The visual system has a limited capacity for capturing and processing the richness and intricate detail of the surrounding environment. Visual information that arrives in the retina is converted from relative light intensities to patterns of excitation and then transmitted to a hierarchy of visual areas, which process and combine increasingly complex features of the visual signal to form a visual percept. At each stage, the amount of task or stimulus-related information a neuron can encode depends on the separability of its responses to different conditions. Using electrophysiological recordings of extracellular spiking activity from single neurons in awake behaving monkeys, we explored ways to quantify information in neuronal firing rates in order to address specific questions about sensory and cognitive signals in visual cortex and frontal cortex during different behavioral contexts. In the first study, we addressed a question of latency differences in the visual pathways that process color using a passive fixation task. Color processing occurs generally along two separate chromatic pathways, and previous work has indicated that information from the two pathways arrive with a relative lag in primary visual cortex. However, to form a perception of color, these two pathways must converge at some stage of visual processing. We used information theory to examine the timecourse of chromatic information in neurons further up the visual hierarchy in area V4, which has been implicated as an area with an important role in color processing. We found that on average, information specific to each pathway arrived simultaneously in V4, suggesting that color signals from the different pathways converge at some point within or before V4 in the visual hierarchy. In order to select behaviorally relevant information from the large amount of visual information available, shifting the focus of gaze (via saccadic eye movements) and directing attention provide ways to allocate processing resources to selected locations in visual space. Studies have shown that perceptual enhancements at behaviorally selected spatial locations are accompanied by enhanced processing in visual cortex. The mechanisms by which neurons in the brain control the selection of sensory signals remain unclear. Previous works suggest that the control of attention and eye movements, as well as the modulation of sensory representations, originate in a distributed network that includes the frontal eye field (FEF) in frontal cortex. We studied responses of single neurons recorded separately in area V4 and the FEF of monkeys engaged in different visuospatial selection tasks. In addition to firing rate responses, we examined the trial-to-trial response variability, which has been suggested to reflect behavioral state. During natural vision, the eyes make frequent movements to selected targets. To better understand how these gaze shifts influence visual processing, we examined selective visual processing in area V4 during saccade preparation. We found that V4 neurons show transiently enhanced stimulus discrimination at the saccade target. This enhancement is due in part to changes in response magnitude, but may also be facilitated by reduced variability (increased reliability) of sensory representations. The similarity to effects of covert attention and experimental manipulations of FEF activity provides further evidence that the mechanisms driving visual modulation during saccade preparation and covert spatial attention rely on common neural resources. To explore signals that likely modulate visual responses through feedback connections, we examined the role of FEF neurons in the maintenance and selection of spatial information. In a task that required remembering and directing spatial attention to a cued location while withholding eye movements, neurons in the FEF exhibited spatially selective persistent activity, which continuously tracked the location of the cue. Moreover, this maintenance of spatial information correlated with successful deployment of attention. Despite robust visual and cognitive firing rate modulations that predicted behavioral performance on the task, declines in response variability appeared to be most effectively driven by visual stimulation, rather than spatial working memory or attention. This indicates that, at least in the FEF, behavioral engagement alone is not sufficient to drive changes in variability. Instead, changes in response variability may reflect shifts in the balance between feedforward and recurrent sources of excitatory drive.

Cortical brain-machine interfaces, or BMIs, is a relatively new field with the potential to provide many different clinical treatments, particularly for fully paralyzed patients. In these applications, multichannel electrode arrays are implanted into motor cortical areas in order to extract useful control signals. My research focuses on taking proof-of-concept academic BMI systems, and solving the engineering challenges that currently prevent them from being used in a clinical setting. These challenges include running a BMI for more than a few hours or a single day, and finding ways to minimize the size, cost, and operational complexity of the complete system. This dissertation includes an analysis of neuron stability over long timescales. I will show that the relationship between neurons in motor cortices and behavior remains stationary over time despite substantial noise, which could mitigate some concerns about long-term BMI performance. I will also discuss the development of HermesC, a wireless system for recording multichannel neural data from freely moving primates. This device dramatically reduces the size and cost of current recording technology for real-time neural prosthetic systems, and could be useful for human clinical trials. It may also enable neural prosthetic studies with animals in a less constrained setting. Combining traditional neural recordings with overnight wireless neural recordings, I will also show that there are substantial changes in neural waveforms from single neurons across days. However, the quality of neural decodes (the extraction of useful control signals) is only slightly improved by sorting individual units rather than using simple threshold crossings. This may enable long term BMI operation because multiunit neural "hash" on electrode arrays tends to persist for a long time, perhaps years, after single neuron signals have declined due to various tissue responses. In fact, other recent work from this project has demonstrated high performance neural decodes using only threshold crossings on arrays ~2.5 years after implantation.

