Home > Research > Publications & Outputs > Mapping the predictors of single word recognition

Electronic data

  • BPS_Presentation

    Other version, 2.17 MB, PDF document

    Available under license: None

View graph of relations

Mapping the predictors of single word recognition: a research synthesis

Research output: Contribution to conference - Without ISBN/ISSN Speech

Published
Publication date30/08/2017
<mark>Original language</mark>English
EventBPS Cognitive Section Conference - Newcastle University, Newcastle, United Kingdom
Duration: 30/08/20171/09/2017

Conference

ConferenceBPS Cognitive Section Conference
CountryUnited Kingdom
CityNewcastle
Period30/08/171/09/17

Abstract

This research synthesis examines 77 reports that have manipulated psycholinguistic variables across contrasting groups of adults or children in word naming and/or lexical decision tasks. Using a random-effects model, meta-analysed effect sizes (Pearson’s r and odds ratios) for frequency, length, consistency, neighbourhood size, age-of-acquisition, imageability and concreteness range from moderate to large for response time and accuracy data.

For lexical decision accuracy scores, the trend is for adults to show stronger effect sizes. In word naming tasks for accuracy, children tend to show stronger effect sizes. For response time data across both tasks, children also tend to show stronger effect sizes.
Adult accuracy appears to be more dependent upon phonological and orthographical properties than semantic properties, however, semantic properties appear to play a role in response times. In contrast, semantic properties of words show a stronger effect in child samples for both accuracy and response time across word naming and lexical decision.

There is a cautionary note, however: confidence intervals are wide and accompanying heterogeneity statistics show very high values, indicating the presence of measurement error as well as expected sampling variation. Differences in experimental design, sample selection and choices for statistical analysis may all serve to inflate the summary effect sizes. Going forward, methods for treating this inflation are suggested and protocols to systematically reduce the heterogeneity are discussed.