The official press release from Berkeley for this paper won't be out for a while, as it's being wrapped into a larger story with my also-accepted paper to Neuron, Dynamic neuroplasticity after human prefrontal cortex damage. Both the PNAS and Neuron papers are experiments where I worked with people who had stokes. This PNAS paper shows how very specific strokes to different parts of the frontal lobes affect working memory. The Neuron paper looks into how different brain regions pick up the slack for the damaged parts of the brain.
In the meantime, since the PNAS paper is online already, I've decided to publish my thoughts on it.
I believe a lot of my blog readers are here via MindHacks or my twitter because of the strange neuroscience historical pieces I've written about self-experimentation, sperm injections, psychic powers, and scientific pharmacological aides. What a lot of you don't know is that I'm trying to pretend to be a for reals scientist, too!
When I first started this blog, I was finishing up my PhD thesis at Berkeley. At the time of my graduation I had three papers out in peer-review, a process that totally destroys all sense of self-worth in a young researcher.
Anyway, the good news is, that this paper was almost a pleasure to have go through the peer-review process. The journal that published my work, PNAS (The Proceedings of the National Academy of Sciences, USA), used to be infamous for its policy wherein researchers who were members of the National Academy of Sciences could "fast-track" their papers into the journal, even by-passing peer-review altogether! Well, since their announcement that this will no longer be done I've become a much bigger fan! Needless to say, I'm quite proud and honored to have my paper accepted there. And now that the media embargo is lifted I can finally talk about it here.
This is my first major research publication that was directly related to the topic of my PhD thesis, "Frontal and Basal Ganglia Contributions to Memory and Attention".
As I said in one of my first posts, "if the public is paying for my research, at the very least they should be able to know what was done and why I thought it was worth doing." To that end, I've tried to do a lot of communicating my research such a giving public lectures (e.g., TEDxBerkeley, Google). I'm also making it a point to write about my peer-reviewed publications. My first scientific write-up was on my hemicraniectomy paper, "Hemicraniectomy: A new model for human electrophysiology with high spatio-temporal resolution" in the Journal of Cognitive Neuroscience (full, free paper here).
Technically this paper was pretty challenging; it's not as mathematically challenging as some of my other work (e.g., "Shifts in gamma phase-amplitude coupling frequency from theta to alpha over posterior cortex during visual tasks"; post coming soon!) It was, however, methodologically more complex. As my PhD advisor and coauthor, Robert Knight, and I said in our letter to the editor when submitting the paper, "[o]ur manuscript provides the first evidence combining human electrophysiology, psychophysics, and focal brain lesions to clarify the role of cortical and subcortical frontal regions in working memory."
To unpack that a bit, we worked with two groups of patients who had strokes. We screened a lot of brain scans prior to the research to look for people who had relatively focal brain lesions in very specific areas of the brain. Because the extent of damage caused by any given stroke is highly variable, it was difficult to identify a group of subjects. Specifically, I was interested in people who had damage only in one hemisphere of the brain, to either the prefrontal cortex (PFC) or to the basal ganglia (BG). The reason for this is that, while decades of previous research has shown that the PFC is important for working memory, only for the last decade or so have people really been looking at the BG as important in cognitive functions.
Or, as we said in science-speak, "[a]lthough the basal ganglia and prefrontal cortex are known to be associated with working memory processes, the precise anatomical and functional roles they play have not been causally demonstrated in humans." I'm very interested in how different brain networks work together in complex cognition, so working with these patients gives a rare opportunity in cognitive neuroscience to look at more causal questions; that is, what brain areas are needed for cognition, not just what brain areas are correlated with it.
The figure at the top of this post shows the average of the two patient groups where the color represents the number of patients with a lesion in that exact brain area.
