A magnetoencephalography spatiotemporal analysis of neural activities during feature binding Francesca M. Filbey a , Tom Holroyd b , Frederick Carver b , Trey Sunderland a and Robert M. Cohen a a Geriatric Psychiatry Branch, National Institute of Mental Health and b NIMH MEG Core Facility, Bethesda, Maryland, USA. Correspondence and requests for reprints to Francesca M. Filbey, Department of Psychology, University of Colorado at Boulder, Muenzinger Building, D326 -B, 345 UCB, Boulder, CO 80309- 0345, USA Tel: + 1302 492 9203; fax: + 1303 492 2967; e-mail: Francesca.Filbey@colorado.edu Received10 August 2005; revised 29 August 2005; accepted 30 August 2005 Previous literature shows that feature binding processes elicit fronto-hippocampal areas. The time course of this process, how- ever, remains unknown.This is the ¢rst study that investigates fea- ture binding using magnetoencephalography. Synthetic aperture magnetometry analysis was used to localize sources of increased power in the y band during the encoding phases of a feature- binding task in the left and right medial frontal gyri (Brodmann’s area 10) and left and right anterior cingulate gyri. Theta band syn- chronization was observed in many of these same areas, but also in other areas not noted to have increased y band power suggesting a broad network of regions subserving the encoding phase of feature binding. NeuroReport 16:1747^1752  c 2005 Lippincott Williams & Wilkins. Keywords: cingulate gyrus, feature binding, magnetoencephalography, prefrontal cortex, y band, time course analysis, working memory Introduction Working memory has been widely studied in the functional magnetic resonance imaging (fMRI) literature. Of the key brain regions associated with working memory functions, the prefrontal cortex is thought to underlie organization, maintenance and manipulation of information [1], the cingulate cortex to reflect the assessment of information [2] and the medial temporal cortex, particularly the hippocampus, for memory retrieval [3–5]. The hippocam- pus, particularly the anterior portion of the hippocampus, in addition to the prefrontal cortex, has also been reported to underlie a specific process in working memory called feature binding [6]. Feature binding has been characterized as the early online process through which different aspects of a stimulus is integrated into a single memory. It has also been described in terms of relational memory binding, which is an important factor for long-term memory representations of relationships among different elements. The hippocampus has been implicated in both early and late binding processes. For example, studies of amnesic patients with hippocampal damage have often reported impaired memory for relationships between features [7,8]. Addition- ally, the hippocampus has also been postulated to play a crucial role in the early perceptual processing and repre- sentation of integrated objects [9,6]. At present, we are only aware of two studies that have attempted to characterize the temporal characteristics of feature-binding processes. One study used cognitive meth- ods to investigate the dynamic binding of feature and location elements [10]. The author posited that dynamic binding has a short lifespan, which is conducive to multiple serial-binding processes. The second study investigating the time course of feature binding utilized simultaneous hemodynamic recordings using fMRI and electromagnetic recordings using event-related potentials [11]. The authors were interested in the feature binding of irrelevant stimuli in a moving array. They discovered that despite instructions to ignore certain stimuli, these ignored stimuli were still processed very rapidly (i.e. within B40–60 ms). These existing studies of the time course of online feature binding demonstrate that binding processes of moving stimuli occur rapidly and effortlessly. Unlike the aforementioned studies, which utilized dy- namic stimuli to measure object-based attention during binding, we wanted to investigate the working memory aspects of feature binding, rather than its perceptual- attention mechanisms. We, therefore, investigated feature binding of stationary stimuli. Theta (y) band activity has often been associated with memory functions [12–16]. It has been suggested that the interactions, particularly during cell assembly formation, between the hippocampus and the neocortex underlie this process [17]. Memory formation and retrieval are then believed to occur during cell assembly formation [17]. Previous work from our laboratory has also indicated a specific role of y rhythms during memory retrieval (Filbey, abstract citation). We found that differences in y band activities in the frontocingular network during feature binding reflected differences in genotype (i.e. apolipoprotein allele e4). These differences paralleled our findings in fMRI. Interestingly, however, these differences were not found in high-frequency bands, such as g (30–80 Hz). In the present study, we used magnetoencephalography (MEG) because it provides fine spatial and temporal information about neural activity to investigate y activity in those brain regions previously reported to be involved in working memory. BRAIN IMAGING NEUROREPORT 0959-4965  c Lippincott Williams & Wilkins Vol 16 No 16 7 November 2005 1747 Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.