blist> <bibtext> Te Nijenhuis, J., & van der Flier, H. (2013). Is the Flynn effect on g ?: A meta-analysis. Intelligence, 41 (6), 802 – 807. https://doi.org/10.1016/j.intell.2013.03.001</bibtext> </blist> <blist> <bibtext> Te Nijenhuis, J., Jongeneel-Grimen, B., & Kirkegaard, E. O. (2014). Are Head Start gains on the g factor? A meta-analysis. Intelligence, 46, 209 – 215. https://doi.org/10.1016/j.intell.2014.07.001</bibtext> </blist> <blist> <bibtext> Te Nijenhuis, J., van Vianen, A. E., & van der Flier, H. (2007). Score gains on g -loaded tests: No g. Intelligence, 35 (3), 283 – 300. https://doi.org/10.1016/j.intell.2006.07.006</bibtext> </blist> <blist> <bibtext> Van Der Maas, H. L. J., Dolan, C. V., Grasman, R. P. P. P., Wicherts, J. M., Huizenga, H. M., & Raijmakers, M. E. J. (2006). A dynamical model of general intelligence: The positive manifold of intelligence by mutualism. Psychological Review, 113 (4), 842 – 861. https://doi.org/10.1037/0033-295X.113.4.842</bibtext> </blist> <blist> <bibtext> Vernon, P. E. (2014). The structure of human abilities (psychology revivals). Routledge.</bibtext> </blist> <blist> <bibtext> Watts, T. W., Gandhi, J., Ibrahim, D. A., Masucci, M. D., & Raver, C. C. (2018). The Chicago school readiness project: Examining the long-term impacts of an early childhood intervention. PLoS One, 13 (7), e0200144. https://doi.org/10.1371/journal.pone.0200144</bibtext> </blist> <blist> <bibtext> Wechsler, D. (1967). Manual for the Wechsler preschool and primary scale of intelligence. Psychological Corporation.</bibtext> </blist> <blist> <bibtext> Wechsler, D. (1974). Manual for the Wechsler intelligence scale for children, revised. Psychological Corporation.</bibtext> </blist> <blist> <bibtext> Wechsler, D. (1981). WAIS-R manual: Wechsler adult intelligence scale-revised. Psychological Corporation.</bibtext> </blist> <blist> <bibtext> Whitehurst, G. J., Arnold, D. S., Epstein, J. N., Angell, A. L., Smith, M., & Fischel, J. E. (1994). A picture book reading intervention in day care and home for children from low-income families. Developmental Psychology, 30 (5), 679. https://doi.org/10.1037/0012-1649.30.5.679</bibtext> </blist> <blist> <bibtext> Wicherts, J. M., & Dolan, C. V. (2010). Measurement invariance in confirmatory factor analysis: An illustration using IQ test performance of minorities. Educational Measurement: Issues and Practice, 29 (3), 39 – 47. https://doi.org/10.1111/j.1745-3992.2010.00182.x</bibtext> </blist> <blist> <bibtext> Wicherts, J. M., Dolan, C. V., Hessen, D. J., Oosterveld, P., Van Baal, G. C. M., Boomsma, D. I., & Span, M. M. (2004). Are intelligence tests measurement invariant over time? Investigating the nature of the Flynn effect. Intelligence, 32 (5), 509 – 537. https://doi.org/10.1016/j.intell.2004.07.002</bibtext> </blist> <blist> <bibtext> Yoshikawa, H., Weiland, C., Brooks-Gunn, J., Burchinal, M. R., Espinosa, L. M., Gormley, W. T., Ludwig, J., Magnuson, K. A., Philips, D., & Zaslow, M. J. (2013). Investing in our future: The evidence base on preschool education. Society for Research in Child Development. https://<ulink href="http://www.srcd.org/event/too-much-evidence-ignore-new-findings-impact-quality-preschool-scale">www.srcd.org/event/too-much-evidence-ignore-new-findings-impact-quality-preschool-scale</ulink></bibtext> </blist> </ref> <aug> <p>By