We chose to study working memory for a variety of reasons. First, philosophically, as I say in the paper, "[v]isual working memory (VWM) is a remarkable skill dependent on the brain’s ability to construct and hold an internal representation of the world for later comparison to an external stimulus." That right there strikes me as an absolutely amazing skill that we have. Second, it's a natural extension of some of the work out of my advisor's lab that showed that, in patients with unilateral PFC lesions, their attention abilities are intact when you present a visual stimulus to their "good" half of their brain, but they're a little worse when you present it to the "bad" half of the brain. Vision, by the way, is lateralized just like movement. Most people know that the right half of your brain controls the left half of your body (and vice versa). Well, the left half of your visual world (not the left eye, but the left half of space) is processed by the right half of your brain. So anyway, these patients with PFC lesions have worse attention when the thing they need to pay attention to is presented to the damaged half of the brain.
What's neat about that Francisco Barceló Nature Neuroscience paper was that they used scalp EEG to show that PFC lesions actually affect the way the visual cortex processes information within the first 100-200 milliseconds. This is called a deficit in top-down processing, because the PFC, a "cognitive" region of the brain (at the "top" of the cognitive ladder) is assisting the visual cortex, a brain region "lower" on the cognitive ladder.
So we wanted to use EEG in our experiment to show that what Barceló showed in PFC patients for attention would be the same for working memory. One of my collaborators, Ed Vogel, had a cool Nature paper where he showed that EEG over visual cortex could actually be used to predict an individual person's working memory capacity with decent accuracy.
But I wanted to extend this idea even further. Another colleague of mine, Earl Miller, had a Nature paper a few years back where they showed that a small, subcortical region in the frontal lobes, the BG, might actually be training the PFC in certain kinds of learning tasks. That, along with a lot of other evidence, supports this "network" idea of brain functioning where different brain regions have to act in concert to give rise to complicated behaviors.
Like I said earlier, working with patients with brain lesions is one of the few methods we have in cognitive neuroscience to get at some questions of causality. You can stick 1000 people into an fMRI machine and show that blood flow increases in some brain region in response to a task, but you still can't say that that brain regions is doing the task. All it takes is one person with a lesion to that region who can still do that task to disprove causal statements in the brain imaging literature.
So we went into this experiment with the idea that, even though PFC lesions tend to be much larger than BG lesions, because of where the BG sit in anatomical relationship to the PFC and the rest of the brain, we hypothesized that patients with the (smaller) BG lesions would do worse at the task.
As you can (maybe) see in the figure above (part C), that's just what we found. Like the Barceló study, we show that patients with PFC lesions are fine when the stimuli enter the "good" half of the brain, but are worse (compared to healthy control subjects) when they enter the "bad" half of the brain. These deficits are both attentional and working memory in nature. In contrast, patients with BG lesions do worse regardless of which hemisphere the visual information enters, but this is specifically a working memory deficit; attention is intact. Especially interesting to me is that the BG patients do extra worse during the first 25 trials or so, which suggests a learning deficit, too.
Okay, so this post is getting really long, so I'll spare all the intricate details, but there's one neat point in the paper that we had to put into the supplemental materials that I really liked. One of the reviewers asked about a piece of our data that I thought was interesting and intriguing as well, so we dug into it a little more. Basically, there was a great PNAS study where (yet another colleague of mine) Ole Jensen showed that brain oscillations may actually be generating those EEG signals that Vogel showed to predict memory capacity.
Since I'm an "oscillations" guy, I really dug this work. So we looked at those same oscillations and found that patients have larger alpha oscillations over the damaged hemisphere (as can be seen in the above figure), specifically over the visual cortex. The more abnormal their alpha was, the less predictive of memory load their EEG was. We interpret this as a loss of top-down facilitation of visual attention/working memory. This latter point is super cool, and one I'd like to investigate in future experiments.
Anyway, take home messages:
1. Networks: important!
2. Basal ganglia: cognitive!
3. Alpha: interesting!
This work was financially supported by the (sadly defunct) American Psychological Association Diversity Program in Neuroscience grant 5-T32-MH18882 (to B.V.) and the National Institute of Neurological Disorders and Stroke grants NS21135, NS21135-22S1, and PO40813 (to B.V. and R.T.K.).
Voytek B and Knight RT (2010). Prefrontal cortex and basal ganglia contributions to visual working memory. Proc Natl Acad Sci USA.