Skip to main content

Comparative Effectiveness of Early Childhood Interventions (public)

Research reportView only

Loading... AlertsChat

September 12, 2025

What are the comparative effect sizes of early childhood educational interventions across different age groups and learning domains?

Early childhood educational interventions demonstrate varying effectiveness across domains, producing effect sizes of 0.10-1.45 standard deviations for cognitive outcomes, 0.18-1.22 for mathematics, 0.10-0.44 for language (up to 1.10 for literacy), and 0.19-0.56 for socioemotional and motor development, with comparable benefits observed across preschool and infant age groups.

Abstract

Early childhood educational interventions yield effect sizes that vary by learning domain, intervention modality, and age group. Studies targeting cognitive outcomes report gains ranging from 0.10 to 1.45 standard deviations—with intensive, multi-year programs (e.g., Ramey and Ramey, 2023) achieving the largest benefits—while language-focused programs generally produce improvements of 0.10 to 0.44 standard deviations, with some literacy interventions reaching 1.10. Mathematics interventions typically exhibit effects from 0.18 to 1.22 standard deviations, especially when curriculum-based and structured (as in Yıldız et al., 2025 and Clements et al., 2011). Socioemotional and motor development interventions tend to yield modest gains (approximately 0.19–0.56), a range further supported by evaluations of responsive caregiving and physical activity programs. Most studies concentrate on preschool-aged children, though interventions for infants show comparable benefits. Greater intervention intensity, duration, and multi-component design are associated with larger effects, according to the reported findings.

Methods

We analyzed 40 sources from an initial pool of 997, using 8 screening criteria. Each paper was reviewed for 6 key aspects that mattered most to the research question. More on methods

Papers identified with Elicit search

n = 997

Papers screened using: Population Age Range, Quantitative Data Availability, Study Design Rigor, Educational Setting, Learning Domain Focus, Typical Development Population, Intervention Component, Educational vs. Medical Focus

n = 997

Papers screened out

n = 957

Papers included for extraction

n = 40

Press enter or space to select a node.You can then use the arrow keys to move the node around. Press delete to remove it and escape to cancel.

Press enter or space to select an edge. You can then press delete to remove it or escape to cancel.

Paper search

Using your research question “What are the comparative effect sizes of early childhood educational interventions across different age groups and learning domains?”, we searched across over 126 million academic papers from the Semantic Scholar corpus. We retrieved the 997 papers most relevant to the query.

Screening

We screened in sources based on their abstracts that met these criteria:

• Population Age Range: Does the study focus on children aged 0-8 years (early childhood period)?
• Quantitative Data Availability: Does the study report quantitative effect sizes or provide sufficient quantitative data to calculate effect sizes?
• Study Design Rigor: Is the study a randomized controlled trial, quasi-experimental study, or controlled study with comparison groups?
• Educational Setting: Is the intervention implemented in early childhood educational settings (e.g., preschools, kindergartens, childcare centers, or home-based educational programs)?
• Learning Domain Focus: Does the intervention target specific learning domains (cognitive, language, literacy, numeracy, social-emotional, or motor skill development)?
• Typical Development Population: Does the study include typically developing children (rather than focusing solely on children with diagnosed developmental disabilities or special needs)?
• Intervention Component: Does the study include an intervention component (rather than being purely observational)?
• Educational vs. Medical Focus: Is the intervention primarily educational in nature (rather than primarily focused on medical or therapeutic treatments)?

We considered all screening questions together and made a holistic judgement about whether to screen in each paper.

Data extraction

We asked a large language model to extract each data column below from each paper. We gave the model the extraction instructions shown below for each column.

• Study Design Type:

Identify the specific type of study design used. Look in the methods section for precise description. Categorize as:

• Randomized Controlled Trial (RCT)
• Quasi-experimental study
• Comparative study
• Longitudinal study

If multiple design elements are present, list all that apply. If uncertain, note “unclear” and provide a brief explanation of why.

• Intervention Type and Components:

Comprehensively list all components of the educational intervention:

• Primary pedagogical approach (e.g., direct instruction, inquiry-based)
• Specific learning domains targeted
• Additional support services (if any)
• Delivery method (group size, individual/group)

Be as detailed as possible. If multiple intervention components exist, list each separately. Include frequency, duration, and specific content of interventions where mentioned.

• Participant Demographics:

Extract the following participant details:

• Age range of children
• Socioeconomic status (if reported)
• Risk factors or special circumstances
• Total sample size
• Gender distribution (if reported)

Use exact numbers or percentages from the study. If a range is given, record the full range. If any demographic information is not reported, clearly mark as “NR” (not reported).

• Outcome Domains and Measurement:

List all primary outcome domains measured:

• Cognitive development
• Language development
• Motor development
• Socioemotional development
• Academic achievement
• Other specific domains

For each domain:

• Note specific measurement tools/instruments used
• Record effect sizes if provided
• Indicate statistical significance
• Note follow-up time points for measurements

If multiple measurements exist for a domain, list all with their respective details.

• Long-term Outcomes:

Identify and extract any long-term outcomes measured beyond immediate child development:

• Educational attainment
• Employment
• Social relationships
• Health indicators
• Economic outcomes

Record:

• Specific outcome measured

• Age/time point of measurement

• Effect size or key findings

• Statistical significance (if reported)

• Geographical and Contextual Factors:

Extract:

• Country or region of study
• Income level classification (low, middle, high-income)
• Urban/rural setting
• Specific contextual challenges or environmental factors

If multiple contexts are represented, list each separately. Use the study’s own classification or standard international definitions for income levels.

Results

Characteristics of Included Studies

Study

Intervention Type

Age Group

Learning Domain

Study Design

Full text retrieved

Love et al., 2013

Direct coaching, video modeling, self-reflection, resource/referral services

Prenatal to 5 years

Cognitive, language, attention, behavior, health, parenting, mental health, employment

Randomized controlled trial (RCT), Longitudinal, Comparative

Yes

Meghir et al., 2023

Early stimulation (home-based), enhanced preschool (center-based)

7-16 months to ~4 years

Cognition, language, executive function, school readiness

Randomized controlled trial (RCT), Longitudinal

No

Olive et al., 2023

Physical activity intervention (AEL)

3-5 years

Executive function, expressive vocabulary, motor skills

Randomized controlled trial (RCT)

Yes

Camilli et al., 2010

Direct instruction, inquiry-based, teacher/small-group

Pre-Kindergarten

Cognitive, social, school progress

Comparative, Quasi-experimental, Randomized controlled trial (RCT)

No

Campbell et al., 2001

Full-time, high-quality educational child care

Infancy to 21 years

Cognitive, academic (reading, math)

Randomized controlled trial (RCT), Longitudinal

No

Campbell and Ramey, 1994

Infant/preschool/primary school educational treatment

Infancy to 12 years

Cognitive, academic achievement

Randomized controlled trial (RCT), Longitudinal

Yes

Reynolds et al., 2001

Comprehensive education, family, health services

3-9 years, followed to 20

Academic, social, educational attainment, crime

Longitudinal, Quasi-experimental, Comparative

Yes

Blewitt et al., 2018

Universal curriculum-based social-emotional learning (SEL)

2-6 years

Social, emotional, behavioral, early learning

Randomized controlled trial (RCT), Quasi-experimental

Yes

Jeong et al., 2021

Parenting interventions (responsive caregiving, stimulation, etc.)

