The word frequency effect in first- and second-language word recognition: A lexical entrenchment account

Abstract

We investigate the origin of differences in the word frequency effect between native speakers and second-language speakers. In a large-scale analysis of English word identification times we find that group-level differences are fully accounted for by the individual language proficiency scores. Furthermore, exactly the same quantitative relation between word frequency and proficiency is found for monolinguals and three different bilingual populations (Dutch–English, French–English, and German–English). We conclude that the larger frequency effects for second-language processing than for native-language processing can be explained by within-language characteristics and thus need not be the consequence of “being bilingual” (i.e., a qualitative difference). More specifically, we argue that language proficiency increases lexical entrenchment, which leads to a reduced frequency effect, irrespective of bilingualism, language dominance, and language similarity.

Keywords:

• Bilingualism
• Word frequency
• Proficiency
• Lexical quality
• Quantitative analysis

Notes

¹ It is customary in IA models to scale resting levels between –0.92 (minimum ∼ lowest word frequency) and 0 (maximum ∼ highest word frequency; see McClelland & Rumelhardt, 1981). In the present simulation, we lowered the scale by multiplication with 1.5. The range thus became –1.38 to 0.

² It is noteworthy that Lemhöfer et al. (2008) found no difference between monolinguals and bilinguals with respect to spoken frequency. These frequencies—taken from the BNC corpus—were calculated on the basis of a relatively small sample, however (i.e., 124 individuals), potentially leading to a high level of idiosyncrasy in the measure.

³ We would like to thank an anonymous reviewer for pointing this out.