Hans Liljenström (ed.)Advances in Cognitive Neurodynamics (IV)Advances in Cognitive Neurodynamics10.1007/978-94-017-9548-7_15

Set-Related Neurocognitive Networks

Steven L. Bressler¹

(1)

Department of Psychology, Center for Complex Systems and Brain Sciences, Cognitive Neurodynamics Laboratory, Florida Atlantic University, Boca Raton, FL, USA

Steven L. Bressler

Email: bressler@fau.edu

Abstract

The term “set” refers to the anticipatory neurocognitive processes that prepare an individual to engage in a particular behavior. Set entails a specific configuration of anticipatory perceptual, motor, or cognitive brain processes that is initiated by task context and actively maintained for subsequent task performance. Set is made possible by prior perceptual, motor, or cognitive experience with the same or a similar task. Set-related processes are thought to establish conditions that guide and channel the fast communications between areas of the cerebral cortex that underlie perception, action, or cognition. It is proposed that set-related processes change the functional connectivity of cortical areas within and between large-scale networks of cognition, or neurocognitive networks. The operation of set has been implicated from the analysis of neuroimaging data recorded in a study of human subjects performing a demanding cued visuospatial attention task. From analysis of long-range directed (top-down) functional connectivity, a neurocognitive network of frontal and parietal cortical areas, called the Dorsal Attention Network, was inferred to modulate activity in retinotopic areas of Visual Occipital Cortex (VOC): top-down influences were larger to VOC subregions representing the attended visual hemifield than to subregions representing the unattended hemifield. This difference was maintained over seconds to minutes throughout the entire task. Bottom-up influences from the two subregions did not differ. The maintenance of task-specific top-down modulation of VOC throughout a recording session suggests that it reflects visuospatial attentional set.

Keywords

PerceptionActionCognitionCerebral cortexAnticipationNeural networkFunctional connectivityTop-downVisuospatial attentionGranger causality

1 Introduction

Set is an essential component of normal human behavior. As people anticipate events in the environment and perform tasks in daily life, they prepare appropriate perceptual, motor, and cognitive processes. Perceptual set is the predisposition to perceive a specific sensory stimulus or stimuli; motor (preparatory) set is the intention to perform a specific action or actions; and cognitive set is the tendency to execute a specific cognitive function or functions. In general, set may be viewed as any configuration of perceptual, motor, and cognitive processes that is initiated by behavioral context and actively maintained for subsequent behavior [19].

Set-related effects are known in a variety of cognitive functions, including motor speech [1], saccadic eye movements [7], visual search [8], rule-based behavioral selection [13, 18], visuospatial attention [14], and visual discrimination [20]. Most studies on the neural basis of set have centered on prefrontal cortex [2, 5, 13, 19], although posterior parietal cortex [10, 15, 22] and basal ganglia [16] have also been implicated.

NeuroCognitive Networks (NCNs) are large-scale systems of distributed and interconnected neuronal populations in the central nervous system organized to perform cognitive functions [3]. NCNs involving prefrontal and posterior parietal cortical areas figure prominently in important aspects of cognition [17]. This report considers the functional configuration of NCNs as a potential mechanism for the instantiation of set in the brain.

Changing the functional interdependency relations among their component brain areas according to task-related processing demands is a potentially powerful mechanism for the set-related configuration of NCNs. It is known that a high-level frontoparietal NCN, called the Dorsal Attention Network (DAN), is responsible for controlling the selection of task-specific sensory information in humans and non-human primates [6]. The frontal and parietal regions of the DAN are consistently activated by central cues indicating where a peripheral object will subsequently appear. Set-related NCN configuration is postulated to occur in visuospatial attention as the top-down functional modulation of “low-level” visual cortical areas by “high-level” cortical areas of the DAN.

2 Methods

Is it possible to quantitatively assess the top-down functional modulation of low-level sensory areas of the brain by high-level control areas during attentional set? Wiener [25] proposed that statistical prediction of the activity in one brain region from that in another region might come from the study of “coefficients of causality running both ways” between regions. This idea would be useful for determining top-down functional modulation related to attentional set if such statistical prediction measurements could be practically derived from neural time series data recorded during set-related behavior.

Can measurements of the type proposed by Wiener actually be made? Following the lead of Wiener, Granger [11] proposed a method for measuring “coefficients of causality” from empirical data by what has come to be known as Granger Causality (GC) [12]. Granger, considering two arbitrary time series, y_t and x_t, proposed to compare two autoregressive (ar) models, called the restricted and unrestricted models, of y_t. The restricted model of y_t is an ar model that includes only past terms of y_t. The unrestricted model of y_t includes those same past terms of y_t, but also includes past terms of x_t. The improvement in predictability of y_t by inclusion of x_t in the unrestricted model, as compared to that of the restricted model, is taken as a measure of statistical “causality”. In other words, if inclusion of x_t in the unrestricted model significantly improves the predictability of y_t, as compared to the restricted model, then x_t has a Granger casual influence on y_t. Determination of this improvement can be made by comparing the variances of the residual time series of each model: if the unrestricted model residual variance is significantly lower than that of the restricted model, then the unrestricted model is better at predicting y_t than the restricted model, and it is said that x_t Granger causes y_t. Not only can y_t be modeled using x_t, but x_t can also be modeled using y_t. Therefore, for any two time series, x_t and y_t, their interdependency relations may be either symmetric, i.e. GC_x→y ≈ GC_y→x, or asymmetric, i.e. GC_x→y ≠ GC_y→x. Multivariate autoregressive modeling may also be used in place of the bivariate models [24]. Autoregressive modeling is increasingly being used to measure directed functional connectivity in the analysis of brain networks.

To follow Wiener’s proposal for measuring the statistical prediction of activity in one brain region from that in another may thus be accomplished by applying autoregressive modeling to neural time series data. To measure directed functional connectivity related to top-down attentional modulation in the brain, then, minimally requires that neural time series data be recorded during attention-demanding task performance, and that directed functional connectivity be measured from those time series. Recording modalities that currently provide potentially suitable time series data include the electroencephalogram (EEG), magnetoencephalogram (MEG), and functional Magnetic Resonance Imaging (fMRI) in normal humans; electrocorticogram (ECoG) in human patients with intracranial electrodes; and local field potential (LFP), spiking single-unit activity (SUA), and multi-unit activity (MUA) in experimental animals.

Only gold members can continue reading. Log In or Register to continue