Rules versus statistics in reading aloud

New evidence on an old debate

Conrad Perry, Johannes C. Ziegler, Mario Braun, Marco Zorzi

Research output: Contribution to journalArticle

15 Citations (Scopus)

Abstract

Nonword reading performance, that is, the ability to generate plausible pronunciations to novel items, has probably been the hardest test case for computational models of reading aloud. This is an area where rule-based models, such as the Dual-Route Cascaded (DRC) model, typically outperformed connectionist learning models. However, what is the evidence that people apply rules when reading nonwords? This was investigated in German. Nonwords were created that allowed us to test whether people apply an abstract rule to determine vowel length or whether they would be sensitive to the statistical distribution of vowel length in the mental lexicon. The human data showed a great amount of variability in nonword pronunciations. Simulations of these nonwords, where the DRC was contrasted with a fully implemented and freely available German version of the connectionist dual process model (German-CDP+), a model that learns the statistical mapping between spelling and sound, showed that CDP+ provided a better account of the data than the DRC. These results support the view that rule based models may simply approximate patterns of language use rather than provide an accurate description of the underlying cognitive machinery.

Original languageEnglish
Pages (from-to)798-812
Number of pages15
JournalEuropean Journal of Cognitive Psychology
Volume22
Issue number5
DOIs
Publication statusPublished - Aug 2010

Fingerprint

Cytidine Diphosphate
Reading
Statistical Distributions
Aptitude
Neural Networks (Computer)
Statistical Models
Language
Learning

Keywords

  • Computational modelling
  • Connectionism
  • German
  • Reading Aloud
  • Rules

ASJC Scopus subject areas

  • Experimental and Cognitive Psychology

Cite this

Rules versus statistics in reading aloud : New evidence on an old debate. / Perry, Conrad; Ziegler, Johannes C.; Braun, Mario; Zorzi, Marco.

In: European Journal of Cognitive Psychology, Vol. 22, No. 5, 08.2010, p. 798-812.

Research output: Contribution to journalArticle

Perry, Conrad ; Ziegler, Johannes C. ; Braun, Mario ; Zorzi, Marco. / Rules versus statistics in reading aloud : New evidence on an old debate. In: European Journal of Cognitive Psychology. 2010 ; Vol. 22, No. 5. pp. 798-812.
@article{d19405025d5343d289e6a558134fc9b6,
title = "Rules versus statistics in reading aloud: New evidence on an old debate",
abstract = "Nonword reading performance, that is, the ability to generate plausible pronunciations to novel items, has probably been the hardest test case for computational models of reading aloud. This is an area where rule-based models, such as the Dual-Route Cascaded (DRC) model, typically outperformed connectionist learning models. However, what is the evidence that people apply rules when reading nonwords? This was investigated in German. Nonwords were created that allowed us to test whether people apply an abstract rule to determine vowel length or whether they would be sensitive to the statistical distribution of vowel length in the mental lexicon. The human data showed a great amount of variability in nonword pronunciations. Simulations of these nonwords, where the DRC was contrasted with a fully implemented and freely available German version of the connectionist dual process model (German-CDP+), a model that learns the statistical mapping between spelling and sound, showed that CDP+ provided a better account of the data than the DRC. These results support the view that rule based models may simply approximate patterns of language use rather than provide an accurate description of the underlying cognitive machinery.",
keywords = "Computational modelling, Connectionism, German, Reading Aloud, Rules",
author = "Conrad Perry and Ziegler, {Johannes C.} and Mario Braun and Marco Zorzi",
year = "2010",
month = "8",
doi = "10.1080/09541440902978365",
language = "English",
volume = "22",
pages = "798--812",
journal = "European Journal of Cognitive Psychology",
issn = "0954-1446",
publisher = "Psychology Press Ltd",
number = "5",

}

TY - JOUR

T1 - Rules versus statistics in reading aloud

T2 - New evidence on an old debate

AU - Perry, Conrad

AU - Ziegler, Johannes C.

AU - Braun, Mario

AU - Zorzi, Marco

PY - 2010/8

Y1 - 2010/8

N2 - Nonword reading performance, that is, the ability to generate plausible pronunciations to novel items, has probably been the hardest test case for computational models of reading aloud. This is an area where rule-based models, such as the Dual-Route Cascaded (DRC) model, typically outperformed connectionist learning models. However, what is the evidence that people apply rules when reading nonwords? This was investigated in German. Nonwords were created that allowed us to test whether people apply an abstract rule to determine vowel length or whether they would be sensitive to the statistical distribution of vowel length in the mental lexicon. The human data showed a great amount of variability in nonword pronunciations. Simulations of these nonwords, where the DRC was contrasted with a fully implemented and freely available German version of the connectionist dual process model (German-CDP+), a model that learns the statistical mapping between spelling and sound, showed that CDP+ provided a better account of the data than the DRC. These results support the view that rule based models may simply approximate patterns of language use rather than provide an accurate description of the underlying cognitive machinery.

AB - Nonword reading performance, that is, the ability to generate plausible pronunciations to novel items, has probably been the hardest test case for computational models of reading aloud. This is an area where rule-based models, such as the Dual-Route Cascaded (DRC) model, typically outperformed connectionist learning models. However, what is the evidence that people apply rules when reading nonwords? This was investigated in German. Nonwords were created that allowed us to test whether people apply an abstract rule to determine vowel length or whether they would be sensitive to the statistical distribution of vowel length in the mental lexicon. The human data showed a great amount of variability in nonword pronunciations. Simulations of these nonwords, where the DRC was contrasted with a fully implemented and freely available German version of the connectionist dual process model (German-CDP+), a model that learns the statistical mapping between spelling and sound, showed that CDP+ provided a better account of the data than the DRC. These results support the view that rule based models may simply approximate patterns of language use rather than provide an accurate description of the underlying cognitive machinery.

KW - Computational modelling

KW - Connectionism

KW - German

KW - Reading Aloud

KW - Rules

UR - http://www.scopus.com/inward/record.url?scp=77954968378&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77954968378&partnerID=8YFLogxK

U2 - 10.1080/09541440902978365

DO - 10.1080/09541440902978365

M3 - Article

VL - 22

SP - 798

EP - 812

JO - European Journal of Cognitive Psychology

JF - European Journal of Cognitive Psychology

SN - 0954-1446

IS - 5

ER -