or Memory interference at the single neuron level
The first task that the subjects had to learn (task A) was a visuomotor rotation (see animation below). In this task, the subjects are asked to control a cursor by moving their hand. Initially, the cursor motion is matched to the hand motion. Then, during visuomotor rotation, the cursor motion is rotated by a given angle such that, for the first trials, the hand moves to the target but the cursor goes in a different direction (before learning in the animation). After a few trials, the subject are able to compensate for this rotation and the cursor moves towards the target when the hand is directed away from it (after learning).
Both the arbitrary association and the visuomotor rotation tasks elicited changes in motor cortex neuron activity. Usually, neurons in that area preferentially fire before movements in a specific direction. However, this pattern can be modulated during the learning process. Indeed, there was a specific increase in firing rate in the learned direction after visuomotor rotation learning (Panel A in the figure below) while there was a color-specific increase in firing rate for the arbitrary association task (Panel B). In other words, there was a direction-specific but color-insensitive increase in firing for visuomotor rotation learning and a direction-insensitive but color-specific increase in firing rate for the arbitrary association task.
In contrast, the motor memory of the visuomotor rotation (task A) was damaged when a visuomotor rotation in the opposite direction (task B2) was learned afterwards (retrograde or retroactive interference). This interference was visible at the single cell levels as well . There wasn't any neurons that encoded the two opposite visuomotor rotations together. Rather, the neurons that were modulated by the learning of the first visuomotor rotation (task A) did not maintain this new firing patterns after the learning of the opposite visuomotor rotation (task B2).
In sum, the interference observed at the behavioral level was also observed at the single cell level in the motor cortex.
I think that this paper is very interesting but it does not present the full story. Indeed, two interfering tasks can be learned together under special conditions. For instance, despite the fact that the learning of two opposite force-fields interfere, visual cues can be used to elicit the learning of both force-fields simultaneously (Cothros et al. 2009).
In addition, interference is not similar to destruction of motor memory. For instance, after interference between task A and B (e.g. two opposite visuomotor rotations), relearning of task A is faster than the original learning of task A and the memory of task A is still observable (Pekny et al. 2011). Therefore, interference does not relate to destruction (as suggested by Zach et al.'s data) but to masking of motor memories. Unfortunately, the study of Zach and colleagues does not uncover whether the memory trace of A is still present in the motor cortex during interference.
Cothros, N., Wong, J., & Gribble, P. L. (2009). Visual cues signaling object grasp reduce interference in motor learning. Journal of neurophysiology, 102(4), 2112-20. doi:10.1152/jn.00493.2009
Pekny, S. E., Criscimagna-Hemminger, S. E., & Shadmehr, R. (2011). Protection and Expression of Human Motor Memories. Journal of Neuroscience, 31(39), 13829-13839. doi:10.1523/JNEUROSCI.1704-11.2011