Increased Stroop interference with better second-language reading skill

Abstract

Skilled readers demonstrate remarkable efficiency in processing written words, unlike beginning readers for whom reading occurs more serially and places higher demands on visual attention. In the present study, we used the Stroop paradigm to investigate the relationship between reading skill and automaticity, in individuals learning a second language with a different orthographic system. Prior studies using this paradigm have presented a mixed picture, finding a positive, a negative, or no relationship between the size of Stroop interference and reading skills. Our results show that Stroop interference in the second language was positively related to reading skill (when controlled for interference in the first language). Furthermore, interference was positively related to objective but not subjective indices of the amount of exposure to the second language. We suggest that the lack of consistency in the results of earlier studies may be due, at least in part, to these studies looking at Stroop interference in isolation, rather than comparing interference between languages.

Keywords:

• Visual word recognition
• Stroop effect
• Automaticity
• Second-language acquisition

Acknowledgments

The authors wish to extend their thanks to the following persons: Frank Amand for technical support, Stephanie Poot for administrative support, and Naoko Sakurai and Stefanie Nijs for their invaluable help with participant recruitment. W.B. is funded by the Research Foundation–Flanders (FWO-Vlaanderen); N.N. is funded by the Faculty of Arts (University of Leuven); J.W. is funded by METH/08/02 and IMPH/06/GHW; H.O.D.B. is funded by CREA/07/004 and the Human Frontier Science Program Organization.

Notes

¹ This may reflect a ∩-shaped relationship between automaticity (and/or executive functions) and interference (given that reading skills tend to decline in old age). Such a relationship can be explained based on models of the dynamics of temporal overlap between two processes (colour naming and word reading; e.g., Zmigrod & Hommel, 2010). Interference is assumed to be maximal when both processes occur in close temporal proximity (i.e., at intermediate proficiency) and less at either high or low proficiency levels, as word processing occurs either too early or too late to interfere with colour naming.

² Both papers also report within-language interference scores, though no analyses on these scores are reported. These results suggest equivalent Stroop interference for responding to L1 stimuli in L1, compared to responding to L2 stimuli in L2, in Sumiya and Healy (2004); conversely, Stroop interference seems larger for L1 than for L2 in Sumiya and Healy (2008). It is interesting to note that the authors further report higher L2 proficiency in their 2004 study, whereas no difference is apparent between L1 and L2 interference scores.

³ For example, studies using verbal responses suggest a complex relationship between within- and between-language interference, which depends on the orthographic similarity between the words in the two languages (see e.g., Fang, Tzeng, & Alva, 1981; though also see Smith & Kirsner, 1982). While it is unclear to what extent this effect still exists when using manual responses, it is unlikely that in our study there is any “carry-over” interference from one language to the other, as there is no visual similarity between the same word in Japanese (in kanji) and in Dutch.

⁴ The proportion of congruent/incongruent trials has been related to the size of the Stroop effect in numerous studies (with greater interference as the proportion of congruent trials increases). A recent study (Schmidt & Besner, 2008) suggests that this relationship emerges due to response contingencies and not because the proportion of congruent trials modulates attention to the written word form (i.e., the extent to which reading is obligatory). An equal proportion of congruent and incongruent trials is also likely to lead to more reliable estimates of reaction times/accuracy for the congruent condition.

⁵ One participant did not complete either the questionnaire or the kanji test and was excluded from analyses that used these variables. Participants who were not currently enrolled in the Japanese Studies programme were excluded from analyses that used the variable “year of study”.

⁶ The same pattern of results emerges when using median RTs instead of means.