Prenatal to 3 years

Cognitive, language, motor, socioemotional, behavior, attachment

Randomized controlled trial (RCT)

Yes

Cahoon et al., 2023

Home-based literacy/math interventions

3.07–5.32 years

Literacy, mathematics

Randomized controlled trial (RCT), Quasi-experimental

Yes

toof

Pageof

View 30 more rows

Intervention Types:

• Curriculum-based interventions:13 studies (including literacy, math, social-emotional learning, spatial skills, etc.).
• Center-based early childhood education/preschool programs:10 studies.
• Home-based or parenting interventions:7 studies.
• Multi-component interventions (education plus health, nutrition, or social services):6 studies.
• Physical or motor skill interventions:3 studies.
• Technology-based interventions (educational apps, computer-supported learning):3 studies.
• Policy or large-scale program evaluations:3 studies.

Age Groups:

• 25 studies targeted preschool-aged children (3-5/6 years).
• 15 studies included prenatal or infancy (0-2 years).
• 5 studies included early primary ages (6-8 years).
• 7 studies followed participants into adolescence or adulthood.
• We didn’t find mention of the age group in 7 studies.

Learning Domains:

• Language and literacy outcomes: 22 studies.
• Cognitive or academic outcomes: 20 studies.
• Math outcomes: 14 studies.
• Social-emotional or behavioral outcomes: 15 studies.
• Motor or physical development outcomes: 5 studies.
• Health outcomes: 4 studies.
• Economic or employment outcomes: 5 studies.

Study Designs:

• 26 studies used randomized controlled trial (RCT) designs.
• 14 studies used quasi-experimental designs.
• 10 studies used comparative designs.
• 17 studies included longitudinal follow-up.
• 6 studies were systematic reviews or meta-analyses.

Effects

Effect Sizes by Age Group

Study

Age Group

Learning Domain

Effect Size Range

Number of Studies

Love et al., 2013

1-5 years

Cognitive, language, socioemotional

0.10–0.20 (cognitive/language), 0.31–0.38 (engagement)

1

Meghir et al., 2023

7-48 months

IQ, school readiness

0.13–0.24 standard deviations

1

Olive et al., 2023

3-5 years

Executive function, vocabulary

Cohen’s d=0.24–0.29

1

Camilli et al., 2010

Pre-Kindergarten

Cognitive, social

Largest for cognitive, no mention found

1

Campbell et al., 2001

3-21 years

Cognitive, academic

Moderate-large, no mention found

1

Campbell and Ramey, 1994

Infancy-12 years

Cognitive, academic

Maintained advantage, no mention found

1

Reynolds et al., 2001

3-20 years

Educational, social

11.2% increase in high school completion, 8.2% decrease in arrest

1

Blewitt et al., 2018

2-6 years

Socioemotional, cognitive

Cohen’s d=0.18–0.54

1

Jeong et al., 2021

0-3 years

Cognitive, language, motor, socioemotional

Standardized mean difference (SMD)=0.19–0.56

1

Cahoon et al., 2023

3-5 years

Literacy, math

Cohen’s d=0.10 (literacy), 0.18 (math)

1

toof

Pageof

View 30 more rows

Learning Domains Targeted:

• Mathematics: 13 studies.
• Cognitive outcomes: 12 studies.
• Language outcomes: 11 studies.
• Socioemotional or emotional outcomes: 6 studies.
• Academic or educational outcomes: 5 studies.
• Literacy and reading outcomes: 5 studies (3 literacy, 5 reading; some overlap).
• Other domains: executive function (2 studies), motor/physical/fine motor skills (4 studies), health (2 studies), social (3 studies), behavior (2 studies), school readiness (1 study), problem-solving/STEM (1 study), spatial skills (1 study), phonological/alphabet (4 studies), pre-literacy (1 study).

Effect Size Reporting:

• Numeric effect size estimates (e.g., standard deviation, Cohen’s d, Hedges’ g, standardized mean difference, correlation coefficient, or percentage) were found in 26 studies.
• 13 studies reported only qualitative descriptors of effect (e.g., “significant”, “modest”, “moderate-large”).
• Of the 26 studies with numeric effect sizes, 2 reported only percentage or non-standard deviation outcomes (e.g., percentage increase in school completion).
• All included studies reported at least a qualitative or numeric effect size.

Age Groups:

• 16 studies targeted preschool or pre-Kindergarten children (3-5/6 years).
• 6 studies targeted infants (0-2 years).
• 3 studies targeted early childhood (0-6/8 years).
• 8 studies included school-age children (6+ years), with some spanning into adolescence or adulthood.
• We didn’t find mention of the age group in 7 studies.

Effect Sizes by Learning Domain

Cognitive Development:

• Effect sizes reported in the included studies range from 0.10–1.45 standard deviations, with larger effects observed in intensive, multi-year interventions (e.g., as reported by Ramey and Ramey, 2023).
• Meta-analyses included in this review (Jeong et al., 2021; Blewitt et al., 2018) reported pooled standardized mean differences of 0.18–0.32 for cognitive and early learning outcomes.

Language Development:

• Effect sizes typically ranged from 0.10–0.44 standard deviations, with some studies reporting larger effects for emergent literacy (Barnett et al., 2018: 1.10).
• Computer-supported and storybook-based interventions yielded small to moderate effects (Verhoeven et al., 2020: 0.28; Carolin and Fardzadeh, 2018: 0.21–0.27).

Mathematics:

• Effect sizes ranged from 0.18–1.22 standard deviations, with the largest effects in structured, curriculum-based interventions (Yıldız et al., 2025; Clements et al., 2011).
• Meta-analyses (Wang et al., 2016; Malofeeva, 2005) reported moderate effects (0.43–0.68; 0.47).

Socioemotional Development:

• Effect sizes were generally small to moderate (Blewitt et al., 2018: Cohen’s d=0.19–0.54; Jeong et al., 2021: standardized mean difference=0.19).
• Classroom-wide and universal social-emotional learning interventions showed consistent, though modest, benefits.

Motor Development:

• Physical activity and fine motor interventions yielded moderate effects (Grady et al., 2024: standardized mean difference=0.19–0.54; Strooband et al., 2020: moderate).

Academic Achievement:

• Reading and math achievement effect sizes ranged from 0.10–0.60 standard deviations, with larger effects in high-intensity or multi-year programs.
• Long-term follow-up studies (Reynolds et al., 2001; Campbell and Ramey, 1995) reported sustained academic benefits.

Other Domains:

• Executive function, school readiness, spatial skills, and problem-solving showed moderate to large effects in targeted interventions (Meghir et al., 2023; Yang et al., 2020; Burns et al., 2025).

Methodological Factors Influencing Effect Sizes

• Study Design:In the included studies, randomized controlled trials and large-scale longitudinal studies tended to report more robust and reliable effect sizes, while quasi-experimental and comparative studies appeared more susceptible to bias.
• Sample Size:Larger studies and meta-analyses provided more precise estimates, though heterogeneity remained a concern.
• Intervention Intensity and Duration:More intensive, longer-duration, and multi-component interventions were associated with larger and more sustained effects.
• Measurement Tools:Use of standardized, validated instruments enhanced comparability; studies using researcher-developed measures often reported larger effects.
• Fidelity and Implementation:Higher fidelity and professional development for educators were associated with greater effects (as reported by Carolin and Fardzadeh, 2018).
• Contextual Moderators:Socioeconomic status, country income level, and risk status of participants moderated effect sizes, with disadvantaged and low- and middle-income country populations often showing larger gains.

