Critical Period Claim Revisited: Reanalysis of Hartshorne, Tenenbaum, and Pinker (2018) Suggests Steady Decline and Learner-Type Differences

reanalysis of the same 669,498-respondent dataset

As Hartshorne et al. 2018

Re-analysis of the original public dataset. The authors' two structural objections are (1) that the original model was fit to AGGREGATED rather than individual data, and (2) that fitting one model across all learner types manufactures an apparent common breakpoint. They refit exponential-learning-with-sigmoidal-decay (ELSD) models separately for monolinguals, bilinguals, early-immersion, later-immersion and non-immersion learners and compare continuous against discontinuous specifications.

A single sharply defined critical age of 17.4 for all learners is described as an artefact. A CONTINUOUS ELSD model fits better when applied separately to monolinguals, bilinguals and early-immersion learners. A DISCONTINUOUS model fits better only for non-immersion learners (critical age 18.6) and later-immersion learners (critical age 19.0). The authors read those two breakpoints as SCHOOLING effects - the age at which formal education ends - rather than as maturational cognitive constraints, and conclude that the critical-period claim is unwarranted.

Inherits the original design's confounds entirely; a reanalysis cannot fix selection into the sample or non-random age of arrival.

Hartshorne (2024) published a rebuttal in Language Development Research arguing that the reanalysis misunderstands what its own tests establish - that they test whether decline is asymmetric rather than whether it is sharp - and that after correction every bilingual type shows a sharp late-adolescent drop. The dispute is live and unresolved; both papers are recorded here rather than one being suppressed.