Integration of “what” and “where” in frontal cortex during visual imagery of scenes

Abstract

Imagination is a key function for many human activities, such as reminiscing, learning, or planning. Unravelling its neuro-biological basis is paramount to grasp the essence of our thoughts. Previous neuroimaging studies have identified brain regions subserving the visualisation of “what?” (e.g. faces or objects) and “where?” (e.g. spatial layout) content of mental images. However, the functional role of a common set of involved regions – the frontal regions – and their interplay with the “what” and “where” regions, has remained largely unspecified. This study combines functional MRI and electroencephalography to examine the full-brain network that underlies the visual imagery of complex scenes and to investigate the spectro-temporal properties of its nodes, especially of the frontal cortex.

Our results indicate that frontal regions integrate the “what” and “where” content of our thoughts into one visually imagined scene. We link early synchronisation of anterior theta and beta oscillations to regional activation of right and central frontal cortices, reflecting retrieval and integration of information. These frontal regions orchestrate remote occipital–temporal regions (including calcarine sulcus and parahippocampal gyrus) that encode the detailed representations of the objects, and parietal “where” regions that encode the spatial layout into forming one coherent mental picture. Specifically the mesial superior frontal gyrus appears to have a principal integrative role, as its activity during the visualisation of the scene predicts subsequent performance on the imagery task.

Introduction

Though often unconsciously, we commonly use mental representations of objects, events and even complex scenes that are not physically present in many of our everyday-life activities. This phenomenon of mental imagination refers to the set of cognitive processes that allows the generation, inspection, and manipulation of such mental representations, and it is likely the cognitive ability at the heart of several core cognitive functions such as object recognition (Riesenhuber and Poggio, 2000), spatial orientation (Goldstein, 2002), learning (Kosslyn et al., 1995), and memory (Cohen et al., 1996).

The neuro-biological mechanism underlying mental imagery is a complex system with a multitude of nodes and interaction-patterns in the brain. Numerous neuropsychological (Farah et al., 1988, Levine et al., 1985) and, more recently, neuroimaging studies (Cohen et al., 1996, D'Esposito et al., 1997, Knauff et al., 2000, Mellet et al., 1998, Mellet et al., 1996, Richter et al., 1997, Trojano et al., 2000) have aimed at unravelling the neural foundations of mental imagery using a wide variety of tasks (for a review, see Kosslyn et al., 2001). However, all of these previous imaging studies focused on only a certain subset of these imagery-related brain regions, using tasks that explored a particular aspect of mental imagery. These studies have shown that analogous to the distinction between the ventral “what” and the dorsal “where” cortical processing streams in visual perception (Ungerleider and Haxby, 1994), a dichotomy exists between the imagination of objects or scenes versus the imagination of spatial features and relationships. In line with this distinction, some studies focussed on examining the visual imagery of objects and subsequently reported imagery-related activity in category-specific occipital–temporal regions as a main finding (Ishai et al., 2002, Ishai et al., 2000, O'Craven and Kanwisher, 2000). In a separate research line, other brain imaging studies revealed the involvement of brain regions within the dorsal pathway during imagery tasks that included a spatial processing component, suggesting that the fronto-parietal networks activated during perceptual visuospatial tasks also underlie the spatial analysis of mentally imagined representations (Cohen et al., 1996, Lamm et al., 2001, Mellet et al., 1996, Tagaris et al., 1996, Trojano et al., 2000). In addition to these task-specific imagery regions within temporal and parietal cortex, both research lines also descriptively reported activation of a set of frontal regions common to all mental imagery tasks. However, while the ventral stream regions were suggested to contribute to the retrieval of content-specific object representations from memory (Ishai et al., 2000), and a functional segregation for spatial image construction and inspection was proposed between left and right posterior parietal cortex (Formisano et al., 2002, Sack et al., 2005, Sack et al., 2002), a systematic investigation of the precise functional role of these frontal regions and their relationship to the task-specific network nodes during mental imagery has been neglected. All previous neuroimaging mental imagery studies, including our own work, have restricted their analyses to the functional localization of a small number of regions, leading to a simplification of the intricate network dynamics during mental imagery.

In this study, we examine the spatio-temporal dynamics of activity in frontal regions and their relation to parietal, occipital–temporal, and primary sensory brain regions, during a behaviourally-controlled task that captures both object and spatial mental imagery. Participants were asked to imagine a previously learned complex visual scene from an auditory cue and, after a certain delay, had to judge whether a visually presented fragment of the scene was mirrored or not. We used event-related functional MRI (fMRI) and latency analysis techniques on the imagery period to capture all imagery-relevant network nodes and their relative temporal onset of activation. In addition, we performed the same scene imagery task during electroencephalography (EEG) measurements to validate our fMRI latency results and to derive more information on the specific functional role of the brain regions from the involved frequency bands. For example, the involvement of the alpha rhythm is often associated with global cognitive engagement (Barrett and Ehrlichman, 1982, Bhattacharya and Petsche, 2005, Gill et al., 1998, Marks and Isaac, 1995, Nikolaev and Anokhin, 1998, Petsche et al., 1997, Petsche et al., 1992), whereas the involvement of the theta or beta rhythm could point to a more specific function such as, respectively, working memory (Bhattacharya and Petsche, 2005, Petsche et al., 1997, Rugg and Venables, 1980) or the integration of information into a rehearsal set (Onton et al., 2005). Our experimental design and methodological approach enabled us to link the spatially-resolved information on the distributed brain network underlying the mental imagery of complex scenes as revealed by fMRI, with the spectral–temporal electrophysiological dynamics within this same network using EEG. We aimed at deriving an exhaustive functional neuro-anatomical model of mental imagery, in which the specific integrative role of the frontal regions and their interplay with the parietal, temporal, and sensory brain regions during complex visual scene imagery is clarified.