TCMC is a feisty meeting. For me, it’s the conference before the conference (Did I mention I’m in DC for the Society for Neuroscience conference, or SfN? Well I am). In one day I absorb more than in the rest of the conference combined. The debates in motor control have evolved over the years in such a way that makes me think science does sometimes get us somewhere. It’s thanks in no small part to a small but incredibly passionate and often entertaining group of motor learning researchers. Some highlights:
Shortened eye movements are a feature, not a bug, of motor disorders
Pavan Vaswani from Reza Shadmehr’s lab at Johns Hopkins gave a well-argued talk on saccadic hypometria. Saccades are eye movements, and sometimes they fall short of their goal, particularly for people with certain motor disorders (Parkinson’s disease, cerebellar ataxia, and Lewy body dementia, to name a few). Pavan argued that rather than seeing these too-small eye movements as part of the problem, perhaps they are part of the solution. In motor disorders, executing movements becomes difficult, and performance becomes wildly variable. If you’re not super confident in your accuracy, it behooves you to be a bit more conservative–the size of your errors will be smaller when you make a series of small movements instead of one big, exuberant one. By looking at the trajectories of the eye movements of people with saccadic hypometria and comparing them to computational models of optimal motor control, Pavan and Reza determined that these movements weren’t “cut short” prematurely, but rather had been intended to be short from the outset. This was an elegant example of a way to find out whether a phenomenon represents a pathology or a compensation for pathology–in this case, short saccades appears to be the latter.
Rats on a treadmill: Running towards understanding the striatum
Pavel Rueda-Orozco from David Robbe’s lab in Barcelona presented data that used a new task they’d designed to see how the brain represents the timing of habitual actions. Anyone who’s ever fed exams into a Scantron machine knows you can increase your machine-feeding accuracy (and save valuable time) by developing a robotic rhythm. If grad students operated less in isolation, I have no doubt there would be chain gang-style chants to help this. But what Pavel and David did was put rats on a treadmill, and train them to wait a certain amount of time before advancing into an endzone to receive a reward. If they advanced too soon, they got a time out, delaying the possibility of future rewards. At first the rats run erratically up near the endzone, falling behind in spurts on the treadmill to learn the proper delay. But they all soon figure out to let the treadmill carry them to the back of the chamber, and to then advance at a steady pace towards the endzone. Adorably, their habits were idiosyncratic–one stud rat they had named “Michael Jackson” would go to the back, hop/dance around for a few seconds, and then sprint. As the rats learned the task, the firings in an area called the dorsolateral striatum, part of the brain’s habit formation circuitry, became correlated with their position and speed. Interestingly, when the researchers put a plate in front of untrained rats to guide them in the same paced run that the trained rats did, these rats’ neurons fired in a way that was uncorrelated with their movements, showing that the firing of these neurons was effectively “keeping track” of how they were doing on this habitual movement. These data shed new light on habit formation. Studies have traditionally run rats through a T-shaped maze to find a reward in one of two arms, while leaving their movement speed unconstrained. By making their speed part of the habit, they showed that the dorsolateral striatum doesn’t so much initiate actions as it does guide their execution with information about context and kinematics.
The cerebellum is more of a storage unit for motor memories than an editor
Alkis Hadjiosif from Maurice Smith’s lab at Harvard pointed out that in several studies of people with cerebellar damage, some have shown that they can adjust their motor behavior more than others. This begged the question: do you need your cerebellum for this or not? What is the deal there? These studies used visuomotor rotation tasks, where people make reaching movements toward targets. As they reach, their on-screen cursor is rotated–if you’re reaching for the 12:00 position on a clock, you would suddenly see your cursor start going towards 10, and your job is to re-train your arm to go towards 2:00 to compensate for it. This task is commonly used to investigate how and whether people can learn new motor skills, like riding a bike or doing a cartwheel. It turns out that in some studies, patients with cerebellar damage reached to many targets, making the average time between a visit to a given target and the soonest return to that same target around 22 seconds. In others, fewer targets were used, making repeat visits to a target about 5-7 seconds apart. The Smith lab had proposed that there are actually two things going on when we adjust motor behaviors: a temporally stable adjustment that lasts a long time, days even, and a temporally labile adjustment that is quicker, more nimble, more reactive to change, but also shorter-lived. With the longer intervals between trials, this had all but disappeared. Cerebellar damage patients could compensate for rotations at shorter delays but not longer ones, suggesting to them that these people relied on their temporally labile adjustments to get the job done. When the delay was long, they were required to retrieve a stable motor memory and just couldn’t do it–strong evidence for the cerebellum’s role as a stabilizer of motor memories.
To be good at a thing, practice that thing, go practice other things, and then come back to the first thing.
Nicholas Wymbs from Pablo Celnik’s lab at Johns Hopkins had people do a task that involved squeezing a pressure sensor to move a cursor to different targets. Of five targets, you’d go maybe 2-4-1-3-5, squeezing different amounts to get the cursor to each one. This task, called the Sequential Visual Isometric Pinch Task (which ends up being fun to say–SVIPT), is another popular one for studying how people learn to do things. They found that people learned best if they first did the task, then came back and did a variable version of the task–learning to do the task with the squeeze-to-distance ratio turned up or down a bit in between tries–and then tried the task again. This is counterintuitive–you’d think to be good at something you should just do that thing over and over again. But actually, variability is a good thing sometimes. It gives you a sense of the landscape. So to be good at free-throws from, oh, what’s that line called in basketball where they throw the um, like, the penalty shots or whatever? Anyway if you want to be good at that, you should go off and try throwing from lots of different places, and then you’ll have a better understanding of how to do exactly THAT one throw, because you have your motor memory of it situated in a landscape of other, neighboring motor memories. Critically, they said the variable SVIPT only worked if people had already learned the one squeeze-to-distance mapping they’d be tested on. And it only worked if they had waited a while since learning it–the memory had to be “consolidated,” or basically all-the-way baked instead of half-baked, before you could return to it, try out some riffs on it, and then see a benefit of your riffing. When people did the variable task, people who were quicker to correct after big jumps in the squeeze-to-distance ratio turned out to be better at the “original” ratio later on. Physical therapists might soon be able to use this information to inform the schedule of their patients’ therapies for maximum benefit.
All right, folks. That’s it for now. Time for me to dash out of my comely Ikea-laden VRBO and off to the “real” conference. I’m already exhausted.