Generalizability and Study Context

• Geographical Diversity:The included studies spanned high-, middle-, and low-income countries, with interventions adapted to local contexts.
• Population Diversity:Most studies focused on low-income or high-risk populations, which may limit generalizability to more advantaged groups.
• Setting:Interventions were delivered in a variety of settings, including center-based, home-based, community, and digital environments. The context of delivery influenced both feasibility and effectiveness, as reported in several studies.
• Long-term Outcomes:Sustained benefits were more likely in intensive, multi-year interventions, but evidence was limited by the small number of long-term follow-up studies.
• Adverse Effects:Few studies reported adverse effects. One study (Love et al., 2013) noted increased parent-reported aggression with formal program participation. We didn’t find mention of other adverse effects in the available full texts or abstracts.

Limitations:

• Heterogeneity in intervention types, outcome measures, and reporting limited direct comparison across studies.
• Some studies lacked detailed reporting of effect sizes, statistical significance, or confidence intervals.
• Risk of bias was present in studies with small samples, non-randomized designs, or low fidelity.
• Generalizability was limited by the focus on high-risk and low-income populations in many studies.

References

Gregory Camilli, Sadako Vargas, S. Ryan, W. Barnett\ (2010).Meta-Analysis of the Effects of Early Education Interventions on Cognitive and Social Development. Teachers College Record

F. Campbell, E. Pungello, S. Miller-Johnson, M. Burchinal, C. Ramey\ (2001).The development of cognitive and academic abilities: growth curves from an early childhood educational experiment. Developmental Psychology

F. Campbell, C. Ramey\ (1994).Effects of early intervention on intellectual and academic achievement: a follow-up study of children from low-income families. Child Development

A. Reynolds, Judy A Temple, Dylan L. Robertson, Emily A. Mann\ (2001).Long-term effects of an early childhood intervention on educational achievement and juvenile arrest: A 15-year follow-up of low-income children in public schools. Journal of the American Medical Association (JAMA)

C. Blewitt, M. Fuller‐Tyszkiewicz, A. Nolan, Heidi J Bergmeier, D. Vicary, and 4 more\ (2018).Social and Emotional Learning Associated With Universal Curriculum-Based Interventions in Early Childhood Education and Care Centers. JAMA Network Open

Joshua Jeong, E. Franchett, C. V. Ramos de Oliveira, K. Rehmani, A. Yousafzai\ (2021).Parenting interventions to promote early child development in the first three years of life: A global systematic review and meta-analysis. PLoS Medicine

Abigail Cahoon, Carolina Jiménez Lira, Nancy Estévez Pérez, Elia Verónica Benavides Pando, Yanet Campver García, and 2 more\ (2023).Meta-Analyses and Narrative Review of Home-Based Interventions to Improve Literacy and Mathematics Outcomes for Children Between the Ages of 3 and 5 Years Old. Review of Educational Research

S. Ramey, Craig T. Ramey\ (2023).Early Childhood Education that Promotes Lifelong Learning, Health, and Social Well-being: The Abecedarian Project and its Replications. Medical Research Archives

F. Campbell, C. Ramey\ (1995).Cognitive and School Outcomes for High-Risk African-American Students at Middle Adolescence: Positive Effects of Early Intervention

A. Reynolds, Judy A Temple\ (1998).Extended early childhood intervention and school achievement: age thirteen findings from the Chicago Longitudinal Study. Child Development

D. Clements, J. Sarama, M. E. Spitler, Alissa A. Lange, Christopher B. Wolfe\ (2011).Mathematics Learned by Young Children in an Intervention Based on Learning Trajectories: A Large-Scale Cluster Randomized Trial

Noreen Yazejian, D. Bryant, S. Hans, Diane M. Horm, L. St Clair, and 2 more\ (2017).Child and Parenting Outcomes After 1 Year of Educare. Child Development

A. Grady, Rebecca Lorch, Luke Giles, Hannah Lamont, Amy Anderson, and 6 more\ (2024).The impact of early childhood education and care‐based interventions on child physical activity, anthropometrics, fundamental movement skills, cognitive functioning, and social–emotional wellbeing: A systematic review and meta‐analysis. Obesity Reviews

C. Ramey, F. Campbell, M. Burchinal, M. Skinner, D. Gardner, and 1 more\ (2000).Persistent Effects of Early Childhood Education on High-Risk Children and Their Mothers

Li Luo, B. Reichow, Patricia A. Snyder, J. Harrington, Joy C. Polignano\ (2020).Systematic Review and Meta-Analysis of Classroom-Wide Social–Emotional Interventions for Preschool Children. Topics in Early Childhood Special Education

Aubrey Wang, Janine M. Firmender, Joshua R. Power, James P. Byrnes\ (2016).Understanding the Program Effectiveness of Early Mathematics Interventions for Prekindergarten and Kindergarten Environments: A Meta-Analytic Review

Marit Carolin, Ebad Fardzadeh\ (2018).University of Southern Denmark The Effectiveness of a Large-Scale Language and Preliteracy Intervention The SPELL Randomized Controlled Trial in Denmark

Weipeng Yang, Haidan Liu, N. Chen, Peng Xu, Xunyi Lin\ (2020).Is Early Spatial Skills Training Effective? A Meta-Analysis. Frontiers in Psychology

A. Reynolds, Judy A Temple, Suh-Ruu Ou, Dylan L. Robertson, J. Mersky, and 2 more\ (2007).Effects of a school-based, early childhood intervention on adult health and well-being: a 19-year follow-up of low-income families. Archives of Pediatrics & Adolescent Medicine

L. Verhoeven, M. Voeten, E. V. Setten, E. Segers\ (2020).Computer-supported early literacy intervention effects in preschool and kindergarten: A meta-analysis. Educational Research and Reviews

Howard Goldstein, Lindsey A Peters-Sanders, Keri M. Madsen, Jeffrey M. Williams, J. Drobisz, and 4 more\ (2023).Efficacy of a Supplemental Small-Group Early Literacy Intervention Implemented by Early Childhood Educators. American Journal of Speech-Language Pathology

Karel F B Strooband, M. Rosnay, A. Okely, Sanne L. C. Veldman\ (2020).Systematic Review and Meta-Analyses: Motor Skill Interventions to Improve Fine Motor Development in Children Aged Birth to 6 Years. Journal of Developmental and Behavioral Pediatrics

J. Love, Rachel Chazan-Cohen, H. Raikes, J. Brooks-Gunn\ (2013).What makes a difference: Early Head Start evaluation findings in a developmental context. Monographs of the Society for Research in Child Development

Noreen Yazejian, D. Bryant, Laura J. Kuhn, Margaret R. Burchinal, Diane M. Horm, and 3 more\ (2020).The Educare intervention: Outcomes at age 3. Early Childhood Research Quarterly

Ensar Yıldız, Özge Koca, Şenel Elaldı\ (2025).Effectiveness of Early Intervention Programs in Developing Early Mathematical Skills: A meta-Analysis. Kuramsal Eğitimbilim

A. Feller, Todd Grindal, Luke W. Miratrix, Lindsay C. Page\ (2016).Compared to What? Variation in the Impacts of Early Childhood Education by Alternative Care-Type Settings

C. Meghir, O. Attanasio, Pamela Jervis, Monimalika Day, Prerna Makkar, and 6 more\ (2023).Early Stimulation and Enhanced Preschool: A Randomized Trial. Pediatrics

A. Holla, Magdalena Bendini, Lelys Dinarte, Iva Trako\ (2021).Is Investment in Preprimary Education Too Low? Lessons from (Quasi) Experimental Evidence across Countries. Policy Research Working Papers

Samantha Burns, Sumayya Saleem, Evelyn McMullen, O. Falenchuk, Linda A. White, and 2 more\ (2025).A systematic review and meta‐analysis of approaches to teaching problem‐solving skills in early childhood education and care settings: A focus on science, technology, engineering and mathematics activities. Revista de educación

F. Campbell, C. Ramey\ (1989).Preschool vs. School-Age Intervention for Disadvantaged Children: Where Should We Put Our Efforts?.

