Critical Evidence: A Test of the Critical-Period Hypothesis for Second-Language Acquisition

2,300,000 immigrants (Spanish and Chinese first-language backgrounds)

Respondents to the 1990 US Census who immigrated to the United States, Spanish or Chinese L1

Cross-sectional analysis of census data. The design is the right one for the question - it tests the DISCONTINUITY that a critical period predicts, rather than merely showing that attainment falls with arrival age, which everyone agrees on. The weakness is the outcome: a self-reported census item on English-speaking ability, not a proficiency test. Education level is controlled and turns out to be a large predictor in its own right.

Large linear effects of level of education and of age of immigration, but adding parameters for a change in slope or a change in mean at a putative critical age accounted for a negligible amount of additional variance. The decline in second-language attainment with arrival age is real and steady; it does not show the discontinuity that is the essential hallmark of a critical period.

High as an observational design. Age of immigration is a family decision correlated with parental education, motivation and resources; the paper controls education but not heritable language aptitude. Note the direction of the bias, though: these confounds would tend to CREATE apparent age effects, and the paper still finds no discontinuity.

Extended by Wiley, Bialystok & Hakuta (2005) using additional census approaches with the same conclusion. The "steady decline, no discontinuity" reading has since been supported by Vanhove's (2013) breakpoint reanalyses and by van der Slik et al.'s (2022) reanalysis of Hartshorne's data for most learner types, and disputed by Hartshorne et al. (2018).