Back in the 1990’s, one of the most intense battlegrounds in systems neuroscience was in monkey posterior parietal cortex. Labs competed to claim what a little strip of cortex called lateral intraparietal area (LIP) really does – decision, movement planning, attention, reward, or all of the above – mostly using single cell recording in behaving monkey. The experiments were (and still are) tough: standard operating procedure requires a well-trained monkey who will perform hundreds if not thousands of trials a day and then isolating neurons one at a time to find ones that respond during some interesting part of the trial. And then lots and lots of repetition so that you can average over many neurons. All things considered, it’s remarkable how much the field has been able to learn with this toolbox.
Fast forward to present day, there’s a new kid on the block. As I discussed a few weeks ago, rodent behavior and physiology is booming. People are taking on questions previously studied mainly in primates and are taking full advantage of the recent storm of new techniques. This is typified by today’s paper by Chris Harvey, Philip Coen, and David Tank, which goes back to the question – what does posterior parietal cortex do during a decision task? They imaged populations of neurons while mice used visual cues to navigate a virtual maze. Just like in primates, individual neurons were selective for different choices that the mouse made. But unlike in primate parietal cortex, where neurons tend to have sustained responses leading up to the time of decision, individual neurons responded transiently in different portions of the trial. So as a population, different choices were represented by distinct sequences of neuronal activity. This kind of sequential firing has been seen in other parts of the rodent brain such as hippocampus, but not in posterior parietal cortex.
So, you might ask, what’s new here? Are you guys just going ga-ga over 2 photon microscopy and the idea of a mouse running on a trackball? Parietal cortex represents decision – didn’t the monkey folks figure that out years ago? Yes, but not with the answer that the representation lies in distinct sequences of firing across many cells. Whether this reflects a real difference between species remains to be seen. Regardless, this study takes a big step in characterizing a part of the rodent brain that, if the primate work is any indication, could be fruitful ground for studying cognition. And it demonstrates how looking at the system at a different level (population dynamics) can give you a different answer than what you might see by averaging over single cells.
Going back to something I discussed in my previous post, it seems that whenever I discuss the new wave of rodent studies with primate researchers, more often than not there is serious resistance to the idea that rodents could be a useful model for studying such things as vision or decision (two currently growing areas in rodent research). “They don’t use vision” or “they don’t have much frontal cortex” seem to be mantras. Those arguments are tired – vision may not be their primary sense, but rats and mice obviously see well enough to navigate. Yes, they have less frontal cortex than monkeys, but monkeys have less frontal cortex than humans and we still study them as a model of human cognition.
Sure, rodents don’t see as well as primates, and simple detection or discrimination tasks that take monkeys no time to learn seem to be bafflingly hard to train in rats – and even harder in mice. There is going to be a limit to the kinds of processes that can be studied in these animals. It may well be that researchers are going to have to get more creative in crafting experiments that will be more suited to their innate behavior (such as foraging and exploring). But rodent vision and decision studies are here to stay and new technology for studying and manipulating neural circuits lays out exciting possibilities for asking questions that complement, not compete, with what can (or should) be asked in experiments on higher animals.