Attention involves selecting the most relevant information from the abundance of information available in the sensory environment. When we attend to an object, neurons in the brain become more selective to processing information about the attended object at the expense of processing information about other, irrelevant objects. While the phenomenology of attention is well characterized, little is known about the mechanisms by which attention modulates neural responses to influence perception. In this thesis, I explore neural mechanisms of attention with a combination of approaches at several levels: 1) whole-brain functional imaging to characterize the spatiotemporal dynamics of forebrain networks that are engaged during attention, 2) extracellular electrophysiology in the midbrain of tranquilized barn owls to measure and characterize gamma oscillations, a prominent neural signature of attention, and 3) pharmacological techniques in a chick midbrain slice preparation to understand the cellular and circuit mechanisms involved in generating and regulating the structure of these oscillations. I conclude by describing ongoing work on developing neuromorphic (hardware) models of putative mechanisms of attention, as well as the development of a novel behavioral paradigm for rapidly and reliably measuring attention in birds. This work highlights the power of integrative and multi-level approaches for exploring the neural basis of cognitive functions, such as attention.

In age-related macular degeneration (AMD) and retinitis pigmentosa (RP), two leading causes of blindness, the photoreceptor layer of the retina is degenerated while the rest of the retina is well preserved. The function of the photoreceptors is very similar to that of solar cells. Upon receiving light, they activate the inner layers of the retina by producing electrical and chemical signals. These signals are then processed and compressed by a complex circuit of retinal neurons (horizontal cells, bipolar cells, amacrine cells, and ganglion cells) and sent to the brain. The brain perceives these data as sight. Electronic retinal implant systems seek to restore sight in AMD and RP by electrical stimulation of the surviving retinal neurons. Currently the more dominant systems consist of a microelectrode array, which is placed directly on the ganglion cells (epiretinal). In this approach, a camera mounted on goggles captures video which is then processed by a pocket computer. The power and data are then transmitted wirelessly to an extraocular receiver unit. Through an intraocular cable, the receiver unit sends appropriate electrical stimuli to the array of microelectrodes. The electrodes then stimulate the ganglion cells by passing current pulses through the tissue. These stimulations are perceived by the brain as spots of light. The epiretinal approach has some disadvantages. Because the electrodes directly stimulate the ganglion cells, the image processing and data compression capabilities of the retina are not utilized. Placing the extraocular receiver unit and connecting it via a cable to the microelectrode array significantly complicates the surgical procedure and increases the chance of post surgical complications. Additionally, the perceived images are independent of the eye movements. We have developed an integrated photovoltaic monolithic silicon retinal implant that requires no electrical power or data connection. In our design, a miniature camera captures video that is processed by a pocket computer before being projected into the eye at a near-infrared wavelength ([Lambda] = 905 nm) onto the silicon implant located in the subretinal space (area in the back of the retina). The implant consists of a two-dimensional array of photovoltaic pixels. The projected image is provided in pulsed fashion and each pixel element consists of up to three series-connected photovoltaic cells such that the pixels deliver current pulses that are sufficiently strong to stimulate the remaining functional neural cells. The current pulses are interpreted as visual images by the brain. Placing the implant in the subretinal space allows for utilization of the existing image processing and data compression functions of the retina. Each pixel receives both power and data directly through laser radiation. This eliminates the need for a separate wireless receiver unit and simplifies the surgical procedures and reduces the post-surgical risks. Additionally, in this approach, eye movements change the perceived images. The novelty of the work reported here is the integration of photovoltaic devices in a MEMS process that allows the implant to deform to the natural curvature of the eye, while also providing isolation between the bodies of the three series-connected subpixels that make up each pixel. This was achieved by patterning the implants into an array of pixels connected together by deformable silicon flexures. In addition, the trenches also provide electrical isolation between the three series-connected photovoltaic subpixels. Curving the implant is advantageous since the complete implant is in focus, resulting in optimum quality of vision perceived. Curved implants can also be substantially larger than planar implants and can hence cover a larger part of the field of view. A curvable implant also causes no mechanical strain or abnormality in the eye. Usage of three series-connected subpixels per pixel improves the impedance matching to the surrounding tissue and enhances the injected current per pixel allowing for higher resolution implants. Fabricated implant with a resolution of 64 pixels/mm2 can inject sufficient current for neural activation at safe optical intensities. Additionally, the exchange of nutrients and waste, which is necessary for the survival of the retinal cells, is provided by diffusion through the trenches that define the silicon devices.

