Causal Interactions in Attention Networks Predict Behavioral Performance

Abstract

Lesion and functional brain imaging studies have suggested that there are two anatomically nonoverlapping attention networks. The dorsal frontoparietal network controls goal-oriented top-down deployment of attention; the ventral frontoparietal network mediates stimulus-driven bottom-up attentional reorienting. The interaction between the two networks and its functional significance has been considered in the past but no direct test has been carried out. We addressed this problem by recording fMRI data from human subjects performing a trial-by-trial cued visual spatial attention task in which the subject had to respond to target stimuli in the attended hemifield and ignore all stimuli in the unattended hemifield. Correlating Granger causal influences between regions of interest with behavioral performance, we report two main results. First, stronger Granger causal influences from the dorsal attention network (DAN) to the ventral attention network (VAN), i.e., DAN3 VAN, are generally associated with enhanced performance, with right intraparietal sulcus (IPS), left IPS, and right frontal eye field being the main sources of behavior-enhancing influences. Second, stronger Granger causal influences from VAN to DAN, i.e., VAN3 DAN, are generally associated with degraded performance, with right temporal-parietal junction being the main sources of behavior-degrading influences. These results support the hypothesis that signals from DAN to VAN suppress and filter out unimportant distracter information, whereas signals from VAN to DAN break the attentional set maintained by the dorsal attention network to enable attentional reorienting.

The MVGC multivariate Granger causality toolbox: A new approach to Granger-causal inference

“…While technological developments supporting ultra-low TRs (Feinberg and Yacoub, 2012) and de-noising (Rasmussen et al, 2012) promise to alleviate these problems, in the absence of these developments a conservative methodology is recommended. Useful strategies include (i) using as short a TR as possible by compromising on coverage; (ii) examining changes in Gcausality between experimental conditions, rather than attempting to identify "ground truth" G-causality patterns; (iii) correlating changes in G-causality magnitude with behavioural variables such as reaction times across trials (or trial blocks) (Wen et al, 2012), and (iv) computing the so-called "difference of influence" term (Roebroeck et al, 2005) which may provide some robustness to HRF variation. Alternative promising approaches include estimation of state-space models which jointly parameterize functional connectivity and hemodynamic responses (Ryali et al, 2011) or blind deconvolution of the HRF to retrieve the underlying neuronal processes (Havlicek et al, 2011;Wu et al, 2013).…”

Empathic control through coordinated interaction of amygdala, theory of mind and extended pain matrix brain regions

“…GCM identifies asymmetric predictive relations between time series. Although it 12 remains unclear whether Granger Causality indicates actual causality (Granger, 1969), 13 differences in Granger Causality across participants predict individual differences in 14 behavior (reaction time) (Wen et al, 2012), illustrating that GCM can provide 15 behaviorally relevant information. In the initial analysis, GCM identifies whether the 16 response in one brain region predicts the subsequent response in another brain region.…”

Brain Connectivity and Visual Attention

“…Stronger Granger causal influences from the DAN to the VAN were positively correlated with improved task performance (red), while stronger Granger causal influences from the VAN to the DAN were negatively correlated with task performance (blue). Reproduced with permission from Wen et al (2012). set maintained by the dorsal attention network to enable attentional reorienting.…”