Learning structural alternations : What guides learners' generalization?

Copyright © 2021 Elsevier B.V. All rights reserved..

Some word-order alternations observed across the world's languages are constrained by specific verb choice, whereas one type of word-order alternation (i.e., scrambling) frequently seen in free word order languages is not lexically-dependent on the verb. Three novel-language learning experiments explore whether speakers latently respect this generalization. If learners show conservativeness that closely reflects statistics from the input, then it would support usage-based and statistical accounts; alternatively, if learners have linguistic biases that allow them to generalize beyond statistics and show generalization similar to typological patterns, then it would support an internal bias account. In each of the three experiments, two groups of English monolinguals learned a Korean-English hybrid language with structural alternations analogous to those found in different categories of natural languages, as defined by whether the language allows scrambling and whether alternations are lexically-dependent on the verb. Learners' generalization patterns in subsequent picture description and acceptability judgment tasks were analyzed. Comprehension data consistently showed that the group which learned alternations found in natural languages with relatively rigid word orders tended to be more verb-wise conservative than the group that learned alternations found in languages with relatively free word orders. Production data trended in the same direction as the comprehension data. Thus, our results suggest that learners have linguistic biases that mirror typological differences that help learners go beyond simple statistics tracking.

Medienart:

E-Artikel

Erscheinungsjahr:

2021

Erschienen:

2021

Enthalten in:

Zur Gesamtaufnahme - volume:215

Enthalten in:

Cognition - 215(2021) vom: 01. Okt., Seite 104828

Sprache:

Englisch

Beteiligte Personen:

Lau, Sin Hang [VerfasserIn]
Momma, Shota [VerfasserIn]
Ferreira, Victor S [VerfasserIn]

Links:

Volltext

Themen:

Artificial language
Journal Article
Learning
Sentence production
Structural alternation
Verb argument structure

Anmerkungen:

Date Completed 20.10.2021

Date Revised 20.10.2021

published: Print-Electronic

Citation Status MEDLINE

doi:

10.1016/j.cognition.2021.104828

funding:

Förderinstitution / Projekttitel:

PPN (Katalog-ID):

NLM327927518