James S. Kim, Josh Gilbert, Qun Yu, Ch. Gale\ (2021).Measures Matter: A Meta-Analysis of the Effects of Educational Apps on Preschool to Grade 3 Children’s Literacy and Math Skills. AERA Open

W. Barnett, Kwanghee Jung, A. Friedman-Krauss, E. Frede, M. Nores, and 3 more\ (2018).State Prekindergarten Effects on Early Learning at Kindergarten Entry: An Analysis of Eight State Programs

C. Lonigan, David J. Purpura, S. B. Wilson, P. Walker, Jeanine Clancy-Menchetti\ (2013).Evaluating the components of an emergent literacy intervention for preschool children at risk for reading difficulties. Journal of Experimental Child Psychology

Helen F. Ladd, C. Muschkin, K. Dodge\ (2014).From Birth to School: Early Childhood Initiatives and Third‐Grade Outcomes in North Carolina

Italo Lopez Garcia, U. Saya, Jill E. Luoto\ (2021).Cost-effectiveness and economic returns of group-based parenting interventions to promote early childhood development: Results from a randomized controlled trial in rural Kenya. PLoS Medicine

A. J. Dowd, Ivelina I Borisova, Ali Amente, Alene Yenew\ (2016).Realizing Capabilities in Ethiopia: Maximizing Early Childhood Investment for Impact and Equity

Elena V. Malofeeva\ (2005).Meta -analysis of mathematics instruction with young children

Suzanne M. Winter, D. Sass\ (2011).Healthy & Ready to Learn: Examining the Efficacy of an Early Approach to Obesity Prevention and School Readiness

Corey A. DeAngelis, Heidi Holmes Erickson, Gary W. Ritter\ (2018).What’s the state of the evidence on pre-K programmes in the United States? A systematic review. Educause Review

L. Olive, R. Telford, E. Westrupp, R. Telford\ (2023).Physical activity intervention improves executive function and language development during early childhood: The active early learning cluster randomized controlled trial. Child Development

Download BIBDownload RISDownload TXT

Report

Status

Gather sources

997 sources found

Details

Screen sources

40 sources included

Details

Extract data

240 data points extracted

Details

Generate report

Save PDF

BIBLaTeX, ZoteroRISZotero, MendeleyTXTAPA BibliographyPDFPDFDOCXMicrosoft Word

Chat

Got some follow-up questions?

Sign up or sign in to chat with this report.

Back

Is Early Spatial Skills Training Effective? A Meta-Analysis

Weipeng Yang, Haidan Liu, N. Chen, Peng Xu, Xunyi Lin

Frontiers in Psychology·

2020·

52 citations

SourceDOI

Plain textPDFFigures

Searching for PDF

Unable to find PDF from source

Back

Study Design Type

Quasi-experimental study, Randomized Controlled Trial (RCT)

Intervention Type and Components

- Primary pedagogical approach: Hands-on exploration, inquiry-based learning - Specific learning domains targeted: Spatial skills (mental rotation, spatial perspective taking) - Additional support services: Feedback mechanisms (modeling, gesture feedback, spatial language feedback) - Delivery method: Classroom-based, individualized activities (video games, robotics) - Frequency and duration: 6-week intervention using iPads, 32-week geometry curriculum - Specific content: Block building activities, mapping activities, use of technology (iPads), multimedia visual perceptual individual training program, spatial reasoning intervention (geometry lessons, quick challenge spatial activities)

Participant Demographics

- Age range of children: 0-8 years - Socioeconomic status: Diverse, but not analyzed - Risk factors or special circumstances: NR - Total sample size: 900 (training), 635 (control) - Gender distribution: Greater effect for girls (g = 0.909) than boys (g = 0.686), but no specific numbers or percentages

Outcome Domains and Measurement

- Cognitive development: Spatial skills (intrinsic-static, intrinsic-dynamic, extrinsic-static, extrinsic-dynamic) - Effect sizes: intrinsic-static (0.456), extrinsic-static (0.770), intrinsic-dynamic (0.952) - Statistical significance: Implied through effect sizes - Follow-up time points: Not mentioned

Long-term Outcomes

- Specific outcome measured: Educational attainment (enhanced math skills and science understanding) - Age/time point of measurement: Not specified - Effect size or key findings: Not specified - Statistical significance: Not reported

Geographical and Contextual Factors

Not mentioned (the paper does not provide specific geographical or contextual factors for the studies included in the meta-analysis)

Spatial skills significantly predict educational and occupational achievements in science, technology, engineering, and mathematics (STEM). As early interventions for young children are usually more effective than interventions that come later in life, the present meta-analysis systematically included 20 spatial intervention studies (2009–2020) with children aged 0–8 years to provide an up-to-date account of the malleability of spatial skills in infancy and early childhood. Our results revealed that the average effect size (Hedges's g) for training relative to control was 0.96 (SE = 0.10) using random effects analysis. We analyzed the effects of several moderators, including the type of study design, sex, age, outcome category (i.e., type of spatial skills), research setting (e.g., lab vs. classroom), and type of training. Study design, sex, and outcome category were found to moderate the training effects. The results suggest that diverse training strategies or programs including hands-on exploration, visual prompts, and gestural spatial training significantly foster young children's spatial skills. Implications for research, policy, and practice are also discussed.

INTRODUCTION

Spatial skills are often applied in problem-solving situations, especially when processing and manipulating visuospatial information (Rafi et al., 2005). Studies have revealed that these skills strongly predict educational and occupational achievements in STEM (science, technology, engineering, and mathematics) domains (Wai et al., 2009; Lubinski, 2010; Uttal and Cohen, 2012;Stieff and Uttal, 2015). Improving spatial skills is therefore an important agenda for both research and educational practice (Hawes et al., 2017) . Although previous studies have showed that early interventions for young children are more effective than interventions that come later in life (Heckman and Masterov, 2007) , to what extent spatial skills training programs can effectively improve young children's spatial development remains understudied. It is worth noting that Uttal et al. (2013) have conducted a meta-analysis of training studies on spatial skills in general populations. However, this seminal work only included research evidence produced in 1984-2009 and did not focus on the training of early spatial skills.

In the past decade, we have observed an increase in research work on early spatial training and its effects, with more types of training approaches being used. In order to achieve an up-to-date understanding of the malleability of spatial skills in infancy and early childhood, the present meta-analytic study aims to synthesize spatial intervention studies that target young children aged 0-8 years from 2009 to 2020. Using a 2 × 2 typology of spatial skills (intrinsic vs. extrinsic and static vs. dynamic; Newcombe and Shipley, 2015) , we meta-analyzed the eligible (quasi-)experimental studies for examining the effect of training on early spatial skills and the potential moderating effects on the relationship between the training and early spatial development.

