What are the Big Ideas in Cognitive Neuroscience?

This year, the Cognitive Neuroscience Institute (CNI) and the Max-Planck-Society organized a symposium on Big Ideas in Cognitive Neuroscience. I enjoyed this fun forum organized by David Poeppel and Mike Gazzaniga. The format included three pairs of speakers on the topics of memory, language, and action/motor who “consider[ed] some major challenges and cutting-edge advances, from molecular mechanisms to decoding approaches to network computations.”

Co-host Marcus Raichle recalled his inspiration for the symposium: a similar Big Ideas session at the Society for Neuroscience meeting. But human neuroscience was absent from all SFN Big Ideas, so Dr. Raichle contacted Dr. Gazzaniga, who “made it happen” (along with Dr. Poeppel). The popular event was standing room only, and many couldn't even get into the Bayview Room (which was too small a venue). More context:

“Recent discussions in the neurosciences have been relentlessly reductionist. The guiding principle of this symposium is that there is no privileged level of analysis that can yield special explanatory insight into the mind/brain on its own, so ideas and techniques across levels will be necessary.”

The two hour symposium was a welcome addition to hundreds of posters and talks on highly specific empirical findings. Sometimes we must take a step back and look at the big picture. But since I'm The Neurocritic, I'll start out with some modest suggestions for next time.

• There was no time for questions or discussion.
• There were too many talks.
• It would be nice for all speakers to try to bridge different levels of analysis.
• This is a small point, but ironically the first two speakers (Gallistel, Ryan) did not talk about human neuroscience.

So my idea is to have four speakers on one topic (memory, let's say) with two at the level of Gallistel and Ryan¹, and two who approach human neuroscience using different techniques. Talks are strictly limited to 20 minutes. Then there is a 20 minute panel discussion where everyone tries to consider the implications of the other levels for their own work. Then (ideally) there is time for 20 minutes of questions from the audience. However, since I'm not an expert in organizing such events, allotting 20 minutes for the audience could be excessive. So the timing could be restructured to 25 min for talks, 10-15 min panel, 5-10 min audience. Or combine the round table with audience participation.

Last year, Symposium Session 7 on Human Intracranial Electrophysiology (which included the incendiary tDCS challenge by György Buzsáki) had a round table discussion as Talk 5, which I thought was very successful.

Video of the Big Ideas symposium is now available on YouTube, but in case you don't want to watch the entire two hours, I'll present a brief summary below.


Big Box Neuroscience

Here's an idiosyncratic distillation of some major points from the symposium.

• The brain is an information processing device in the sense of Shannon information theory.
• The brain does not use Shannon information.
• Memories (”engrams”) are not stored at synapses.
• We learn entirely through spike trains.
• The engram is inter-spike interval.
• The engram is an emergent property.
• Emergent properties are for losers.
• Language is genetically predetermined.
• There is something called mirror neural ensembles.
• Recursion is big.
• Architectures are important.
• Build a bridge from networks to models of behavior.
• Use generative models to construct theories.
• Machine learning will save us.
• Go back to behavioral neuroscience.

Maybe I'll explain what this all means in the next post. You can also check out the official @CogNeuroNews coverage.


ADDENDUM (April 18 2017): The sequel is finally up: The Big Ideas in Cognitive Neuroscience, Explained


Footnote

¹ Controversy is always entertaining, and these two had diametrically opposed views.