In December I gave a talk at the new and beautiful Howard Hughes Janelia Farm Research Campus in Virginia. I talked about my work on how we resolve ambiguous or multiple stimuli. My slides are here although the movies don’t work. The talk is based mostly on work in two papers (C.R. Laing and C.C. Chow. `A spiking neuron model for binocular rivalry’, J. Comp. Neurosci. 12, 39-53 (2002) and S. Moldakarimov, J.E. Rollenhagen, C.R. Olson, and C.C. Chow, ‘ Competitive dynamics in cortical responses to visual stimuli’, Journal of Neurophysiology 94, 3388-3396 (2005), both of which can be downloaded from here) with some new stuff that Hedi Soula and I have been working on (and mostly off) for the past four years. I’m hoping that we’ll finally finish the paper this year.
When the eyes are presented with multiple or ambiguous stimulus, several things can happen. You can perceive multiple images, which is what you usually do when you look out into the natural world. You can resolve an ambiguity. For example, you could be presented with a dark and shadowy image and your brain decides on what is figure and what is ground. However, if you are presented with something truly ambiguous like the Necker cube then your perception will be multistable. You’ll see one thing and then another and back again. The most striking form of multistable perception is binocular rivalry, which occurs when each eye is presented with a completely unrelated image like horizontal stripes to the left eye and vertical stripes to the right eye. For a range of contrasts, your perception will alternate between vertical and horizontal stripes. The alternations are stochastic with a gamma-like distribution for the dominance times with a mean of a second or so, which varies from person to person.
The interesting thing about these different psychological responses is that they are correlated to neural responses. As I wrote before about the neural code, there are neurons in the brain that respond to different stimuli. Hence, there are neurons that fire when you see an apple and others that fire when you see an orange. If you see an apple and an orange simultaneously, both sets of neurons will fire. However, as shown in experiments, the firing rates of the neurons “normalize” when multiple images are observed. So if neuron A fires at rate R when you look at an apple and rate S when you look at an orange, it will fire at a rate between S and R when you look at both. The rates sum sub-linearly or normalize. This is thought to be a form of gain control.
Now the thing that really got my attention was the result shown by David Leopold and Nikos Logothetis and that is a) monkeys experience binocular rivalry and b) there are neurons in the brain that are correlated with what the money perceives. So neurons that are responsive to vertical stripes will only fire when the monkey reports that it is seeing vertical stripes. To me this was a revelation because it points to neurons that are correlated with our conscious perception. Even though the eyes are presented with two images and neurons in the primary visual cortex are active for both, there are neurons higher up in the visual stream that are correlated with what you perceive. It turns out that there are neurons in primary visual cortex that also alternate but the fraction that alternates increases as you go further up the visual pathway. This has also been seen in humans with fMRI imaging. To me this says that there is some set of neurons (which need not be fixed) in the brain that makes up the conscious part of you.
My work with Carlo Laing and later with Samat Moldakarimov (with data from Carl Olson and Julie Rollenhagen) was to show that all of these neural dynamics could be explained with a simple canonical cortical circuit with recurrent excitation and opponent inhibition in the presence of synaptic depression. In particular, we could explain a peculiar phenomenon called Levelt’s second proposition that states that during binocular rivalry if you reduce the contrast to one eye you increase the amount of time you see the image of the other eye but do not affect the dominance time of what you see in the manipulated eye. ( Rubin Moreno-Bote, Nava Rubin, John Rinzel and others have since shown that it is a little more complicated than this.) Another thing we could do was to derive the dominance time distribution based on an exponential decay to a stochastic threshold. However, we never had a full microscopic theory for the stochastic distribution in terms of a population of neurons. That is what Hedi and I have been working on for the past few years. Along the way we derived a simple model of stochastic neural activity that can account for finite population size effects (H. Soula and C.C. Chow, `Stochastic dynamics of a finite-size spiking neural network’, Neural Comp., 19, 3262-92 (2007), also available on my web page). One puzzle is that data shows that a lot of neurons participate in the alternations. Naively, since correlations between neurons are generally weak, this would imply that any noise present in the neurons would be averaged away and mean field theory would apply leading to periodic oscillations. One of the things we are arguing is that the noise responsible for the dominance time distribution for rivalry arises because a small set of neurons initiate alternations. I’ll write more on this when we finish the paper.