본문 바로가기
읽기 · 학술 · 공개 라이선스 원문

Neural Correlates of Verbal Working Memory: An fMRI Meta-Analysis

Emch, von Bastian, Koch · 2019

공개 학술지 Frontiers in Human Neuroscience 에 실린 총설 논문입니다(2019년). CC BY 4.0 으로 공개된 글이고, 그림 · 표 · 참고 문헌 표시를 뺀 본문을 실었습니다.

출처 Emch et al. (2019), Frontiers in Human Neuroscience · 라이선스 CC BY 4.0 · 원문

장 3개 올림 이 장 1981낱말 TOEFL® 읽기 문제 5문항
다음 장 →

이 장은 원문만 있습니다. 문장별 해석은 앞 장부터 차례로 붙입니다.

1장 · Neural Correlates of Verbal Working Memory: An fMRI Meta-Analysis

Verbal Working memory (vWM) capacity measures the ability to maintain and manipulate verbal information for a short period of time. The specific neural correlates of this construct are still a matter of debate. The aim of this study was to conduct a coordinate-based meta-analysis of 42 fMRI studies on visual vWM in healthy subjects (n = 795, males = 459, females = 325, unknown = 11; age range: 18–75). The studies were obtained after an exhaustive literature search on PubMed, Scopus, Web of Science, and Brainmap database. We analyzed regional activation differences during fMRI tasks with the anisotropic effect-size version of seed-based d mapping software (ES-SDM). The results were further validated by performing jackknife sensitivity analyses and heterogeneity analyses. We investigated the effect of numerous relevant influencing factors by fitting corresponding linear regression models. We isolated consistent activation in a network containing fronto-parietal areas, right cerebellum, and basal ganglia structures. Regarding lateralization, the results pointed toward a bilateral frontal activation, a left-lateralization of parietal regions and a right-lateralization of the cerebellum, indicating that the left-hemisphere concept of vWM should be reconsidered. We also isolated activation in regions important for response inhibition, emphasizing the role of attentional control in vWM. Moreover, we found a significant influence of mean reaction time, load, and age on activation associated with vWM. Activation in left medial frontal gyrus, left precentral gyrus, and left precentral gyrus turned out to be positively associated with mean reaction time whereas load was associated with activation across the PFC, fusiform gyrus, parietal cortex, and parts of the cerebellum. In the latter case activation was mainly detectable in both hemispheres whereas the influence of age became manifest predominantly in the left hemisphere. This led us to conclude that future vWM studies should take these factors into consideration.

Introduction

Working memory (WM) is a cognitive system that holds information available that is needed for complex cognition in the present moment (Baddeley, 2010; Oberauer and Hein, 2012). It has been shown that WM capacity is a strong predictor of a wide range of complex cognitive tasks such as analytic problem solving, language acquisition, and reading comprehension (Daneman and Carpenter, 1980; Engle et al., 1999; Wiley and Jarosz, 2012). There have been several attempts to understand the organization of human WM. The arguably most influential model is the multiple-component model proposed by Baddeley and Hitch (1974). The authors hypothesized the existence of a “central executive” component, which controls the incoming information and passes the information to two subsystems: the “phonological loop” and the “visuospatial sketchpad.” Within the phonological loop, due to the interplay of its two components—the phonological store and the articulatory loop—the verbal material representation can be kept in an active state. Verbal information is processed in perceptual systems before it enters the phonological loop in which it is temporarily stored in the phonological store and maintained through the articulatory loop using subvocal rehearsal of the information. In addition to subvocal rehearsal, the articulatory loop is also thought to be involved whenever verbal information is presented visually: whereas auditory verbal information (e.g., spoken words) can directly enter the phonological store, visually presented verbal information (e.g., written words) must first be recoded into phonological information. In other words, subvocalization is necessary in order to reroute visually derived verbal material into the phonological store (Buchsbaum and D'Esposito, 2008). The visuospatial sketchpad is responsible for integrating visual and spatial information. Later, the “episodic buffer” was added (Baddeley, 2000). It binds the information from the different subsystems into integrated episodes. Alternative models proposed that WM holds any type of information in a state of heightened availability (Oberauer, 2010; Cowan et al., 2012) whereas others models have emphasized on the role of attentional control in WM (e.g., Kane and Engle, 2003; Unsworth and Engle, 2007). These different theoretical conceptualizations of WM are not necessarily mutually exclusive (Cowan et al., 2012), with common features including a variety of processes such as encoding, maintaining and retrieving information of various domains (e.g., letters, geometric forms, or words), and some attentional control mechanism that supports dealing with interference from irrelevant or distracting information. Thus, the neural correlates of WM may vary depending on the processes, the type of information, and the modality of stimulation (auditory or visual). Given the variety across studies with regard to WM domain and the lack of process differentiation in most studies, the present meta-analysis focused exclusively on visually presented verbal working memory (vWM) across all processes involved in WM.

