Summary:
We finished the discussion on the “master” review by Wang and discussed computational roles of oscillations (see the 2nd ref.).
We started with a fundamental puzzle of oscillations in the cortex – how do we record oscillations LFP, when single neurons are quite far from that (usually highly Poisson)? Wang discusses a basic model with Poisson neurons that shows that if you add a small oscillatory current (that is probably produced by the previous mechanisms we spoke about) they can be maintained by this type of networks. We worked to understand how this model can determine the frequency of the oscillation – it turns out that this crucially depends on the filtering properties of single neurons, which emphasizes the importance of knowing your neurons. We also discussed the general utility of models when they necessarily ignore at least part of the underlying biological variation – the models assume homogeneous filtering properties, but thats not how biology works (at all). Lastly, this model enables us to predict which types of neural networks will create what range of frequencies – I-I networks mostly stand generate ultra-fast oscillations while E-I networks are more in the gamma range. Summing up à coupled oscillators are much less likely in the cortex (neurons aren’t oscillatory), but you can have Poisson neurons and still get oscillations (and these will be less stable etc., like what we actually find).
We then moved to Wang’s discussion of functional\cognitive properties, which started with a big disclaimer about the methodologies currently used in the field. We strongly agree with his line of criticism, centering around the fallacy of assuming an oscillation without showing any peaks in the spectrum \ coherence plot. We agreed to continue this discussion in a future meeting centering on the work of Brad Voytek. Additionally, Wang mentions synchrony is quite small between the LFP and single neurons and between neurons, so it raises an important question about the importance of these phenomena. Wang believes that at least in some cases this small synchrony can still be very important, and we tend to agree, but it’s important to keep this in mind.
We discussed phase coding in detail, and specifically the phenomenon of phase precession of place cells in the hippocampus. This is a paradigmatic case where the LFP carries additional information above spike rates, though importantly, we don’t know what the “readout” here, so the question is still relatively open (and also – this doesn’t happen in bats, oddly). This brought us to the question of whether it is at all crucial to look for active roles of oscillations, and we were reminded of important cognitive scientists who don’t look for the causal effects of oscillations, but simply use them as a useful marker to study cognitive processes. Lastly, discussing computational roles of oscillations we emphasized the importance of spike timing to all roles (conveying information, regulating information flow or learning), and we understood that this is an ill posed question because you always have to understand what your reference frame is (spike timing relative to what?).
Future meetings will discuss - a) data! Open meeting to discuss our data. b) Brad Voytek’s work and specifically methods for separating oscillatory peaks from 1/f “noise”.
References:
- Wang, X. J. (2010). Neurophysiological and computational principles of cortical rhythms in cognition. Physiological reviews. https://doi.org/10.1152/physrev.00035.2008
- Sejnowski & Paulsen (2006). Network oscillations: emerging computational principles. JoN. https://doi.org/10.1523/JNEUROSCI.3737-05d.2006
Participants:
- Gal Vishne
- Nir Ofir
- Nora Vrieler
- Moran Aharoni
- Maya Inbar
- Lena Gorina