Impossible but not improbable events boost children's word learning, with step-like rather than graded performance¶

Claim · All claims
Evidence · 1 study (2 entries) · review r2 · q1
Currency · newest study 2019 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q1 i? Children who saw a 50% probable outcome failed to learn a novel word while children who saw a 0% impossible outcome succeeded. → Stahl Aimee E. 2019 q1 i? Children seeing improbable outcomes (10%, 5%, or 2.5% likelihood) all failed to learn, with no graded improvement as events became more surprising. → Stahl Aimee E. 2019 (2)

Evidence¶

Stahl Aimee E. 2019¶

Stahl Aimee E., Feigenson Lisa. (2019). Violations of Core Knowledge Shape Early Learning. Topics in Cognitive Science. https://doi.org/10.1111/tops.12389

q1 · i? · review · r2

Gumball machine experiment (Stahl & Feigenson, in preparation b) with 2- and 3-year-olds: a pink toy emerged from a machine containing 10 purple and 10 pink toys (probable) or 20 purple toys (impossible), followed by novel word teaching and a referent-selection test.

"children who saw the probable outcome (50% likelihood) failed to learn the novel word, whereas children who saw the impossible outcome (0% likelihood) succeeded"

Stahl Aimee E. 2019 (2)¶

Stahl Aimee E., Feigenson Lisa. (2019). Violations of Core Knowledge Shape Early Learning. Topics in Cognitive Science. https://doi.org/10.1111/tops.12389

q1 · i? · review · r2

Three additional groups saw pink toys emerge at 10%, 5%, or 2.5% likelihood; all failed to learn, showing step-like rather than continuous performance, contrasting with prediction-error accounts where learning scales with statistical expectancy.

"In all of these cases, children failed to learn the novel word used to label the pink toy. Interestingly, there was no evidence of graded performance; learning did not improve as the event became more surprising"

Discussion¶

Learner Variables¶

• Vocabulary — outcome: instruction changes it

Related Claims¶

• Three- to six-year-old children learn novel verbs and nouns from a single exposure following core knowledge violations, including outside the laboratory — related
• Eleven-month-old infants learn a new object-sound mapping after a surprising core knowledge violation but not after an expected outcome — related
• Core knowledge violations may be 'moderately impossible,' aligning with the Goldilocks effect on attention allocation — a broader claim this one bears on
• Eleven-month-old infants who see a core knowledge violation preferentially explore the violating object over a novel distractor — related
• Learning enhancements from core knowledge violations are entity-specific: no benefit for unrelated bystander objects — related