Spatial Skills in the Early Years

Spatial skills refer to the cognitive processing of spatial information, which "concerns shapes, locations, paths, relations among entities and relations between entities and frames of reference" (Newcombe and Shipley, 2015, p. 180) . There are two traditions of conceptualizing spatial skills, including the psychometric approach and the classification system approach (Uttal et al., 2013). The former relies on exploratory factor analysis for identifying the key components of spatial skills, while the latter is rooted in a system comprised of two fundamental distinctions, i.e., between intrinsic and extrinsic information and between static and dynamic tasks (Uttal et al., 2013; Newcombe and Shipley, 2015) . In this study, we extended the line of research on spatial skills training by following the 2 × 2 framework of spatial skills used in Uttal et al.'s (2013) seminal meta-analysis. According to Newcombe and Shipley (2015) , the 2 × 2 typology of spatial skills leads to four categories of spatial skills and various assessments, as shown in Table 1 . Based on the 2 × 2 framework of spatial skills (Newcombe and Shipley, 2015) , the measurements of spatial skills can be put into categories as aligned with the four categories of spatial skills.

Spatial skills or spatial thinking skills are found to undergo considerable development during infancy and early childhood (0-8 years of age) (Newcombe and Frick, 2010) . Prior research evidence indicated that infants as young as 4 months could show precursors of mental transformation (Rochat and Hespos, 1996; Hespos and Rochat, 1997) . Frick and Wang (2010) also found that 13-to 16-month-old infants could perform mental rotation tasks after practice. Besides mental rotation, Bai and Bertenthal (1992) showed that 8-month-old infants had the ability of perspective taking when they moved to keep track of the location of an object. Preschoolers aged 3-5 years were also shown to be able to locate an object relative to a different viewpoint (Newcombe and Huttenlocher, 1992) . However, individual differences exist in the early development of spatial skills (Hazen, 1982; Harris et al., 2013) .

Extrinsic-static Identifying the spatial location of objects relative to others

To represent the location of objects in a map Rod and Frame Test, performance of spatial relations, etc.

Extrinsic-dynamic Transformation of the inter-relations of objects in movement

To enable perspective taking in understanding astronomy

Piaget's Three Mountains Task, water tilting task, etc.

Newcombe and Shipley (2015) ; Uttal et al. (2013).

The significance of early spatial skills has been demonstrated by an extensive body of research, which links the early development of spatial thinking to map use (Liben et al., 2013) , numerical skills (Zhang, 2016;Cornu et al., 2018;Fanari et al., 2019), arithmetic development (Zhang et al., 2014), math reasoning (Casey et al., 2015), math knowledge (Rittle-Johnson et al., 2019), early writing skills (Bourke et al., 2014), motor skills (Jansen and Heil, 2010) , and executive functions (Lehmann et al., 2014; Frick and Baumeler, 2017). However, several lines of evidence suggest that there are early sex and socioeconomic status (SES) differences in spatial skills, with advantages for males and those with higher SES on spatial tests (Levine et al., 1999 (Levine et al., , 2005;; Quinn and Liben, 2008). Therefore, it is of importance to know whether early spatial skills can be improved, especially in girls and socially disadvantaged children.

Neurological evidence supports that early intervention can enhance the neural functioning for spatial thinking (Gersmehl and Gersmehl, 2007) . Prior studies also showed that the effects of early spatial training could be transferred to children's math skills (Cheng and Mix, 2014;Bower et al., 2020;Ribeiro et al., 2020;Thomson et al., 2020) and science understanding (Bower, 2017). For instance, Ribeiro et al. (2020) and Thomson et al. (2020) revealed that parental support such as spatial concept support and spatial language use in block building tasks or toy play situations tended to enhance young children's math performance. However, whether spatial skills training and support could lead to a substantial magnitude of improvement in early spatial development, as well as how it can be brought in an early childhood setting and incorporated into an early childhood curriculum, deserves more research.

Malleability of Spatial Skills and Early Interventions

Previous research supports that spatial skills are malleable and can be improved through spatial training or instruction. However, most of the solid evidence for supporting the malleability of spatial skills is revealed by studies in the population of adolescents and adults (Uttal et al., 2013). In the most recent meta-analysis of spatial skills training studies conducted by Uttal et al. (2013), 217 intervention studies were included for analysis, revealing that the average effect size for spatial skills training relative to control was Hedges's g = 0.47 (SE = 0.04). However, of the 217 studies, only 53 studies focus on children younger than 13 years, with very few focusing on infants, toddlers, and preschoolers. Therefore, it remains to be further explored how to promote spatial skills in the early years.

It is worth noting that most of the training interventions were conducted in a much more controlled setting rather than the naturalist educational setting (Uttal et al., 2013; Hawes et al., 2017) . Recent studies (e.g., Newcombe and Frick, 2010) have suggested that integrating spatial content into formal and informal instruction is meaningful for improving spatial functioning and reducing digital divides as related to sex and SES. As a result, more research is needed to test whether there is a difference in training effects across diverse settings, as well as demographic factors such as sex and SES. This will be a significant step forward in searching for an early spatially enriched curriculum (or "spatial curriculum" as promoted by Uttal, 2012) demonstrating the educational relevance of spatial training in the early years.

In terms of classroom-based spatial training, some have been conducted in early childhood settings. For instance, Ehrlich et al. (2006) found that gesturing provided meaningful cues about 5-year-old children's spatial strategies, which implied that gesture-based spatial training in the early childhood setting could be effective in improving mental rotation skills. In an experimental study, Casey et al. (2008) used block building activities to promote 6-year-old kindergarteners' spatial skills. They found that storytelling would provide a practical and useful context for teaching spatial content, while block building could develop children's various spatial skills (Casey et al., 2008). Petty and Rule (2008) also demonstrated the impact of mapping activities as supported by the use of materials such as toy figures, toy buildings, and photograph maps on the spatial skills of children aged 2.5-9, through a pretestposttest quasi-experimental study. Furthermore, Hawes et al. (2015) conducted a randomized controlled trial among 6-to 8year-olds to test the impacts of spatial skills training in regular classroom settings. Their research used iPad devices as the platform of early spatial skills training, and the intervention lasted 6 weeks. Evidence indicated that as compared to children in the control group, children who received the computerized spatial training demonstrated enhanced spatial skills (i.e., mental rotation) (Hawes et al., 2015). To make the spatial training more situated in the classroom, Hawes et al. (2017) further designed a 32-week geometry curriculum and conducted another experimental research study with 6-year-olds in their school. Results revealed that those young children's spatial and numerical skills (i.e., spatial language, visual-spatial reasoning, mental rotation, and symbolic number comparison) had been effectively improved using the spatially enriched approach to early geometry instruction (Hawes et al., 2017) .

In the past decade, there have been an increasing number of studies on the effects of early spatial skills training. In general, these studies seem to support that young children would significantly benefit from participating in intentional spatial tasks or activities. However, the effects of early spatial skills training have not been systematically investigated.

The Present Meta-Analytic Review

As mentioned above, spatial skills are shown to be malleable; therefore, early spatial skills training activities comprised of interactive components such as hands-on exploration and environmental feedback (e.g., visual cues) are expected to show positive effects. This theoretical assumption can be further supported by understanding the early development of spatial skills (i.e., early spatial development).

