Cross-cortical Coherence during Effector Decision Making

1. Cross-cortical Coherence during Effector Decision Making Chess Stetson Andersen Laboratory Caltech Sloan-Swartz Meeting 2009/07/28

2. How do different parts of the brain work together? In other words, how do they communicate and form functional networks?

3. How do different parts of the brain work together? Oscillatory synchrony is a popular answer – different neural populations oscillate at the same phase of a brain rhythm in order to talk. from Womelsdorf et al., Science, 2007

4. How do we look for synchrony? We record single-unit spikes and local field potentials (LFPs), which correlate with subthreshold membrane potential. from Poulet & Petersen, Nature, 2008

5. Where do we look for synchrony? We are particularly interested in communication between separate cortical areas. Under the theory that LFPs represent input to a region (Mitzdorf, 1982), we might expect the output of one area to synchronize with the input to another.

6. What measure do we use? • FT(x) = the multi-taper Fourier transform of x • C(x,y) is a complex number representing the phase of the offset between signals x and y at every frequency • We will focus on |<C(x,y)>|, the epoch-averaged coherence, expressing the consistency in phase-lag between signals, independent of signal power • z-transform generates a z-score, according to the # of degrees of freedom, s.t. coherence->0 under the null hypothesis

7. The common driver problem Apparent communication between two signals could actually result from a 3rd area nefariously driving them both. We would like a measure that isolates the cross-cortical component.

8. The partial coherence... ...isolates the coherence between two signals, apart from the influence of a third Any third party who drives both cortical areas will likely be available in both LFPs. To the extent that we can identify this 3rd party, we can sift it out.

9. Which brain regions to choose? • MIP(medial intraparietal cortex) and PMd(dorsal premotor cortex) are monosynaptically connected parts of the brain • MIP is thought of as the final stage of the dorsal pathway of visual processing • PMd is thought to be involved in the planning of motor behavior • Cortically distant, functionally similar PMd MIP

10. (600 ms) (600-1000ms) A typical task for MIP & PMd No target or effector cue during the delay period Delay Reach

11. 60 80 0 0 Mean firing rate (Hz) 0 .6 Time Into Delay Period (s) Delay Reach

12. 60 80 0 0 Mean firing rate (Hz) 0 .6 Time Into Delay Period (s)

13. Why decision-making? • Visual stimuli exert powerful effects on neural synchrony • Decisions are internally generated -- removing visual cues rules out a likely source of common drive

14. Why effector decision-making? During effector decision-making, when reward contingencies are near 50/50 (Barraclough et al., Nat Neuro, 2004), information evolves slowly within MIP/PMd. Coherence is often associated with visual attention (Gregoriou, et al., Science, 2009), but cortical communication could be more general. During effector decision-making, are the cells with something to say the ones most likely to communicate?

15. How do different parts of the brain work together? Between two cortical locations involved in making an effector decision... ...do cells which show early evidence of the decision have more cross-cortical influence?

16. Do effector-decision cells show more cross-cortical coherence?

17. Do effector-decision cells show more cross-cortical coherence? REACH/SACCADE 30 8 50 Mean firing rate (Hz) z-trans., part. coherence Frequency (Hz) 0 -2 0 -.6 0 -.6 -.6 0 -.6 Time Into Delay Period (s)

18. z-trans., part. coherence REACH/SACCADE 50 100 0 0 30 50 8 0 -2 0 50 60 0 0 60 50 Mean firing rate (Hz) Frequency (Hz) 0 0 -.6 0 -.6 -.6 0 -.6 Time Into Delay Period (s)

19. Effector decision cells are more cross cortically coherent Spike-LFP pairs including “effector decision” spikes * * * 1 Ensemble “effector decode” Spike-LFP pairs without significant tuning during the delay 0.75 % correct avg. z-trans part. coherence 0.5 MEMORY PERIOD 0 0 700 time (ms) into memory period Cells with significant effector tuningduring the memory period 0 50 100 Frequency (Hz) Cells with non-significant tuning * , p<.05

20. Not all coherent cells make effector decisions 30 14 50 z-trans., part. coherence Frequency (Hz) Mean firing rate (Hz) -2 0 0 -.6 0 -.6 -.6 0 -.6 Time Into Delay Period (ms)

21. 1 * * * 0 0 50 100 Effector decision cells are more cross-cortically coherent during the memory delay z-trans part. coherence * Frequency (Hz) , p<.05

22. More fronto-parietal coherence in the WHERE task, as compared to increased parieto-frontal coherence in the HOW task. 1.2 “WHERE” TO MOVE “HOW” TO MOVE 0.6 z-trans part. coherence 0 -0.6 Time Into Delay Period (ms) from Pesaran, Nelson and Andersen, Nature, 2009 pre-filtered for significantly coherent cells no pre-filtering for significant coherence 0 600 1200 1800

23. Conclusions • Significant, sustained coherence appears in the delay period in an effector choice task. • Cells that make effector decisions are more likely to be cross-cortically coherent, supporting the idea that coherence represents influence across cortical regions. • The effector decision makes it unlikely that this cross-cortical coherence has exclusively to do with spatial attention. • There is an overall increase in parieto-frontal coherence during effector decision-making, potentially suggesting that the type of decision affects the direction of communication.

24. Thanks • The Swartz foundation • The Andersen laboratory Richard Andersen, Tessa Yao, Viktor Shcherbatyuk, Xoana Troncoso, Linus Schumacher, and many other talented folks • Bijan Pesaran, for many helpful conversations • Monkeys L & Z