Visual Verbal Working Memory

Several fMRI studies over the past years have specifically investigated the brain areas involved in vWM (Honey et al., 2000; Veltman et al., 2003; Chen and Desmond, 2005; Narayanan et al., 2005; Wolf et al., 2006). They basically corroborated the general notion that a variety of brain networks are activated during vWM mainly including areas in the prefrontal cortex (PFC) and the parietal cortex as well as cerebellar and basal ganglia regions (Paulesu et al., 1993; Petrides et al., 1993; Desmond et al., 1997; Crosson et al., 1999; Lewis et al., 2004; Chang et al., 2007; Buchsbaum et al., 2011; Thürling et al., 2012; Moore et al., 2013; Chai et al., 2018). Previous meta-analyses have indicated that the left PFC might be predominantly involved in vWM processes whereas the right PFC seems to be more strongly involved in spatial WM, leading to a lateralization of this region due to different input (Wager and Smith, 2003; Owen et al., 2005). However, there is no general consensus on the functional organization of the PFC (Eriksson et al., 2015). Functional neuroimaging studies suggested that the articulatory loop is associated with the left inferior frontal cortex—where Broca's area is located –, left supplementary motor area (SMA), left premotor cortex (BA6), and left insula. The phonological store has been shown to be associated with the left BA 40, corresponding to the left supramarginal gyrus located in the left inferior parietal lobule. Thus, these regions are essential for any kind of vWM task (Paulesu et al., 1993; Smith and Jonides, 1998; Henson et al., 2000; Buchsbaum and D'Esposito, 2008). Moreover, parietal activation has been interpreted as a buffer for modality-specific information. Whereas, the relevance of prefrontal and parietal regions for vWM has long been recognized, the cerebellum came into focus only some years ago. Originally regarded mainly as a structure involved in motor control and coordination, its involvement in higher-order cognitive processes, such as vWM, is no longer called into question (Ravizza et al., 2006; Hayter et al., 2007; Cooper et al., 2012; Thürling et al., 2012; Tomlinson et al., 2014). More specifically, it has been suggested that the cerebellum plays a relevant role in subvocal rehearsal, but the specific contribution of the cerebellum to the various processes involved in vWM is still a matter of debate (Desmond et al., 2003; Pleger and Timmann, 2018). Like the cerebellum, the basal ganglia (BG) are critical structures for motor control by enhancing desired motor behaviors and suppressing undesired ones (Alexander et al., 1986; Mink, 1996). In addition, the BG are involved in various cognitive processes, such as language production and working memory (McNab et al., 2008). Again, for many years, fMRI studies on vWM tended to focus on cortical structures such as parietal and frontal regions, underestimating the relevance of BG structures such as caudate, putamen and globus pallidus. Finally, limbic areas, such as cingulate, are known to be involved in vWM, but its contribution has likewise long been underestimated (Moore et al., 2013).

Influencing Factors in the Neural Correlates of vWM

Activation in these brain regions can be influenced by several factors, such as age, gender, and type and difficulty (i.e., WM load) of the fMRI task. Moreover, the activation can be assumed to depend on individual performance (e.g., response velocity/speed as assessed by mean response times) and the statistical threshold which analyses are based on.