The underlying mechanism of early spatial development is complex and dynamic, as comprised of multiple elements, including natural maturation, cultural scaffolding, environmental feedback, and active exploration (Newcombe and Learmonth, 1999) . It involves both quantitative and qualitative aspects of cognitive change and continuity (Newcombe and Learmonth, 1999), which could be explained by Piaget's theory of cognitive development and Vygotsky's social development theory. The spatial development framework (Piaget, 1953;Piaget and Inhelder, 1956) describes children's progressive understanding of spatial relationships, from appreciating limited objects in the topological stage to considering distances and angles in the Euclidean stage. Although Piaget's cognitive constructivist approach has minimal emphasis on the role of cultural scaffolding, the functioning of schema through assimilation and accommodation provides implications that children's cognitive development can benefit from their interaction with the (physical) world in which they are living. Apart from Piaget, Vygotsky's (1978) sociocultural approach suggests that social interaction plays a fundamental role in cognitive development, which also applies to the specific development of spatial cognition.

Accordingly, the theoretical mechanism of early spatial development has assumed that environmental feedback and guidance in spatial training will improve an individual's ability to handle and manipulate specific spatial tasks. This meta-analysis assessed the extent to which spatial skills training programs could effectively improve young children's spatial development. Some meta-analytic or systematic reviews have examined the effectiveness of spatial skills training or related experiences (e.g., Baenninger and Newcombe, 1989;Spence and Feng, 2010;Uttal and Cohen, 2012;Uttal et al., 2013). However, to our knowledge, to date, there has been no systematic and dedicated research to examine the effect of spatial training on improving the spatial skills of children aged 0-8 years. To address this knowledge gap, we explored the effects of spatial skills training in the crucial life periods of infancy and early childhood, lasting from birth to 8 years. The following research questions thus guided this meta-analytic study:

1. What is the effect of early training on the spatial skills of children aged 0-8? 2. What variables moderate the effect of early spatial skills training?

Literature Search

The first author and the third author conducted an extensive automated search of electronic articles through the databases of PsycINFO, ERIC, EBSCO, ProQuest, and Scopus from February 1, 2009, through February 1, 2020. The literature search aimed to thoroughly identify randomized controlled trials or (quasi-)experiments studying the effects of early childhood interventions on the spatial skills development of children aged 0-8 years. Three different sets of terms with two Boolean operators (AND and OR) and the truncation character ( * ) were utilized to search for and download relevant literature from the databases: predictors (specific terms included "curriculum, " "intervention, " "approach, " "training, " and "program"), outcomes (specific terms included "spatial * , " "space, " "map, " "form perception, " "visual * , " and "visuospatial"), and sample (specific terms included "preschool, " "pre-K, " "prekindergarten, " "prekindergarten, " "kindergarten, " "primary school, " "elementary school, " "younger children, " "infant, " "toddler, " and "young children"). We created the search terms through extensive piloting. We used the operators "AND, " to connect search terms between the categories, and "OR, " to connect search terms within each category.

Inclusion and Exclusion Criteria

Two researchers (the first two authors) independently selected and reviewed a subset (25%) of the articles following the inclusion criteria:

1. Included studies were (quasi-)randomized controlled trials or (quasi-)experimental designs. 2. Participants were 0-8 years of age (i.e., mean age of the participants). 3. Spatial skills were measured as outcomes of the intervention. 4. The reported information was sufficient enough for effect sizes to be calculated. 5. English was the written language used. We excluded correlational studies (e.g., Levine et al., 2012) and reviews (e.g., Zimmermann et al., 2019). Non-full-text documents were also excluded because they may lack sufficient and credible information for meta-analysis.

Study Selection

Based on the above inclusion and exclusion criteria, the two researchers divided 25% of the selected articles into three categories: eligible, possibly eligible, and ineligible. The interrater reliability was good (Cohen's kappa coefficient κ = 0.70) (Cohen, 1960). In view of the differences, the two researchers discussed the adequacy of the articles marked as "possibly eligible" and made the final inclusion decision based on full common consensus. The first author finished the selection of the remaining articles (75%).

As shown in Figure 1 , which follows the PRISMA statement (Moher et al., 2009) , of the 505 records initially identified, 445 were excluded by title and abstract based on the predefined inclusion and exclusion criteria. Of the 60 records remaining and screened, nine were duplicates. We then performed manual searches of the reference list of eligible research reports and repeated this process until no other studies were found, thus adding eight full-text articles. Twenty studies were eventually included, resulting in 50 independent effect sizes.

Data Extraction

To identify interesting variables for research synthesis, Lipsey (2009) proposed three groups of study descriptors: extrinsic variables, method variables, and substantive variables.

1. Extrinsic variables are represented by fixed characteristics of the study, such as the date of publication, publication type, and funding source. We coded the date of publication in this meta-analysis. 2. Method variables are related to the control of the implementation fidelity and the psychometric properties of the measures. We included the type of study design and the category of spatial skills measures as the two method variables for the moderator analysis.

Data Analyses

We used the Comprehensive Meta-Analysis Version 3 (CMA v3; Borenstein et al., 2013) statistical software package to compute and analyze all the meta-analytic data, as follows:

Computing Effect Sizes

We calculated the effect sizes using Hedges's g, as the sample sizes in the included studies were mostly small (below 50) (Cohen, 2013; Hedges and Olkin, 2014) . This metric is appropriate, as it corrects biases due to sample size (Cohen, 2013). The coefficient of Hedges's g represents the difference in means between the two groups relative to the pooled and weighted standard deviation (Cohen, 2013). One effect size was calculated for each outcome category in each study.

Since the data for this meta-analysis were obtained from a series of published studies conducted by different people, it is unlikely that all studies are functionally identical (Borenstein et al., 2007(Borenstein et al., , 2011)). In this case, it is suggested that the random effects model is a more reasonable option for the meta-analysis (Borenstein et al., 2007(Borenstein et al., , 2010(Borenstein et al., , 2011)). However, when the number of studies is small (N < 10), the variance estimate between the studies is usually low, so it is better to calculate the average difference according to the fixed effect model (Borenstein et al., 2010). Therefore, this meta-analysis used a random effects model to calculate the overall effect size and chose either the random effects approach (N ≥ 10) or the fixed effect approach (N < 10) to calculate and compare the effect sizes across studies involving different categories of outcomes in the moderator analyses.

Publication Bias

We verified the possibility of publication bias using the trimand-fill method and a funnel plot of standard error by Hedges's g (Duval and Tweedie, 2000). The trim-and-fill analysis only slightly reduced the estimated average effect sizes. The estimated mean values of the trim-and-fill analyses were all significantly different from zero. The results of the additional analysis did not find any variable that could be used as an alternative interpretation of the current results. In addition, a funnel plot was generated against the results to examine the effect size distribution relative to the sample sizes (see Figure 2 ). Since most of the studies were symmetrically distributed around the average effect size, there was little publication bias observed (Borenstein et al., 2009). Therefore, we report the combined results of the 20 studies and 50 effect sizes in this meta-analysis.

Analyzing Variance in Effect Sizes

We studied the variability of the effect sizes across studies through the heterogeneity test (Hedges and Olkin, 2014; Schmidt and Hunter, 2014;Cooper, 2016). We thus identified moderators that may not have been studied in a single experiment and that may affect the magnitude of the training effects (Cooper, 2016).

