The locus of the orthographic consistency effect in auditory word recognition: Further evidence from French

Abstract

Ventura, Morais, Pattamadilok, and Kolinsky (2004) found, for spoken Portuguese words, an orthographic consistency effect in lexical decision but not in standard shadowing (on-line repetition): words ending with phonological rimes that have several spellings led to longer decision times than words ending with phonological rimes that have only one spelling. This pattern of results was replicated in this study, using French, a language presenting a much higher degree of orthographic inconsistency than Portuguese. The observation of systematic associations between the effects of word orthographic consistency, word frequency and lexicality supports the hypothesis that lexical processing is critical to the occurrence of the consistency effect. Finally, the comparison of the word consistency effects obtained in French and in Portuguese suggests that their size depends on the overall consistency of the language's orthographic code.

Acknowledgements

Preparation of this article was supported by the Human Frontier Science Program RGP 53/2002 ‘An interdisciplinary approach to the problem of language and music specificity’, by an FRFC convention (2.458.02F), by a joint grant of Fundação para a Ciência e a Tecnologia – Ministério da Ciência, Tecnologia e Ensino Superior – and European Community FEDER funding (Project POCI/PSI/56901/2004, ‘Visual phonology and auditory orthography’) and by the Centro de Psicologia Clínica e Experimental – Desenvolvimento, Cognição e Personalidade of the Universidade de Lisboa, Portugal. The first and the last authors are Research Fellow and Senior Research Associate, respectively, of the Belgian Fonds National de la Recherche Scientifique (F.N.R.S.). We thank Johannes Ziegler and two anonymous reviewers for helpful discussions and comments on an earlier version of this article.

Notes

¹Throughout the paper, the term ‘rime’ refers to the phonological rime and the term ‘body’ refers to the orthographic rime.

²The outliers were identified on the basis of the RT distribution. They corresponded to the values that were smaller (or larger) than 1.5 interquartile range from the score at the 25th percentile (or the 75th percentile). The same criterion was used in all subsequent experiments.

³As suggested by an anonymous reviewer, one could reduce the impact of extreme scores without removing data by performing the analyses on the median instead of the mean RTs. We also did this. The results obtained in these analyses were identical to those obtained in the analyses performed on the mean RTs (when the three outliers: /fras/, /tryr/, and /kluf/ were not excluded from the analyses) with the consistency effect for pseudowords being significant only in the analysis by subjects, t ₁(25) = 2.17, p<.05; t ₂=1.

⁴The analyses performed on the median RTs of pseudowords also showed an unreliable consistency effect, t ₁(26) = 4.38, p<.0001; t ₂<1.

⁵The pseudoword mean duration was considered as an important factor according to the assumption that participants had to wait until the end of the stimuli to decide whether or not they were pseudowords. This was indeed the case for 72 (out of 80) pseudowords for which the deviation point is at the last phoneme.

⁶The analyses performed on the median RTs of pseudowords also showed an unreliable consistency effect, t ₁(24) = 3.6, p=.001; t ₂<1.

⁷We chose to compare the effect size obtained in different tasks rather than to enter task as a between-subject factor in an ANOVA and look for an interaction because, contrary to the latter method, the effect size comparison is independent of the variability and the distribution of the data inherent to each task. In so doing, the data obtained in different tasks or with different materials could be directly compared even though their distributions or error terms are not equivalent. However, we also performed an ANCOVA taking into account task as a between-subject factor and RT on fillers as covariate (given that the overall RTs obtained in lexical decision were longer than those obtained in shadowing). The analysis showed a significant interaction between consistency and task, F ₁(1, 75) = 75.27, p<.0001; frequency and task, F ₁(1, 75) = 5.37, p<.025; and consistency, frequency, and task, F ₁(1, 75) = 5.38, p<.025.