Children actually CAN compute scalar implicatures & only fail when we proffer "worlds" that invite unintended inferences. Check out new work by @ebruevcen.bsky.social showing early success in a standard TVJ task for both disjunction & some/all & how this totally changes our lab's account of failire.
2/ The standard story: kids hear "Some of the horses jumped" when ALL did, and accept it as a perfectly good description.
But WHY? One answer: kids can't access stronger relevant alternatives (like "all") to compute the implicature.
6/ Our take: kids spontaneously access relevant alternative utterances and compute scalar implicatures when allowed to generate their own interpretations before evaluating experimenter-provided worlds, and that previous studies underestimate this ability by artificially proffering irrelevant worlds.
3/ But here's what act-out studies show: kids almost never act on all objects when they hear "some," or both when they hear "or." But that doesn't mean they computed the implicature. Acting on one object doesn't tell us whether they'd reject "both" So we designed a novel two-phase task: Act-Out+TVJ
5/ Same pattern in Exp. 2 with "some" ("Lucy colored some of the cows").
Act-Out phase: kids colored 1, 2, or 3 cows, almost never all four. TVJ phase: 82% rejected the puppet coloring "all four"
New preprint w/ @drbarner.bsky.social, and probably the most fun project I've worked on!!
Past studies find kids fail to compute scalar implicatures. Do those tasks test spontaneous interpretation, or do they proffer worlds kids would never consider? 🧵 osf.io/preprints/ps...