A heterogeneity test compares the variance shown by a set of effects with the assumed variance due to sampling error (Higgins et al., 2003; Cooper, 2016). If the heterogeneity test results indicate that the difference in a set of effects can be attributed only to the sampling error, then the data can be assumed to represent the population of participants (Hunter et al., 1982) . We used the inter-group statistic, Q, to assess whether the group average effect is homogeneous (Yang et al., 2019). A statistically significant Q indicates that the grouping factor contributes to the variance in effect size; in other words, the grouping factor has a significant effect on the measurement of outcomes (Higgins et al., 2003) .

Effects of Early Spatial Interventions

We meta-analyzed 20 intervention studies on spatial skills for children aged 0-8 years. There were 900 children in the training group and 635 children in the control group. Table 2 presents the effect sizes and key characteristics of the included studies.

As shown in Table 2, 2 .

Although publication biases always exist in any meta-analysis (Lipsey and Wilson, 1993) (see the funnel plot in Figure 2 ), the random effects analysis results revealed that the average effect size (Hedges's g) for training relative to control was 0.96 (SE = 0.10).

Moderator Analyses

We further analyzed the moderating effects of several study descriptors, including the type of study design, sex, age, outcome category (i.e., type of spatial skills), research setting (e.g., lab vs. classroom), and type of training. We used the Q statistic to assess the significance of the heterogeneity test in the effect size. Table 3 presents the results of the moderator analysis of the effects of these six study descriptors on the spatial skills of the participating children.

As shown in Table 3 , the type of study design [within subjects (g = 0.328) < between subjects (g = 0.529) < mixed (g = 0.759)], sex [girls (g = 0.909) > boys (g = 0.686) > mixed (g = 0.499)], and outcome category [generic (g = 0.326) < intrinsic, static (g = 0.456) < extrinsic, static (g = 0.770) < intrinsic, dynamic (g = 0.952)] were found to moderate the training effects. However, there was no significant difference in age, type of training, and research setting as related to children's spatial skills outcomes. a Fixed effect approach is used for the moderator analysis of this variable. b Random effects approach is used for the moderator analysis of this variable.

DISCUSSION

Although existing meta-analyses have demonstrated that spatial skills are malleable and can be improved by training (Baenninger and Newcombe, 1989;Uttal et al., 2013), none of them exclusively focuses on the effect of training on young children's spatial skills.

To the best of our knowledge, this meta-analysis is the first attempt of its kind to systematically review and investigate the effects of spatial skills training in children aged 0-8 years.

Early Intervention Matters in the Development of Spatial Skills

This meta-analysis revealed that diverse training strategies or programs including hands-on exploration, visual prompts, and gestural spatial training could significantly foster young children's spatial skills. This finding demonstrated that young children's spatial skills could be significantly improved if they are given specific training, with an average effect size (Hedges's g) of 0.96 for training relative to control. The effect size obtained in the current meta-analysis is greater than the average effect (g = 0.47) indicated in Uttal et al.'s (2013) results. Therefore, our finding seems to support the argument that spatial skills, a kind of cognitive trait, are more malleable in the early years of life than the later stages such as adolescence and adulthood. However, this argument warrants further investigation, as only published papers are included in this meta-analysis, and publication bias may exist (Thornton and Lee, 2000). The positive effect of early spatial skills training revealed in this study aligned with the theoretical links between action and cognition for understanding the underlying mechanism of effective early spatial training strategies or programs. According to Newcombe and Frick (2010) , mental rotation and spatial perspective taking are the most crucial precursory forms of spatial skills in the early years, which are commonly related to motor development. Motor activities can thus facilitate children's performance in mental rotation and spatial perspective-taking tasks by engaging them in active movement (Newcombe and Frick, 2010) . As found in the present meta-analysis, most of the effective spatial training used video games, play, handson exploration, spatial tasks, or classroom-based courses as the intervention or stimuli. What they have in common is that hands-on exploration, visual prompts, and gestures are used to support the process of actively practicing spatial skills in various activities (e.g., Frick et al., 2009; Borriello and Liben, 2018;Bower et al., 2020). It is possible that the engagement in manipulating visuospatial information would require the involvement of different neural processes. This could further shape the neural functioning related to spatial skills. However, the neural mechanism has not yet been thoroughly unveiled in spatial training studies and requires more future research to make sense of the positive effects of early intervention on children's spatial skills.

Differences in the Response to Training: Study Design, Sex, and the Category of Spatial Skills

Our results revealed that the type of study design, sex, and outcome category moderated the effects of early spatial skills training. However, the moderator analyses revealed that age, research setting, and type of training did not have a significant, moderating effect on the training outcomes. The combined effect sizes indicated that different groups of age, training settings, and training approaches did not generate significantly different effect in promoting young children's spatial functioning. These findings suggest that various approaches such as hands-on exploration, visual prompts, and gestural spatial training could all lead to improvements in spatial skills across different age groups in the early years. This aligns with theoretical arguments given by Ehrlich et al. (2006) that environmental input plays a crucial role in the development of spatial skills, even though biology also contributes to spatial skills.

As revealed in this meta-analysis, research setting did not play a moderating role; however, as argued by Klahr and Li (2005) , there is an urgent need for studies on integrating cognitive research in laboratories with teaching in classrooms. A recent experimental study conducted by Hawes et al. (2017) provided empirical evidence that a classroom-based spatially enriched geometry course with a relatively long duration of 32 weeks could lead to young children's considerable progress in spatial skills. This research agenda requires more attention and endeavors, as our evidence indicated that classroom-based spatial skills training might be more effective (g = 1.16 > 0.69 in the laboratory setting). Below we further discuss the confirmed moderating factors.

Study Design

This meta-analysis revealed that the study design quality moderated the training effects. Although we only included studies using a (quasi-)experimental design, there are three different levels of quality regarding the rigor of design. The results showed that those experiments with both a within-and betweensubjects design (N = 38) had the largest effect sizes regarding the training effect (average g = 0.759). However, it is unclear why within-subject comparison does not lead to a higher extent of positive training effect on average. There are two possible explanations. First, this may be caused by the effect of publication bias, as academic journals tend to be in favor of between-subject experimental research with more positive results (Song et al., 2010). Second, the existence of a control group seems to increase the effect sizes of training; therefore, we suggest that there could be negative effects brought by the lack of targeted spatial skills training for specific assessments. As this may be contradictory to the potential learning of test-taking strategies by children in the control group (Müller et al., 2012) , more research is needed to directly investigate these claims regarding the effect in the control group such as practice effects in spatial skills assessments among young children.

Sex

Existing meta-analyses demonstrated that men outperform women on measures of mental rotation and spatial perception (Linn and Petersen, 1985; Voyer et al., 1995; Maeda and Yoon, 2013) . The male performance advantage in spatial skills seems to start as early as infancy and early childhood (Levine et al., 1999; Moore and Johnson, 2008; Quinn and Liben, 2008). Our metaanalysis revealed that early spatial skills training would lead to greater effect for girls (g = 0.909) than boys (g = 0.686). Our finding supports the suggestion given by Newcombe and Frick (2010) that the integration of spatial learning opportunities into early childhood education could not only promote spatial skills in general but also reduce early sex differences that may impede female citizens' full participation in the current digital world. Such an encouraging consequence of introducing spatial skills training in early childhood settings further demonstrates that experiences with spatially enriched stimuli and activities would benefit children in their spatial cognition and reduce the sex differences in this cognitive trait (Baenninger and Newcombe, 1989; Moore and Johnson, 2008) .