Remy Pages; John Protzko and Drew H. Bailey</p> <p>Reported by Author; Author; Author</p> </aug> <nolink nlid="nl1" bibid="bib24" firstref="ref3"></nolink> <nolink nlid="nl2" bibid="bib25" firstref="ref4"></nolink> <nolink nlid="nl3" bibid="bib61" firstref="ref5"></nolink> <nolink nlid="nl4" bibid="bib69" firstref="ref6"></nolink> <nolink nlid="nl5" bibid="bib96" firstref="ref7"></nolink> <nolink nlid="nl6" bibid="bib19" firstref="ref10"></nolink> <nolink nlid="nl7" bibid="bib23" firstref="ref11"></nolink> <nolink nlid="nl8" bibid="bib39" firstref="ref12"></nolink> <nolink nlid="nl9" bibid="bib72" firstref="ref13"></nolink> <nolink nlid="nl10" bibid="bib20" firstref="ref14"></nolink> <nolink nlid="nl11" bibid="bib40" firstref="ref15"></nolink> <nolink nlid="nl12" bibid="bib103" firstref="ref16"></nolink> <nolink nlid="nl13" bibid="bib15" firstref="ref17"></nolink> <nolink nlid="nl14" bibid="bib41" firstref="ref18"></nolink> <nolink nlid="nl15" bibid="bib71" firstref="ref19"></nolink> <nolink nlid="nl16" bibid="bib46" firstref="ref21"></nolink> <nolink nlid="nl17" bibid="bib92" firstref="ref22"></nolink> <nolink nlid="nl18" bibid="bib48" firstref="ref24"></nolink> <nolink nlid="nl19" bibid="bib47" firstref="ref25"></nolink> <nolink nlid="nl20" bibid="bib54" firstref="ref27"></nolink> <nolink nlid="nl21" bibid="bib94" firstref="ref29"></nolink> <nolink nlid="nl22" bibid="bib14" firstref="ref30"></nolink> <nolink nlid="nl23" bibid="bib37" firstref="ref31"></nolink> <nolink nlid="nl24" bibid="bib49" firstref="ref32"></nolink> <nolink nlid="nl25" bibid="bib95" firstref="ref33"></nolink> <nolink nlid="nl26" bibid="bib93" firstref="ref35"></nolink> <nolink nlid="nl27" bibid="bib73" firstref="ref38"></nolink> <nolink nlid="nl28" bibid="bib27" firstref="ref39"></nolink> <nolink nlid="nl29" bibid="bib83" firstref="ref42"></nolink> <nolink nlid="nl30" bibid="bib82" firstref="ref43"></nolink> <nolink nlid="nl31" bibid="bib17" firstref="ref45"></nolink> <nolink nlid="nl32" bibid="bib30" firstref="ref46"></nolink> <nolink nlid="nl33" bibid="bib31" firstref="ref47"></nolink> <nolink nlid="nl34" bibid="bib32" firstref="ref48"></nolink> <nolink nlid="nl35" bibid="bib91" firstref="ref49"></nolink> <nolink nlid="nl36" bibid="bib102" firstref="ref50"></nolink> <nolink nlid="nl37" bibid="bib11" firstref="ref52"></nolink> <nolink nlid="nl38" bibid="bib33" firstref="ref53"></nolink> <nolink nlid="nl39" bibid="bib77" firstref="ref54"></nolink> <nolink nlid="nl40" bibid="bib78" firstref="ref55"></nolink> <nolink nlid="nl41" bibid="bib12" firstref="ref56"></nolink> <nolink nlid="nl42" bibid="bib10" firstref="ref58"></nolink> <nolink nlid="nl43" bibid="bib45" firstref="ref59"></nolink> <nolink nlid="nl44" bibid="bib75" firstref="ref61"></nolink> <nolink nlid="nl45" bibid="bib79" firstref="ref62"></nolink> <nolink nlid="nl46" bibid="bib74" firstref="ref65"></nolink> <nolink nlid="nl47" bibid="bib18" firstref="ref70"></nolink> <nolink nlid="nl48" bibid="bib86" firstref="ref76"></nolink> <nolink nlid="nl49" bibid="bib87" firstref="ref77"></nolink> <nolink nlid="nl50" bibid="bib89" firstref="ref78"></nolink> <nolink nlid="nl51" bibid="bib88" firstref="ref79"></nolink> <nolink nlid="nl52" bibid="bib97" firstref="ref82"></nolink> <nolink nlid="nl53" bibid="bib98" firstref="ref83"></nolink> <nolink