You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

Poster B52, Wednesday, November 8, 3:00 – 4:15 pm, Harborview and Loch Raven Ballrooms

Converging evidence from univariate and multivariate fMRI analyses suggests a phonological buffer in the left supramarginal gyrus

Qiuhai Yue1, Randi C. Martin1, A. Cris Hamilton1, Nathan S. Rose2;1Rice University, Houston, TX, USA, 2University of Notre Dame, Notre Dame, IN, USA

Phonological short-term memory (pSTM) refers to the capacity for retaining speech sounds for a short period of time. The current study evaluated two theories of pSTM: the buffer account, which proposes a dedicated pSTM store (e.g., Baddeley, 2000), and the embedded processes account, which argues that pSTM consists of the activated portion of phonological long-term memory (LTM) (e.g., Cowan, 2001). Previous findings from patients showing poor phonological STM with preserved word recognition support the buffer account (Warrington & Shallice, 1969; Vallar & Baddeley, 1984), and lesion localization suggests that the left inferior parietal lobe supports the phonological buffer (Shallice & Vallar, 1990). Evidence from neuroimaging studies on healthy subjects has been mixed, with some showing sustained delay-period activity in regions outside those for the LTM representation of speech sounds (Paulesu et al., 1993) but others showing sustained activity solely in LTM regions (e.g., superior temporal lobe) (Ravizza et al., 2011). However, most studies have used visually presented verbal stimuli (e.g., Ravizza et al., 2011), making it unclear whether subjects relied on phonological retention. In addition, recent studies on visual STM using a multivariate approach (e.g., multivariate pattern analysis; MVPA) have complicated the interpretation of delay-period activity, with some studies showing that visual mnemonic information could be decoded from LTM regions during the delay period, even though no sustained activity was observed (Riggall & Postle, 2012). The present study addressed the limitations of previous studies by using auditory stimuli, and employing both univariate and multivariate analyses. We scanned 15 healthy young adults while they performed an immediate discrimination task (perception) and a delayed recognition task (STM). In both, we manipulated the stimulus type (nonwords vs. chords), and for the STM task we manipulated memory load (one vs. three-items). Using the perception task, a region in the left superior temporal cortex was found to support speech recognition. However, contrary to the predictions of the embedded processes account, this region failed to show a memory load effect, or any sustained activation, during the delay period. Moreover, MVPA decoding during the delay period was unsuccessful in this region by a perception classifier or an encoding classifier. While successful decoding was observed with a maintenance classifier, decoding accuracy was unrelated to behavioral performance. In contrast, consistent with the buffer approach, univariate analyses revealed a set of fronto-parietal regions, including the left supramarginal gyrus, which showed sustained activity and load effects during the delay period. Using MVPA, stimulus decoding was successful with both encoding and maintenance classifiers and, in the left supramarginal gyrus, decoding accuracy was related to verbal but not nonverbal memory performance. In addition, a functional connectivity analysis found that, as memory load increased, the left temporal cortex involved in speech perception became more strongly connected with the left inferior parietal region involved in maintenance. Thus, converging evidence from both univariate and multivariate approaches provided greater support for a buffer than an embedded processes account for pSTM, with a region in the left supramarginal gyrus serving as a buffer for maintaining phonological representations.

Topic Area: Phonology and Phonological Working Memory

Back to Poster Schedule