Category of Spatial Skills Assessment

This meta-analysis revealed that the category of spatial skills measures moderated the training effects. Results indicated that different categories of spatial tasks respond differently to training, with the mean weighted effect sizes for intrinsic-static, extrinsic-static, and intrinsic-dynamic kinds of assessment at 0.456, 0.770, and 0.952, respectively. The moderating role of the kinds of spatial skills assessment is consistent with the result revealed in Uttal et al.'s (2013) meta-analysis. However, in the early years, children tended to perform better in mental rotation as featured in the intrinsic-dynamic category of assessment instead of the extrinsic-static category. Although our finding seems to align with an extensive body of literature that records infants' and young children's performance in mental rotation tasks (e.g., Moore and Johnson, 2008; Frick and Wang, 2014; Lehmann et al., 2014) , more direct research is needed to ascertain what the exact differences of effects are when measuring children's spatial skills using different assessments.

Limitations of This Meta-Analysis

One of the limitations of our meta-analysis is that as the number of studies involved is relatively small, the effect sizes across studies are considerably heterogeneous. The variance in effect size may explain why heterogeneity between groups is not significant for the results of moderator analyses of certain research descriptors (e.g., type of training and research setting). Although the publication biases were shown to be acceptable using the trim-and-fill method, the generalization of our findings to other contexts and populations should be conducted with caution due to the small number of eligible studies included. Moreover, only published English papers were included in this meta-analysis due to the inaccessibility of other types of articles. This may have led to biases in our metaanalysis because studies reporting a significant impact are more likely to be published than studies not reporting statistical significance (Rosenthal, 1979). Also, our moderator analyses did not cover the factors of SES, initial level of performance on spatial tasks, and intervention duration. The included studies reported that their participants were from families of diverse socioeconomic backgrounds; therefore, we were not able to analyze the moderating effect of SES in the current meta-analysis. Although this meta-analysis attempted to control study design, it was still unable to adequately capture or control certain variables, such as trainers' qualifications and the duration of training, because these variables were not clearly reported in the included studies.

Last but not least, this meta-analysis did not include nonexperimental research as well as those studies on transfer effects of spatial skills training to untrained tasks. The current meta-analysis only included studies examining the relationship between training programs and the development of spatial skills. However, meta-regression can also be used to examine the relationship between spatial training and children's spatial skills and other related outcomes (e.g., math skills, scientific task performance, and executive function), so that correlational studies can be meta-analyzed. Correlational studies may be valuable for exploring the complex behavioral and neural mechanisms behind the training effect. Subsequent qualitative systematic reviews or meta-regression analyses of the processes and mechanisms through which early spatial skills can be enhanced would be of great importance.

Implications for Research, Policy, and Practice

Our research contributes to the literature in the field of spatial thinking by showing whether and how early intervention approaches and programs can promote young children's spatial functioning through meta-analytic evidence. Our meta-analysis thus expands this line of research on the malleability of spatial skills in the early years and provides the following implications for future research, policy-making, and practice in early childhood education.

First, early spatial intervention matters. Our evidence indicated that the malleability of spatial skills is stronger in younger children, as compared to the average effect size (g = 0.47) found in the general population (Uttal et al., 2013).

Second, a spatially enriched curriculum should play a more vital role in early childhood education via the integration of effective practices such as spatial play (block building) and purposeful use of visual and verbal cues. This is also supported by our evidence that classroom-based spatial skills training is more effective (g = 1.16) than laboratory-based training (g = 0.69). To implement effective spatially oriented curricula in early childhood settings, more specific research is needed to design, implement, and evaluate classroom-based spatial training programs for young children.

Third, as linked to the previous implication, both early childhood policymakers and practitioners should consider scaling up effective classroom-based spatial training. Publicity and promotion require not only more research endeavors but also initiatives in policy and practice so as to bridge the gap between the laboratory environment and authentic learning settings and foster early spatial skills among children from diverse backgrounds, especially those placed in socially disadvantaged environments such as poverty and adverse parenting practices.

Fourth, to support children with difficulties in spatial functioning, spatially relevant game tasks can be used. For instance, visuospatial representation and transformation activities based on Quick Draws (Tzuriel and Egozi, 2010), playing with robotics (Keren et al., 2012) , rotating objects on mobile devices or computers (Ping et al., 2011;Hawes et al., 2015;Cornu et al., 2019), and tangram-related activities (Chabani and Hommel, 2014) are shown to significantly foster young children's spatial skills. Moreover, adult educators such as teachers and parents can provide children with more opportunities of manual exploration of the object, such as building blocks (Möhring and Frick, 2013) , and intentionally give various types of feedback (e.g., modeling, gesture feedback, and spatial language feedback) during spatially relevant activities (Bower et al., 2020). Some early interventions such as a multimedia visual perceptual individual training program (Chen et al., 2013) and spatial reasoning intervention including geometry lessons and quick challenge spatial activities (Hawes et al., 2017) can also be provided. However, more studies are needed to explore how to tailor spatial training programs to the specific abilities and disabilities of individual children.

Fifth, as early spatial skills training is demonstrated to more effectively enhance girls' spatial functioning and minimize the male advantage in this aspect, girls should be given the priority to engage in spatially enriched experiences.

Last but not least, more future research is warranted to explore the behavioral and neural mechanisms underlying the effects of spatial training in the early years. Two aspects should be focused on: study design and assessment. On the one hand, future research should draw upon a more rigorous design using randomized controlled trials and even a longitudinal design to investigate the training effects in the long run. One the other hand, there is an urgent need to conduct specific research on measuring children's spatial skills using different assessments. Moreover, how the improvement of early spatial skills may be linked to fostering other core skills such as numeracy, math reasoning, early writing skills, and executive functions can be explored in the future.

CONCLUSION

This meta-analysis supports the notion that effective spatial learning components could be infused into early childhood settings, so as to spatialize the curriculum and encourage children learn to think spatially (Newcombe and Frick, 2010; Bruce et al., 2015). To implement effective spatially oriented curricula in early childhood settings (Newcombe and Frick, 2010; Uttal and Cohen, 2012), early childhood researchers, policymakers, and practitioners should work together to intentionally support children's hands-on, proactive manipulation and processing of spatial information. The US National Research Council (2006) has released a national report to call for a curriculum and support system for spatial thinking in the K-12 educational context. Taking off from this research and policy achievement, high-quality, evidence-based, contextually appropriate spatial curricula should also be developed and provided for children to promote their spatial intelligence and help them become better prepared for the high-tech world.

AUTHOR CONTRIBUTIONS

WY designed the research and drafted the manuscript. WY, HL, NC, and PX collected and extracted data for analysis. XL provided important ideas and substantial feedback for the study and edited the manuscript. All of the authors read and approved the final manuscript.

Conflict of Interest:

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Funding

Frick, A., and Wang, S. (2010). "Round and round she goes: Effects of hands-on training on mental rotation in 13-to 16-month-olds, " in Poster presented at the XVIIth Biennial International Conference on Infant Studies (Baltimore).National Research Council.(2006).Learning to Think Spatially.Washington, DC:The National Academies Press. FUNDINGThis study was funded by First-rate Undergraduate Course Construction Project on Science Education in Early Childhood in Fujian Province in 2019.

Loading PDF...

StripeM-Inner

Paper sources

Abstract screening pilot

Abstract screening results

Extraction pilot

Extraction results

Research report

Modify setup