A roommate in cream: Morphological family size effects on interlingual homograph recognition

Abstract

In monolingual studies, target word recognition is affected by the number of words that are morphologically related to the target. Larger morphological families lead to faster recognition. We investigated the role of the morphological family size (MFS) effect in bilingual word recognition. First, re-analysis of available English lexical decision data from Dutch–English bilinguals reported by Schulpen, Dijkstra, and Schriefers (2003) revealed a facilitatory English MFS effect in purely English words and in Dutch–English interlingual homographs (such as ROOM, a word that exists both in English and in Dutch, where it means ‘cream’). For interlingual homographs, the Dutch MFS simultaneously induced inhibitory effects, supporting a language non-selective access process. The MFS effect was independent of the relative frequency of the two readings of the homographs. Task-dependence of the MFS effect was demonstrated in generalized Dutch–English lexical decision data, which led to facilitatory effects of both families. Finally, the pervasiveness of the MFS effect was demonstrated in a Dutch lexical decision task performed by the same type of bilinguals. Facilitatory effects of Dutch MFS were found for Dutch monolingual words and interlingual homographs, which were also affected by inhibitory effects of English MFS. The results are discussed in relation to the task-sensitive BIA+ model of bilingual word recognition.

Acknowledgments

The authors wish to thank Marc Brysbaert and one anonymous reviewer for helpful comments and suggestions on a previous version of this manuscript. The second, fourth, and fifth authors have been supported by the Dutch Research Council (NWO) through a PIONIER grant to R. H. Baayen. During the last stages of preparation of the manuscript, the second author received additional support from the Medical Research Council (UK) and the European Community, under the ‘Information Society Technologies Programme’ (IST-2001-35282).

Notes

¹Dutch frequency appeared to be significantly non-lognormally distributed according to the very conservative Shapiro-Wilk test (W=0.97, p=.02). However, inspection of a quantile-quantile plot showed that this deviation was quite small.

²Error data from lexical decision experiments are strongly non-normally distributed, due to the very small numbers of incorrect responses provided by participants. This violates the normality precondition for a traditional ANOVA. Instead, we use a logistic regression analysis which is less sensitive to the deviation from normality.

³Although Dutch frequency appeared to be significantly non-lognormally distributed, visual inspection of a quantile-quantile plot showed that the deviation from lognormality was very small. Moreover, separate normality tests revealed that neither the subset of interlingual homographs (W=0.97, p>.05), nor the subset of Dutch monolingual words (W=0.98, p=.60), deviated significantly from lognormality.

⁴Although this correlation was only marginally significant, it reached full significance by a non-parametric Spearman correlation (r _s =.37, p=.02).