TY - JOUR
T1 - Generalizing beyond the input
T2 - The functions of the constructions matter
AU - Perek, Florent
AU - Goldberg, Adele E.
N1 - Funding Information:
We are grateful to Stefan Gries, Clarice Robenalt, and Laura Suttle for statistical advice, as well as to Martin Pickering and two anonymous reviewers of an earlier draft for very helpful comments and suggestions. This research was supported by a postdoc scholarship granted to the first author by the DAAD ( German Academic Exchange Service ), and by an Einstein visiting fellowship from the Einstein Foundation in Berlin to the second author.
Publisher Copyright:
© 2015 Elsevier Inc.
PY - 2015/10/1
Y1 - 2015/10/1
N2 - A growing emphasis on statistics in language learning raises the question of whether and when speakers use language in ways that go beyond the statistical regularities in the input. In this study, two groups were exposed to six novel verbs and two novel word order constructions that differed in function: one construction but not the other was exclusively used with pronoun undergoers. The distributional structure of the input was manipulated between groups according to whether each verb was used exclusively in one or the other construction (the lexicalist condition), or whether a minority of verbs was witnessed in both constructions (the alternating condition). Production and judgments results demonstrate that participants tended to generalize the constructions for use in appropriate discourse contexts, ignoring evidence of verb-specific behavior, especially in the alternating condition. Our results suggest that construction learning involves an interaction of witnessed usage together with the functions of the constructions involved.
AB - A growing emphasis on statistics in language learning raises the question of whether and when speakers use language in ways that go beyond the statistical regularities in the input. In this study, two groups were exposed to six novel verbs and two novel word order constructions that differed in function: one construction but not the other was exclusively used with pronoun undergoers. The distributional structure of the input was manipulated between groups according to whether each verb was used exclusively in one or the other construction (the lexicalist condition), or whether a minority of verbs was witnessed in both constructions (the alternating condition). Production and judgments results demonstrate that participants tended to generalize the constructions for use in appropriate discourse contexts, ignoring evidence of verb-specific behavior, especially in the alternating condition. Our results suggest that construction learning involves an interaction of witnessed usage together with the functions of the constructions involved.
KW - Argument structure
KW - Artificial language learning
KW - Generalization
KW - Language acquisition
KW - Novel construction learning
KW - Statistical learning
UR - http://www.scopus.com/inward/record.url?scp=84935905032&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84935905032&partnerID=8YFLogxK
U2 - 10.1016/j.jml.2015.04.006
DO - 10.1016/j.jml.2015.04.006
M3 - Article
AN - SCOPUS:84935905032
SN - 0749-596X
VL - 84
SP - 108
EP - 127
JO - Journal of Memory and Language
JF - Journal of Memory and Language
ER -