Second Language Learning through Immersion: A Review of U.S. Programs

narrative review of US immersion programme evaluations

Elementary students in US French and Spanish immersion programmes, with reference to the Canadian French immersion literature

Narrative review by one of the field's founders, drawing on programme evaluations that are almost entirely non-randomised comparisons of immersion enrollees against neighbourhood peers. Two structural problems run through the whole literature it summarises and are not solved in it: families choose immersion, and weaker students differentially LEAVE immersion, so the surviving immersion cohort is selected twice.

The canonical summary of the immersion literature, and worth separating into its two halves. On FIRST-language and content outcomes, immersion students match or exceed comparison peers by the end of elementary school despite receiving much of their instruction in another language - the finding that makes immersion defensible. On SECOND-language outcomes, immersion students reach high functional proficiency, clearly above conventional foreign-language instruction, but do NOT reach native-like productive competence: receptive skills approach native levels while speaking and writing retain persistent non-native grammar. Immersion buys usable fluency, not a native speaker.

High. Self-selection into immersion and differential attrition out of it both bias the comparison in the same, favourable direction, and neither is addressed by the underlying evaluations. This is precisely the confound the Portland lottery removes - and note that the lottery's English-reading effect (0.13-0.22 SD) is much smaller than the observational literature's.

The two-part conclusion - no cost to L1, substantial but non-native L2 - has been restated by every subsequent immersion review, but always over the same self-selected evaluations. The L1 half is now corroborated by lottery evidence; the L2 half is not, because the lottery study did not measure L2.