The Evidence on Teaching

Multiple Inference and Gender Differences in the Effects of Early Intervention: A Reevaluation of the Abecedarian, Perry Preschool, and Early Training Projects

Anderson, M. L. · 2008

grade Bcritiqueindependentmixednumbers spot-checked
Sample
299 across three trials (Abecedarian 111 + Perry 123 + Early Training 65)
Population
The three classic boutique early-childhood RCTs, reanalyzed jointly.
Design
Applies false-discovery-rate and family-wise-error corrections plus summary-index aggregation across the full outcome families, and tests cross-trial replication.
Key findings
Once you correct for testing dozens of outcomes, the Perry evidence base shrinks to two robust effects. Benefits are concentrated in FEMALES for attainment (pooled adult .27, SE .09) and are absent for males (-.05). This is the single most important corrective to the Perry/Abecedarian policy literature.
Counterfactual
Abecedarian controls got iron-fortified formula, free diapers and social services — NOT nothing. Perry controls got nothing.
Genetic confound
N/A (statistical reanalysis of RCTs).
Replication notes
The decisive multiplicity check. NOTE the balance: Anderson's own cross-trial replication verdict is 'In general, yes' for the female attainment effects — multiplicity kills the MALE and CRIME results, not the entire literature.
DOI / URL
10.1198/016214508000000841

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
PERRY standalone FDR reanalysis — the key resultoutcomes surviving FDRAt q=.05 only TWO survive: early MALE IQ and FEMALE high-school graduation (3 more at q=.10)mixedadultbusiness-as-usualadulthoodattainment
Cross-trial replication of Perry's extra outcomesreplication6 of 8 outcomes clearing naive p<.10 fail to replicate in Abecedarian or Early Trainingmixedadultbusiness-as-usualadulthoodattainment
Pooled summary index, TEEN stageES (SE)females .51 (.13); males .08 (.14); gender difference p=.029mixedgrade-6+business-as-usualover-2yrattainment
Pooled summary index, ADULT stageES (SE)females .27 (.09); males -.05 (.11); difference p=.027mixedadultbusiness-as-usualadulthoodattainment
Adult Perry crime effectsFDRNo adult Perry crime effect rejects at q=.05mixedadultbusiness-as-usualadulthoodbehaviour

Cited by

Multiple Inference and Gender Differences in the Effects of Early Intervention: A Reevaluation of the Abecedarian, Perry Preschool, and Early Training Projects · The Evidence on Teaching