When
LouAnn Gerken
Professor, Psychology and Cognitive Science
Director, Language Development Laboratory
University of Arizona
Child~Adult Differences in Learning Two Linguistic Generalizations
Abstract: Infants and children learn language differently than adults, as evidenced by superior mature
performance on many measures of language for earlier-exposed learners. In this talk, I attempt
to shed light on the nature of age-effects on language learning by comparing younger vs. older
learners’ abilities to infer two types of linguistic generalizations. One generalization involves
second order relations among segments in a word-like string (e.g., in CVCV nonwords, C2 is
voiced if C1 is voiced, or C2 is voiceless if C1 is voiceless). Such second order relations
characterize many morpho-syntactic rules of language, which are particularly difficult to learn
by adults, as well as by children with developmental language disorder (DLD). Studies from our
lab and others suggest that typically developing infants and children are superior to adults
learning such generalizations, both in terms of speed and overall success. I contrast this type
of generalization with one involving family resemblance, in which nonwords must exhibit two
out of three relevant properties to fit the generalization (e.g., C1 voiced, C2 voiced, V1 front).
Adults, including those with DLD, readily learn this generalization, while infants and toddlers do
not. I will suggest that one reason for this pattern of data is that younger humans, unlike either
Large Language Models or older humans, compute generalizations over very local input data
(good for within-unit sequential dependencies), as opposed to over huge or merely large
numbers of stored forms (good for between-unit family resemblance relations). Given time and
courage, I will speculate on what this account means for DLD and for language learning more
generally.