Age

Older adults compared to younger adults have been found to show a more bilateral pattern of prefrontal cortex activity under comparable task demands, a finding which constituted the basis of the Hemispheric Asymmetry Reduction in Older Adults (HAROLD) model (Cabeza, 2002; Cabeza et al., 2004). One hypothesis is that bilateral activity in older adults could reflect a functional compensatory mechanism, in which age-related asymmetry reductions compensate neurocognitive decline leading to a less lateralized brain activity. This is known as the compensation view. A second hypothesis is the so-called dedifferentiation view which assumes a less specific recruitment of neural networks due to gradual changes occurring with age. In a PET study, Reuter-Lorenz et al. (2000) showed that PFC activity in younger adults was left lateralized for verbal and right lateralized for spatial stimuli, whereas older adults presented a bilateral PFC activation for verbal and visual tasks. This model is not only supported by functional neuroimaging results but also by behavioral results from a letter matching task (Reuter-lorenz et al., 1999). Apart from these models, a number of other theories related to age differences in brain activation have been proposed, such as the Posterior-Anterior Shift in Aging (PASA). This theory assumes both frontal over-activation and posterior midline cortex under-activation in older adults compared to younger ones (Davis et al., 2009). The Compensation-Related Utilization of Neural Circuits Hypothesis (CRUNCH) proposes that people will activate more cortical regions if task difficulty increases (Reuter-Lorenz and Cappell, 2008). Finally, the Scaffolding Theory of Aging and Cognition (STAC and STAC-r) suggests that the increased frontal activation with age is a marker of the adaptive human brain indicating a compensation for the structural and functional decline going along with aging (Park and Reuter-Lorenz, 2009). This theory takes a holistic view by considering compensation a normal process involved in our daily lives in order to be able to achieve our goals.

Gender

The influence of gender in the context of WM and, more specifically, vWM, is still rather controversial, with some studies reporting no gender effects (Bell et al., 2006; Schmidt et al., 2009) and others reporting significant differences between male and female participants (Lejbak et al., 2011; Zilles et al., 2016). The controversial results might be due to the potential influence of sex hormones, which have been shown to influence several cognitive functions including vWM (Mordecai et al., 2008; Joseph et al., 2012). Sex hormones are known to fluctuate with, for instance, menstrual cycle or hormonal contraception. However, most studies did not provide any information on these aspects which may explain the result heterogeneity to some degree.

Additional Factors (Tasks, Load, Mean, Reaction Time)

A previous meta-analysis showed differences in brain activity due to WM task type (Rottschy et al., 2012). They found that n-back and Sternberg tasks, which are typical fMRI WM tasks, not only showed differences in mental processes but also in brain activation. Moreover, tasks can vary in their difficulty through modulating the WM load (i.e., the number of items that need to be remembered). Load effects reflect the neural activation related to the increasing memory demands of information (Cowan et al., 2012; Cowan, 2017). Rottschy et al. (2012) found that load effects were mainly associated with activation in the bilateral inferior frontal gyrus. Finally, Honey et al. (2000) demonstrated that prolonged mean reaction times (RT) in response to a vWM task could influence activation in WM related brain regions. Therefore, these findings suggest that all the previously mentioned potential factors should be taken into consideration.

Aim of the Study

Against this background, the first aim of the present study was to provide an updated and extended meta-analysis of the neural correlates of vWM in healthy humans using a coordinate-based meta-analysis. The second aim was to find out more about the role of the potential moderators (age, gender, type and difficulty of the fMRI task, mean RT, and statistical threshold). Although task performance which is related to the difficulty level could be another potential factor, it was not taken into consideration due to the heterogeneous assessment in the selected studies (i.e., absolute correct values, percentage of correct values, accuracy) as pointed also by Meule (2017). To our knowledge this is the first meta-analysis to study these factors in vWM. A better knowledge about their influence on the neural correlates of vWM will increase understanding of the general mechanisms of vWM as well as help to improve methods and analyses of future vWM studies.

TOEFL® 읽기 문제 5문항. 이 장의 한 대목을 그대로 옮기고 문항을 붙였습니다. 대목을 읽고 문제를 풉니다.

문제 풀기 이 장에서 대목 보기

이 글은 위 출처의 원문을 그 라이선스(CC BY 4.0)에 따라 옮겼습니다. 참고문헌 목록과 인용 번호, 그림 · 표 · 각주 · 수식 · 부록은 빼고 절별로 나누어 실었으며, 그 밖의 문장은 원문 그대로입니다. 원 저작자와 발행 기관은 이 화면을 만들거나 검토하지 않았고, 이곳의 해석과 낱말 풀이는 공식 번역이 아닙니다.