Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Poster Slams
The role of research design in the reproducibility of L1 and L2 language networks: A review of bilingual neuroimaging meta-analyses.
Poster C11 in Poster Session C, Friday, October 7, 10:15 am - 12:00 pm EDT, Millennium Hall
Lindy Comstock1; 1UCLA
The functional organization of first (L1) and second (L2) language processing in bilinguals remains a topic of great interest to the neurolinguistics community. Today, papers investigating L1 and L2 language processing in bilinguals number in the hundreds, allowing for no less than six neuroimaging meta-analyses devoted to aspects of bilingual language use (Cargnelutti, Tomasino, & Fabbro, 2019; Indefrey, 2006; Liu & Cao, 2016; Luk, Green, Abutalebi, & Grady, 2012; Sebastian, Laird & Kiran, 2011; Sulpizio, Del Maschio, Fedeli, & Abutalebi, 2019). It is well known that meta-analyses are a reliable method to increase the statistical power and generalizability of fMRI findings, such that differences in L1 and L2 processing that may be too subtle to be detected in individual fMRI studies should become apparent in a meta-analysis. But why then is there still little consensus between the reported findings of even the most recently published (and therefore, most highly-powered) meta-analyses? While weaknesses stemming from limitations in the literature, such as the absence of a common method to define and measure descriptive categories like proficiency level, degree of language exposure, and AoA are often cited, an equally plausible explanation lies in a lack of understanding about the technical differences in how meta-analyses are conducted. These details are often given minimal attention, elided from the method section, or are not stated in a manner that the general public may easily recognize or understand. Moreover, the specific nature of many research questions in neurolinguistics has facilitated experimental paradigms that defy easy categorization or comparison, frequently leading to the combination of data types that are not compatible in effort to increase the size of the overall data set. Ultimately, an aggregation of substantially different data types may weaken the validity of meta-analysis findings. This presentation begins with a review of the key factors by which meta-analyses may differ, including data types, task designs, thresholding, and analysis strategy. Next, how these methodological choices can affect the results of a meta-analysis is explained, and a case is made for greater attention to task categorization. Finally, some results of a task-based meta-analysis are presented (Comstock & Oliver, 2021).
Topic Areas: Methods, Multilingualism