nlid="nl54" bibid="bib99" firstref="ref84"></nolink> <nolink nlid="nl55" bibid="bib29" firstref="ref85"></nolink> <nolink nlid="nl56" bibid="bib34" firstref="ref89"></nolink> <nolink nlid="nl57" bibid="bib81" firstref="ref90"></nolink> <nolink nlid="nl58" bibid="bib53" firstref="ref92"></nolink> <nolink nlid="nl59" bibid="bib56" firstref="ref94"></nolink> <nolink nlid="nl60" bibid="bib59" firstref="ref95"></nolink> <nolink nlid="nl61" bibid="bib50" firstref="ref96"></nolink> <nolink nlid="nl62" bibid="bib51" firstref="ref97"></nolink> <nolink nlid="nl63" bibid="bib62" firstref="ref98"></nolink> <nolink nlid="nl64" bibid="bib80" firstref="ref99"></nolink> <nolink nlid="nl65" bibid="bib68" firstref="ref100"></nolink> <nolink nlid="nl66" bibid="bib26" firstref="ref101"></nolink> <nolink nlid="nl67" bibid="bib36" firstref="ref102"></nolink> <nolink nlid="nl68" bibid="bib35" firstref="ref103"></nolink> <nolink nlid="nl69" bibid="bib65" firstref="ref104"></nolink> <nolink nlid="nl70" bibid="bib58" firstref="ref105"></nolink> <nolink nlid="nl71" bibid="bib55" firstref="ref107"></nolink> <nolink nlid="nl72" bibid="bib101" firstref="ref108"></nolink> <nolink nlid="nl73" bibid="bib21" firstref="ref109"></nolink> <nolink nlid="nl74" bibid="bib57" firstref="ref111"></nolink> <nolink nlid="nl75" bibid="bib64" firstref="ref113"></nolink> <nolink nlid="nl76" bibid="bib44" firstref="ref115"></nolink> <nolink nlid="nl77" bibid="bib16" firstref="ref116"></nolink> <nolink nlid="nl78" bibid="bib43" firstref="ref117"></nolink> <nolink nlid="nl79" bibid="bib85" firstref="ref118"></nolink> <nolink nlid="nl80" bibid="bib28" firstref="ref119"></nolink> <nolink nlid="nl81" bibid="bib63" firstref="ref120"></nolink> <nolink nlid="nl82" bibid="bib52" firstref="ref121"></nolink> <nolink nlid="nl83" bibid="bib38" firstref="ref123"></nolink> <nolink nlid="nl84" bibid="bib67" firstref="ref124"></nolink> <nolink nlid="nl85" bibid="bib22" firstref="ref135"></nolink> <nolink nlid="nl86" bibid="bib42" firstref="ref139"></nolink> <nolink nlid="nl87" bibid="bib90" firstref="ref141"></nolink> <nolink nlid="nl88" bibid="bib100" firstref="ref142"></nolink> <nolink nlid="nl89" bibid="bib13" firstref="ref143"></nolink> <nolink nlid="nl90" bibid="bib66" firstref="ref144"></nolink> <nolink nlid="nl91" bibid="bib84" firstref="ref145"></nolink> <nolink nlid="nl92" bibid="bib76" firstref="ref146"></nolink> <nolink nlid="nl93" bibid="bib60" firstref="ref148"></nolink> <nolink nlid="nl94" bibid="bib70" firstref="ref151"></nolink>
Header DbId: eric
DbLabel: ERIC
An: EJ1349873
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: The Breadth of Impacts from the Abecedarian Project Early Intervention on Cognitive Skills
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Pages%2C+Remy%22">Pages, Remy</searchLink> (ORCID <externalLink term="http://orcid.org/0000-0002-0396-6498">0000-0002-0396-6498</externalLink>)<br /><searchLink fieldCode="AR" term="%22Protzko%2C+John%22">Protzko, John</searchLink><br /><searchLink fieldCode="AR" term="%22Bailey%2C+Drew+H%2E%22">Bailey, Drew H.</searchLink>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="SO" term="%22Journal+of+Research+on+Educational+Effectiveness%22"><i>Journal of Research on Educational Effectiveness</i></searchLink>. 2022 15(2):243-262.
