1. Neurocognitive Networks and Task Set
Steven L. Bressler
Cognitive Neurodynamics Laboratory
Center for Complex Systems & Brain Sciences
Department of Psychology
Florida Atlantic University

3. Overview NeuroCognitive Networks (NCNs) and Task Set
NCN Measurement by Directed Functional Connectivity (DFC)
NCNs in Visual Spatial Attention
NCNs in Visual Pattern Discrimination

4. Aleksandr Luria
“The concept of localization of functions … has come to mean a network of complex dynamic structures or combination centers, consisting of mosaics of distant points of the nervous system, united in a common task.”
Higher Cortical Functions in Man, 1962

5. NeuroCognitive Networks
Bressler, Scholarpedia, 2008 Bressler & Menon, TICS, 2010 Fuster & Bressler, TICS, 2012 Meehan & Bressler, NBR, 2012
NeuroCognitive Networks (NCNs) are large-scale systems of distributed and interconnected neuronal populations in the brain organized to perform cognitive functions.

6. NCN Functional Organization
NCNs are functionally organized by the processing demands of cognition.
NCN functional organization is constructed by the mutual constraints imposed on constituent neuronal populations as they interact.
NCN functional organization is observed by changes in network node activity and edge strength (functional connectivity).
Functional connectivity may be undirected (e.g. coherence measures) or directed (e.g. Granger causality measures).

7. Directed Functional Connectivity by Wiener-Granger Causality
For two simultaneous time series, one series is called causal to the other if we can better predict the second series by incorporating knowledge of the first.
Norbert Wiener (The Theory of Prediction, 1956)
Granger (1969) formulated this idea in terms of autoregressive models.

8. Task Set in the Brain
Task set is the collection of perceptual, cognitive, and motor predispositions in the brain that facilitates performance of a task.
Example: activity of an orchestrated set of neurons throughout the motor system reflects preparation for movement (Evarts et al., 1984).

9. NCNs and Task Set
I propose that task set reflects the functional organization of NCNs. I will illustrate with 2 examples from my work:
NCNs are shaped by tonic task parameters in a human fMRI BOLD experiment on visual spatial attention.
NCNs are pre-configured before task performance in a monkey LFP experiment on visual pattern discrimination.

10. NeuroCognitive Networks in Attention Corbetta et al., Neuron, 2008
The Dorsal Attention Network (DAN) is a system of frontal and parietal regions consistently activated by cues indicating where a visual object will appear.
The DAN is postulated to exert attentional top-down control of visual cortical.

11. Posner Spatial Cueing Task Sylvester, Shulman, Jack, Corbetta J Neurosci, 2007
Six subjects performed the Posner Spatial Cueing Task at Washington University School of Medicine, St Louis.
Attention was only required to upper visual field.
Event-related fMRI BOLD data were acquired with a Siemens Allegra 3T scanner using an asymmetric spin-echo echoplanar sequence (TR=2.064s, TE=25ms, flip angle=90°, 32 contiguous 4 mm axial slices, 4x4 mm in-plane resolution).

12. Directional Asymmetry in
Visual Spatial Attention Bressler, Tang, Sylvester, Shulman, Corbetta
J Neurosci, 2008
Granger Causality was consistently greater for DAN  VOC than VOC  DAN across subjects and pairs, but not in randomized controls.
By subject
By ROI pair
Across-Cortex Voxel-Randomized Control
Trial- Randomized Control

13. NeuroCognitive Networks in Visual Spatial Attention Meehan et al
NeuroCognitive Networks in Visual Spatial Attention Meehan et al., in prep
NCN diagram showing interactions within & between DAN & VOC in visual spatial attention.
The diagram was constructed by a modified FARM procedure that measures conditional DFC between brain regions.

14. Top-Down Modulation in Visual Spatial Attention is Spatially Selective
Tang et al., in prep
Top-down influences from DAN to VOC are consistently greater than bottom-up influences throughout the trial.
Top-down influences to ventral VOC (representing upper VF – where attention was directed and targets appeared) are consistently greater than those to dorsal VOC (representing lower VF – where attention was not directed and targets did not appear) throughout the trial.
target
Top-Down blue line
Bottom-Up red line
Ventral VOC solid line
Dorsal VOC dashed line

15. Simultaneous Recordings
Visual Pattern Discrimination Task (Experiments performed by Drs. Richard Nakamura & Richard Coppola at the National Institute of Mental Health)
Lateral View
Electrodes
Simultaneous Recordings
Stimuli
-115 ms
0 ms
300 ms
Response
Decision
Sensation
Expectation
Time

16. Neurocognitive Networks in Visual Expectation Bressler et al, Stat Med, 2007
Prestimulus top-down influences are carried by synchronized beta rhythms in a neurocognitive network of V1 and extrastriate (V4, TEO) sites.

17. Top-Down Modulation in Visual Pattern Discrimination is Task-Specific Richter et al., in prep
The data were bisected into training and test sets, which were then used to bootstrap trial bivariate AR models for each striate-extrastriate pair for each contingency. 
A SVM, constructed from the training set, was used to classify the test set.  This procedure was repeated for random bisections of the data.
Bootstrap distributions of Support Vector Machine classification of prestimulus top-down beta-band GC between 2 stimulus-response task contingencies. Veridical bootstrap distribution (red) vs randomized control distribution (blue).   (*** p < 10e-5 based on a non-parametric randomization test; contingencies: line-go/diamond-nogo vs line-nogo/diamond-go).

18. Effect of Top-Down Modulation on Visual Evoked Response
The size of the N1 evoked response at a V1 site to a stimulus (e.g. right slanted line) depends on task contingency.
Example for 1 stimulus type at 1 V1 site
Mean rectified difference between contingencies over all stimulus types and all V1 sites in 2 monkeys

19. Take-Home Message
Task set is created in the brain by the functional organization of NeuroCognitive Networks in cognitive task performance.
Set is an essential aspect of cognitive function, and should make artificial neural network models more versatile and powerful.

20. Implications for Cognitive Science
A cognitive system is an organized body of information that associates knowledge in a coherent framework (Rescher, 1979).
Cognitive systems may bear isomorphic relations to neurocognitive network structures in the brain (Bressler, 1999).
Organized associational knowledge is observed in the “cognit” structure of large-scale cortical networks (Fuster & Bressler, 2012).

21. Reprints Available at http://www.ccs.fau.edu/~bressler/