The Critical Period Hypothesis in Second Language Acquisition: A Statistical Critique and a Reanalysis

reanalysis of published age-of-acquisition datasets

Participants of the reanalysed studies - adult second-language learners of varying ages of acquisition

A methodological critique with reanalysis. The argument is that the analyses conventionally used to test the critical-period hypothesis - comparing correlations in arbitrarily split age bands, or comparing group means either side of a chosen breakpoint - cannot in principle distinguish a genuine discontinuity from a steady decline, because the split point is chosen rather than estimated. Vanhove refits the same data with breakpoint regression that estimates the breakpoint.

The standard tests are not diagnostic of a critical period. When the breakpoint is estimated rather than assumed, the published datasets do not show the discontinuity the hypothesis requires; a steadily declining function fits at least as well. The paper is the methodological ancestor of the later van der Slik et al. reanalysis of Hartshorne et al.

Not applicable; no new data.

The methodological point has been made independently several times (Hakuta et al. 2003; van der Slik et al. 2022) and has not been rebutted on its own terms - the counter-argument from Hartshorne is that his model does estimate rather than assume the breakpoint.