Parietal Lobe: From Action Organization to Intention Understanding
Inferior parietal lobule (IPL) neurons were studied when monkeys performed motor acts embedded in different actions, and when they observed similar acts done by an experimenter. Most IPL neurons coding a specific act (e.g. grasping) showed markedly different activiations depending on contexts (e.g. grasping for eating or for placing).
Many motor IPL neurons also discharged during the observation of acts done by others. The context-dependence also applied to when these actions were observed.
These neurons fired during the observation of an act, before the beginning of the subsequent acts specifying the action - thus these neurons not only code the observed motor acts but also allow the observer to understand the agent’s intentions.
Posterial parietal cortex has traditionally been considered as the association cortex. Besides “putting together” different sensory modalities, also
- Codes motor actions
- Provides the representations of these motor actions with specific sensory information.
Experiment
Set 1
Investigates neuron selectivity during actioin performance:
- Monkey starts from a fixed position, reache for and grasp a piece of food located in front of it and brigh the food to mouth.
- Monkey reach for and grasp an object, and place it in a box on the table.
- Monkey reach for and grasp an object, and place it into a container near its mouth.
Most observed neurons discharged differentially with respect to context.
Control for confounds:
- Grasp an identical piece of food in both conditions - rewarded with the same piece of food in situation 2.
- Situation 3 aims to determine if arm kinematics determined the different firing behavior - turns out to be pretty much identical to situation 2.
- Force exerted grasping - mechanical device that could hold the objects with two different strengths used, no change in neuron selectivity.
Set 2
Study 41 mirror neurons that discharge both during grapsing observation and grasping execution, in an experiment where the monkey observed an experimenter performing those actions.
Results:
- Some neurons are selective for grasping to eat, others for grasping to place - most have the same specificity as when the monkey performed those actions.
- Most neurons have the same selectivity regardless of whether the grasped object is food or solid. Some discharged more strongly in the placing conditions when the grasped object was food.
The authors suggest that while counter-intuitive, this should makes sense since motor acts are not related to one another independent of the global aim of the action but appear to form prewired intentional chains in which each motor act is faciliatted by the previously executed one. This reminds me of Churchland and Shenoy’s theory about action planning as a trajectory in the Null-Space.
Consequences
The authors suggest that the monkey’s decision on what to do with the object is undoubtedly made before the onset of the grasping movement. The action intentionis set before the beinning of the movements and is already reflected in the first motor act. This motor reflectioin of action intention and the chained motor organization of IPL neurons have profound consequences on […] understanding the intention of others.
While I can see where he reaches this conclusion, and it would nicely explain why passive observation training for BMI works (even though mostly M1 and S1 neurons are recorded), but this explanation would make more sense if most of the difference in discharge rate happens BEFORE the monkey physically grasp the object/food…
It has been generally accepted that mirror neurons allow the observing individual to understand the goal of the observed motor act. The authors’ interpretation then deduces that because the monkey knows the outcome of the motor act it executes, it recognizes the goal of the motor act done by another individual when this act triggers the same set of neurons that are active during the execution of that act.
Applications to BMI
This paper suggests mirror neurons as a mechanism for evaluation of intention, it also suggests that depending on context and object, a simple grasp action may be coded differently. Thus a BMI arm trained on passive observation of grasping an object to place might not work well when the user tries to grasp a food to eat. This is similar to when the Schwarz 10-DOF arm encountered, when the human subject had trouble grasping object. The problem only went away when the training was done in a virtual space.