The role of language proficiency, cognate status and word frequency in the assessment of Spanish–English bilinguals’ verbal fluency

Abstract

Purpose: Assessment tools are needed to accurately index performance in bilingual populations. This study examines the verbal fluency task to further establish the relative sensitivities of letter and category fluency in assessing bilingual language skills in Spanish–English bilinguals.

Method: English monolinguals and Spanish–English bilinguals had 1 minute to name words belonging to a category (e.g. animals) or starting with a letter (e.g. A). Number of words retrieved, proficiency, cognate and frequency effects were examined.

Result: In their dominant language (English), bilinguals and monolinguals showed similar fluency patterns, generating more words in category than letter tasks. This category advantage disappeared for bilinguals tested in their non-dominant language (Spanish). Further, bilinguals retrieved a higher percentage of cognates (e.g. lagoon-laguna) than monolinguals across tasks and languages. In particular, as proficiency increased in their non-dominant language, bilinguals were more likely to produce cognates (including cognates with lower word frequencies).

Conclusion: While bilinguals and monolinguals performed largely the same, bilinguals showed fine-grained differences from monolinguals in both their dominant and non-dominant languages. Based on these results, it is recommended that clinicians evaluate findings from bilinguals’ verbal fluency tasks with attention to language proficiency, cognate words produced and relative to normative data that match their clients’ language histories.

Keywords:

• Verbal fluency
• bilingual assessment
• language proficiency

Acknowledgement

We would like to thank members of the Bilingualism and Cognition Laboratory at San Diego State University, in particular Anna Zak, Roxanna Palma, Rhiann Grills, and Sarah Stoelb, and the Bilingualism and Psycholinguistics Research Group at Northwestern University, for research assistance and helpful discussions of this work. This project was supported in part by SDSU Grant 242338 and an ASHFoundation New Investigator Grant to HKB, and by Grant NICHD1R01HD059858 to VM.

Declaration of interest

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.

Notes

1 The four excluded participants, who were outliers in verbal fluency performance beyond 2.5 SD of the mean, included one monolingual and three bilinguals. The bilinguals also differed from other bilinguals’ linguistic profiles because one had learned Gujarati as a child, one spoke Yucatecan Spanish and one spoke Venezuelan Spanish. The monolingual did not deviate from other monolinguals’ background profiles and no other demographic characteristics of the excluded participants deviated from characteristics of included participants. Examination of means suggested similar patterns in the excluded participants as in the overall sample. Because outliers tend to skew the distribution, it is common practice to exclude individuals who deviate from the group mean by more than 2.5 SD to reduce the impact of the outlier in data-sets. The practice of removing outliers has been shown not to artificially inflate p-values across experiments (Bakker & Wicherts, 2014).

2 After exclusion of outlying participants, comparable yet slightly uneven distributions in counterbalancing remained. For letter sets, 12 bilinguals (48%) and 10 monolinguals (42%) received the EPM set in English; 13 bilinguals (52%) and 14 monolinguals (58%) received the ALC set. For category sets, 13 bilinguals (52%) and 15 monolinguals (62.5%) received Animals-Fruits-Clothes cues; 12 bilinguals (48%) and nine monolinguals (37.5%) received Vegetables-Colours cues. Finally, of the higher-proficiency bilinguals, five (42%) received the EPM cues in English, seven (58%) received the ALC cues, seven (62.5%) received the Animals-Fruits-Clothes cues and five (48%) received the Vegetables-Colours cues. Of the lower-proficiency bilinguals, six (50%) received the EPM cues, six (50%) received the ALC cues, six (50%) received the Animals-Fruits-Clothes cues and six (50%) received the Vegetables-Colours cues. Table II illustrates the similarity of performance across sets.

3 Frequencies were drawn from comparable English and Spanish corpora, making such an analysis possible (English spoken words based on a 450 million word corpus and Spanish spoken words based on a 100 million word corpus, with similar sources for both, Davies, 2002, 2008).