Poster D12, Friday, August 17, 4:45 – 6:30 pm, Room 2000AB

Nonword repetition recruits distinct and overlapping nodes of language and working memory networks

Terri L. Scott1, Sara C. Dougherty1, Ja Young Choi1,2, Tyler K. Perrachione1;1Boston University, 2Harvard University

Phonological working memory (PWM) is the process by which we temporarily maintain representations of speech sounds in short-term memory. This ability is believed to be important for language and reading acquisition, and is measured clinically using tests of nonword repetition. The operationalization of PWM is not always clearly dissociated from verbal working memory (VWM), which is theorized to be supported by a phonological loop independent from core language processing (Baddeley & Hitch 1974; Baddeley 1986, 2003). However, more recent studies propose that phonological working memory necessarily recruits core language regions otherwise specifically implicated in encoding phonological information (McGettigan et al. 2010; Barry et al. 2011; Perrachione et al. 2017; Scott et al. 2018). Using functional magnetic resonance imaging (fMRI), this work explores how neural structures responsive to PWM-load are differentially sensitive to modulation by language and VWM. Twenty adult participants (12 female; age 19-32, M=24.1 years) underwent functional magnetic resonance imaging (fMRI) while completing three separate tasks designed to engage PWM (nonword repetition), VWM (digit span), or language processing (auditory language localizer; Scott et al. 2017). Brain regions modulated by working memory were localized by contrasting high (long nonwords/digit sequences) vs. low (short nonwords/sequences) working memory loads. The language processing task required participants to listen to clips of intact speech, contrasted with unintelligible degraded speech. Nonword repetition activation was measured during a sparse-sampling block design fMRI (TR=2.25s, TA=0.75s, 3mm isotropic, 45 slices, 5 simultaneous slices). Digit span and the language localizer activation was measured during continuous-sampling fMRI (TR=0.75s, TA=0.75s, 3mm isotropic, 45 slices, 5 simultaneous slices). Group-constrained subject-specific (GSS; Fedorenko et al. 2010; Julian et al. 2012) analyses were employed to define and interrogate functional regions of interest (fROIs) in individual subjects. We previously identified five regions (bilateral STG, L-PT, L-PreCG, and right cerebellum) that are commonly sensitive to PWM load in our participants (Scott et al. 2018). Of these regions, only L-PreCG and L-PT both showed similar patterns of activity during language and VWM (L-PreCG: r=0.42 ± 0.08; L-PT: r=0.37 ± 0.09). fROIs were defined within these regions based upon each subject’s top 10% of voxels in the critical contrast of high vs. low PWM load and then responses to the critical contrasts for language and digit span were measured. Responses to language were found to be significantly larger than VWM in both L-STG (p << 0.01) and R-STG (p << 0.01). By examining the convergences and divergences of task activation in language processing, VWM, and PWM, we show that, of the regions commonly activated during PWM, L-PreCG and L-PT evince highly similar patterns of activity during language and VWM, whereas significant differences between language and VWM were measured in bilateral STG. Taken together; these results suggest that PWM processing involves a combination of core language sensitive regions in STG with dual language/VWM convergent regions in the dorsal stream.

Topic Area: Phonology and Phonological Working Memory

Back