Abstract
There are conflicting ideas about the neural basis of short-term memory (STM) storage. Recent neuroimaging studies using multivariate pattern analyses (MVPA) reported evidence for visual STM storage of gratings, colors, motion and even complex visual patterns in content-selective occipital and parietal regions, which provides evidence against the classical model of predominant STM storage in prefrontal cortex. Here we studied how Chinese Characters are encoded in the human brain during short-term memory. Chinese characters have a simple pronunciation (one syllable per symbol) but also a complex visual appearance, which strongly encourages the use of verbal memorization strategies. In each trial, Chinese native speakers were instructed to memorize one out of two characters using a retro-cue. After a prolonged delay, they were tasked with identifying this character amongst a series of partly obscured test characters. Questionnaire data provided evidence for the predominant use of verbal strategies. Patterns of fMRI activity during maintenance were analyzed using searchlight-based MVPA (cvMANOVA) to identify brain regions encoding content-specific information about Chinese symbols. We found three brain regions containing significant stimulus information during the delay: the anterior Broca's area (ABA), the left premotor cortex (lPMC), and early visual cortex (EVC). EVC, however, also showed similar levels of information for not remembered stimuli indicating no involvement in memory storage. Both ABA and lPMC were lateralized to the left hemisphere and showed little content-specific information regarding other complex visual stimuli like color or motion patterns. ABA, but also left lPMC, have been previously implicated in language comprehension and production. In summary, our study provided evidence for verbal STM storage of Chinese symbols in language-related regions, which supports a distributed network model of STM storage and might constitute the neural substrate of Baddeley's 'phonological loop'.
Meeting abstract presented at VSS 2017