|dc.description.abstracteng||As humans, the only way for us to interact with the world around us is by utilizing our highly trained motor system. Therefore, understanding how the brain generates movement is essential to understanding all aspects of human behavior. Despite the importance the motor system, the manner in which the brain prepares and executes movements, especially grasping movements, is still unclear. In this thesis I undertake a number of electrophysiological and computational experiments on macaque monkeys, primates showing similar grasping behavior to humans, to shed light on how grasping movements are planned and executed across distributed brain regions in both parietal and premotor cortices. Through these experiments, I reveal how the use of large-scale electrophysiological recording of hundreds of neurons simultaneously in primates allows the investigation of network computational principles essential for grasping, and I develop a series of analytical techniques for dissecting the large data sets collected from these experiments.
In chapter 2.1 I show how large-scale parallel recordings can be leveraged to make behavioral predictions on single trials. The methods used to extract single-trial predictions varied in their performance, but population-based methods provided the most consistent and meaningful interpretation of the data. In addition, the success of these behavioral predictions could be used to make inferences about how areas differ in their contribution to preparation of grasping movements. It was found that while reaction time could be predicted from the population activity of either area, performance was significantly higher using the data from premotor cortex, suggesting that population activity in premotor cortex may have a more direct effect on behavior.
In chapter 2.2 I show how preparation and movement intermingle and interact with one another on the continuum between immediate and withheld movement. Our population-based and dimensionality reduction techniques enable interpretation of the data, even when single neuron tuning properties are highly temporally and functionally complex. Activity in parietal cortex stabilizes during the memory period, while it continues to evolve in premotor cortex, revealing a decodable signature of time. Furthermore, activity during movement initiation clusters into two groups, movements initiated as fast as possible and movements from memory, showing how a state shift likely occurs on the border between these two types of actions.
In chapter 2.3 I show that the question of how motor cortex controls movement is an ongoing issue in the field. I address crucial details about recent methodology used to extract rotational dynamics in motor cortex. I show how a simple neural network simulation and novel statistical test reveal properties of motor cortex not examined before, showing how models of movement generation can be essential tools in adding perspective to empirical results.
Finally, in chapter 2.4 I show how the specificity of hand use can be used as a tool to dissociate levels of abstraction in the visual to motor transformation in parietal and premotor cortex. While preparatory activity is mostly hand-invariant in parietal cortex, activity in premotor cortex dissociates the intended hand use well before movement. Importantly, we show how appropriate dimensionality reduction techniques can disentangle the effects of multiple task parameters and find latent dimensions consistent between areas and animals.
Together, the results of my experiments reinforce the importance of seeing the motor system not as a collection of individually tuned neurons, but as a dynamic network of neurons continuously acting together to produce the complex and flexible behavior we observe in all primates.||de