言語入力に含まれる誤りの人工言語習得に及ぼす効果

守 一雄

doi:10.4992/jjpsy.51.179

THE EFFECT OF MISTAKES IN THE INPUT ON THE ACQUISITION OF A MINIATURE ARTIFICIAL LANGUAGE

KAZUO MORI

Author information

JOURNAL FREE ACCESS

1980 Volume 51 Issue 4 Pages 179-187

DOI https://doi.org/10.4992/jjpsy.51.179

Details

Published: November 20, 1980 Received: November 09, 1979 Available on J-STAGE: July 16, 2010 Accepted: - Advance online publication: - Revised: -
Correction information
Date of correction: July 16, 2010 Reason for correction: - Correction: ABSTRACT Details: Wrong : There have been two major turning points in the theories of language acquisition with regards to whether the input to language learning children should be well-formed or not. Until 1950's, when it was believed that a child was learning his first language through imitation and/or verbal conditioning, it seemed to be taken for granted that the utterances the child listened to were all well-formed. The first turning point against this view was given by the rationalistic view of language acquisition. Chomsky (1965) proposed the innate language acquisition device (LAD) to explain the acquisition of language. He pointed out that although the linguistic competence was free from errors the linguistic performance was not so well-formed as it had been thought; and in fact, some other researchers (e.g., Bever, Fodor, & Weksel, 1965) showed that there were a lot of errors and hesitations in adult's speech. The second turning point was made by recent trend in the studies on mothers' speech. Although adult-to-adult speech contained a lot of speech errors as the former researchers had pointed out, mothers' speech to their children was found to be surprisingly well-formed. Newport (1976), for example, found in her study only one incidence of ungrammatical utterances out of 1500. The similar findings have been reported by many other researchers (cf., Snow & Ferguson, 1977). However, this does not necessarily mean that well-formedness is a necessary condition for language acquisition. It might be possible for a child to acquire a language even when he was exposed to an input which contained a fair number of errors and mistakes. It is not possible, however, to prove this experimentally by supplying some errors in the input to a child learning a language.
The present study was designed for serving some data for this controversy using adult subjects learning a miniature artificial language. Fifty college students learned a miniature artificial language under five different conditions: In the control condition subjects were presented 80 sample sentences with their referents four times and required to learn the correspondence between the words and the referents and infer the grammatical structure of the language. In four experimental conditions, the same procedures were used except that either 16 or 32 grammatical mistakes (GM-16, GM-32), or 16 or 32 referential mistakes (RM-16, RM-32) were inserted in these 80 sample sentences. After each presentation of the 80 sample sentences a test was given to measure the subjects' progress in learning the structure of the language. After the four presentations and four tests, a vocabulary test was given to the subjects to measure how much vocabulary they acquired.
Experimental results suggest that subjects can acquire a miniature artificial language even if they are exposed to an ill-formed input corpus that contains a certain number of grammatical mistakes or referential mistakes. This finding would lend support to the hypothesis that well-formedness of linguistic input would not be a necessary condition for children's language acquisition.
Right : Fifty college students learned a miniature artificial language under five different conditions: In the control condition subjects were presented 80 sample sentences with their referents four times and required to learn the correspondence between the words and the referents and infer the grammatical structure of the language. In four experimental conditions, the same procedures were used except that either 16 or 32 grammatical mistakes (GM-16, GM-32) or 16 or 32 referential mistakes (RM-16, RM-32) were inserted in these 80 sample sentences. The results showed that even under the GM-16, RM-16, and RM-32 conditions some subjects could acquire the language much as the subjects under the control condition. These results were discussed from the viewpoint of child language acquisition.

Abstract

Fifty college students learned a miniature artificial language under five different conditions: In the control condition subjects were presented 80 sample sentences with their referents four times and required to learn the correspondence between the words and the referents and infer the grammatical structure of the language. In four experimental conditions, the same procedures were used except that either 16 or 32 grammatical mistakes (GM-16, GM-32) or 16 or 32 referential mistakes (RM-16, RM-32) were inserted in these 80 sample sentences. The results showed that even under the GM-16, RM-16, and RM-32 conditions some subjects could acquire the language much as the subjects under the control condition. These results were discussed from the viewpoint of child language acquisition.

Corresponding author

Correction information

Register with J-STAGE for free!