– Name: Avail
  Label: Availability
  Group: Avail
  Data: Routledge. Available from: Taylor & Francis, Ltd. 530 Walnut Street Suite 850, Philadelphia, PA 19106. Tel: 800-354-1420; Tel: 215-625-8900; Fax: 215-207-0050; Web site: http://www.tandf.co.uk/journals
– Name: PeerReviewed
  Label: Peer Reviewed
  Group: SrcInfo
  Data: Y
– Name: Pages
  Label: Page Count
  Group: Src
  Data: 20
– Name: DatePubCY
  Label: Publication Date
  Group: Date
  Data: 2022
– Name: TypeDocument
  Label: Document Type
  Group: TypDoc
  Data: Journal Articles<br />Reports - Research
– Name: Audience
  Label: Education Level
  Group: Audnce
  Data: <searchLink fieldCode="EL" term="%22Early+Childhood+Education%22">Early Childhood Education</searchLink>
– Name: Subject
  Label: Descriptors
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Early+Intervention%22">Early Intervention</searchLink><br /><searchLink fieldCode="DE" term="%22Cognitive+Ability%22">Cognitive Ability</searchLink><br /><searchLink fieldCode="DE" term="%22Intelligence+Quotient%22">Intelligence Quotient</searchLink><br /><searchLink fieldCode="DE" term="%22Children%22">Children</searchLink><br /><searchLink fieldCode="DE" term="%22Adolescents%22">Adolescents</searchLink><br /><searchLink fieldCode="DE" term="%22Young+Adults%22">Young Adults</searchLink><br /><searchLink fieldCode="DE" term="%22Early+Childhood+Education%22">Early Childhood Education</searchLink><br /><searchLink fieldCode="DE" term="%22Disadvantaged%22">Disadvantaged</searchLink><br /><searchLink fieldCode="DE" term="%22Program+Effectiveness%22">Program Effectiveness</searchLink><br /><searchLink fieldCode="DE" term="%22Thinking+Skills%22">Thinking Skills</searchLink>
– Name: Subject
  Label: Geographic Terms
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22North+Carolina%22">North Carolina</searchLink>
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1080/19345747.2021.1969711
– Name: ISSN
  Label: ISSN
  Group: ISSN
  Data: 1934-5747<br />1934-5739
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: Early life interventions impacting cognitive abilities are most often followed by post-treatment fadeout. Some have hypothesized that persistence is unlikely when gains are specific to trained skills and distinguishable from impacts on general cognitive ability (classically modeled as a hierarchical factor, so-called psychometric g). Using measurement invariance testing and multiple-indicators multiple-causes models, we investigated impacts on IQ subtests from the Abecedarian early childhood intervention (n = 107). We found that (1) observed impacts on IQ scores from age 5 to age 21 were consistent with persistent positive effects on g; (2) subtest-specific variance that was differentiable from changes on g did fade. Together, these findings indicated that Abecedarian early impact persisted across a range of cognitive skills, providing some evidence for the hypothesis that breadth and persistence of impacts from educational interventions are related.
– Name: AbstractInfo
  Label: Abstractor
  Group: Ab
  Data: As Provided
– Name: DateEntry
  Label: Entry Date
  Group: Date
  Data: 2022
– Name: AN
  Label: Accession Number
  Group: ID
  Data: EJ1349873
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ1349873
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1080/19345747.2021.1969711
    Languages:
      – Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 20
        StartPage: 243
    Subjects:
      – SubjectFull: Early Intervention
        Type: general
      – SubjectFull: Cognitive Ability
        Type: general
      – SubjectFull: Intelligence Quotient
        Type: general
      – SubjectFull: Children
        Type: general
      – SubjectFull: Adolescents
        Type: general
      – SubjectFull: Young Adults
        Type: general
      – SubjectFull: Early Childhood Education
        Type: general
      – SubjectFull: Disadvantaged
        Type: general
      – SubjectFull: Program Effectiveness
        Type: general
      – SubjectFull: Thinking Skills
        Type: general
      – SubjectFull: North Carolina
        Type: general
    Titles:
      – TitleFull: The Breadth of Impacts from the Abecedarian Project Early Intervention on Cognitive Skills
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Pages, Remy
      – PersonEntity:
          Name:
            NameFull: Protzko, John
      – PersonEntity:
          Name:
            NameFull: Bailey, Drew H.
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 01
              Type: published
              Y: 2022
          Identifiers:
            – Type: issn-print
              Value: 1934-5747
            – Type: issn-electronic
              Value: 1934-5739
          Numbering:
            – Type: volume
              Value: 15
            – Type: issue
              Value: 2
          Titles:
            – TitleFull: Journal of Research on Educational Effectiveness
              Type: main
ResultId 1