A computational analysis of semantic structure in bilingual verbal fluency performance Vanessa Taler a,b,⇑ , Brendan T. Johns c , Katherine Young a,b , Christine Sheppard a,b , Michael N. Jones c a University of Ottawa, Canada b Bruyère Research Institute, Canada c Indiana University, Bloomington, United States article info Article history: Received 13 August 2012 revision received 19 August 2013 Available online 26 September 2013 Keywords: Verbal fluency Bilingualism Semantic memory Computational modeling abstract Groups of English monolingual and English–French bilingual participants completed letter and category fluency tasks, either only in English (monolinguals) or in English, French, free- switch and forced-switch conditions (bilinguals). Response patterns were modeled using a semantic space approach that estimates the weight of frequency and semantic similarity information in determining output patterns. Overall, semantic similarity had a stronger influence on output patterns than did frequency. In the forced English–French switching condition, the weight of similarity information was reduced and the weight of frequency information was increased, suggesting that the increased executive demands related to language switching result in alterations in the semantic structure of fluency output. More- over, the frequency and similarity model parameters were negatively correlated in all tasks, indicating that they may be in competition during verbal fluency tasks. Ó 2013 Elsevier Inc. All rights reserved. Introduction Bilingualism is extremely prevalent, with more than 50% of the world’s population being bilingual or multilin- gual (Grosjean, 2008). Recent research has demonstrated differences in cognitive function between bilinguals and monolinguals, with bilinguals exhibiting lower perfor- mance than monolinguals in language-related tasks, but better performance on tasks of executive control (for a re- view, see Bialystok, 2009; Bialystok, Craik, & Luk, 2008). Verbal fluency tasks are among the most commonly used measures to assess language function in neuropsychologi- cal testing. These tasks require the participant to generate as many words as possible according to a specific criterion within a given time period (typically 1 min). Most com- monly, the criterion is either a phonemic cue (e.g., words starting with the letter F) or a category (e.g., animals). Ver- bal fluency requires both language and executive function; the subject must organize verbal retrieval and recall, initi- ate responses, monitor prior responses, and inhibit inap- propriate responses (Henry, Crawford, & Phillips, 2004). However, the demands on semantic knowledge vary depending on the criterion. Category fluency intrinsically requires rapid access to semantic knowledge. Letter flu- ency, in contrast, can be performed without access to semantic knowledge (Rohrer, Salmon, Wixted, & Paulsen, 1999), although output is typically influenced by semantic organization to some extent in healthy adults (Schwartz, Baldo, Graves, & Brugger, 2003). One influential approach to analyzing fluency output examines measures of clustering and switching (Troyer, Moscovitch, & Winocur, 1997; Troyer, Moscovitch, Wino- cur, Leach, & Freedman, 1998). Clustering refers to the grouping together of items from a given subcategory; for example, a participant may produce a subgroup of farm 0749-596X/$ - see front matter Ó 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jml.2013.08.004 ⇑ Corresponding author. Address: School of Psychology, University of Ottawa, 136 Jean Jacques Lussier, Vanier Hall, Ottawa, Ontario K1 N 6N5, Canada. E-mail address: vtaler@uottawa.ca (V. Taler). Journal of Memory and Language 69 (2013) 607–618 Contents lists available at ScienceDirect Journal of Memory and Language journal homepage: www.elsevier.com/locate/jml