By restoring the ability to move and communicate with the world, brain machine interfaces (BMIs) offer the potential to improve quality of life for people suffering from spinal cord injury, stroke, or neurodegenerative diseases, such as amyotrophic lateral sclerosis (ALS). BMIs attempt to translate measured neural signals into the user's intentions and, subsequently, control a computer or actuator. Recently, compelling examples of intra-cortical BMIs have been demonstrated in tetraplegic patients. Although these studies provide a powerful proof-of-concept, clinical viability is impeded by limited performance and robustness over short (hours) and long (days) timescales. We address performance and robustness over short time periods by approaching BMIs as a systems level design problem. We identify key components of the system and design a novel BMI from a feedback control perspective. In this perspective, the brain is the controller of a new plant, defined by the BMI, and the actions of this BMI are witnessed by the user. This simple perspective leads to design advances that result in significant qualitative and quantitative performance improvements. Through online closed loop experiments, we show that this BMI is capable of producing continuous endpoint movements that approach native limb performance and can operate continuously for hours. We also demonstrate how this system can be operated across days by a bootstrap procedure with the potential to eliminate an explicit recalibration step. To examine the use of BMIs over longer timescales, we develop new electrophysiology tools that allow for continuous multi-day neural recording. Through application of this technology, we measure the signal acquisition stability (and instability) of the electrode array technology used in current BMI clinical trials. We also demonstrate how these systems can be used to study BMI decoding over longer time periods. In this demonstration, we present a simple methodology for switching BMI systems on and off at appropriate times. The algorithms and methods demonstrated can be run with existing low power application specific integrated circuits (ASICs), with a defined path towards the development of a fully implantable neural interface system. We believe that these advances are a step towards clinical viability and, with careful user interface design, neural prosthetic systems can be translated into real world solutions.

The fruit fly, Drosophila melanogaster, is a key model species for biological research. Trained humans can manipulate, inspect and dissect individual flies, but these operations are often rate-limiting bottlenecks for screening and experimentation. Here I present a high-speed, economical robot for handling non-anesthetized adult flies. Using machine vision the robot tracks a fly's thorax and gently grabs it ~400 ms after targeting. The robot can then translate and rotate the picked fly, inspect its phenotype, dissect or release it, and thereby rapidly prepare multiple flies sequentially for a wide range of experimental formats. In one illustration, the robot restrained flies and dissected the cuticle to permit two-photon imaging of neural dynamics. In another, the robot sorted flies by sex. The robot's tireless capacity for accurate, repeatable manipulations will enable experiments and biotechnology applications that would otherwise be totally infeasible, especially those requiring high-throughput capture, testing and assessment of individual fly attributes.

Movement preparation allows the rapid and accurate execution of voluntary movements, and can be influenced by factors that may change from moment to moment, such as attention and differences in stimulus properties. Consequently, movement preparation unfolds differently across many repetitions of the same movement. Averaging neural responses across many repetitions is necessary to interpret single-cell recordings, but diminishes our ability to characterize the dynamics of the underlying process. A central question in neuroscience, and also of fundamental clinical importance, is to understand how these plans develop in the brain. Several research groups are starting to build prosthetic devices that are controlled directly by neural activity in motor areas of the brain (Nicolelis, 2001; Donoghue, 2002; Musallam et al., 2004; Schwartz, 2004; Santhanam et al., 2006; Hochberg et al., 2006; Mulliken et al., 2008; Andersen et al., 2010), but the extent to which these can be developed may hinge critically upon our understanding of the neural basis of motor preparation. Simultaneous recording from populations of neurons allows dynamics of movement preparation to be estimated on single trials. Our goal is to characterize these dynamics, to gain insight into the process underlying movement preparation. Here, we recorded peri-saccadic activity from ensembles of neurons in an oculomotor area, prearcuate cortex, in two monkeys. While monkeys performed visually-guided eye movements, we measured firing rates of a population of neurons using a 96-electrode array. We studied (1) the responses from a heterogeneous population of prearcuate cortex neurons involved in decision-making and movement preparation, (2) the relationship between saccade initiation times and responses from the neural population, and (3) how these responses compared to those recorded in PMd, a cortical area involved in arm movements. The array data from prearcuate allowed us to compare responses from individual neurons with previous findings, but also allowed us to analyze the population dynamics of movement planning, by using techniques applied to the reach system. We found that ensemble responses from diverse populations of prearcuate neurons (1) can be visualized as trajectories moving through a low-dimensional neural state space, (2) reflect visual, decision-, and movement-related aspects of the task, and (3) correlate with a monkey's reaction time on a trial-by-trial basis. Further, the single-trial relationship between ensemble activity in prearcuate cortex and saccadic reaction times was qualitatively and quantitatively very similar to the relationship between ensemble activity in PMd and corresponding reach reaction times. This framework for analyzing neural population activity and dynamics should permit new extensions of single-neuron-level models, and may offer further insight into general mechanisms of movement preparation across motor systems.

The ultimate purpose of the motor system is clear: it exists to control the body. However, despite the motor system being among the longest-studied brain structures, it remains unclear how -- mechanistically -- motor cortex performs this function. Here, a mechanistic approach was taken to investigate how primary motor cortex (M1) and dorsal premotor cortex (PMd) control movement. That is, the goal was to elucidate the dynamics of the motor cortex 'machine.' Monkeys were trained in reaching tasks, and neural signals were recorded from their brains as they performed them. Two broad classes of analysis were used. First, cell-by-cell analyses were combined with cell-type analyses, which permitted examining the activity patterns of excitatory and inhibitory neurons separately. Second, techniques based on dynamical systems analysis (such as dimensionality reduction) were applied, which permitted analysis of neural populations as a whole and abstraction to a somewhat higher level of system function. Three major results and a technical advance are presented. Firstly, we investigated how it is possible for an animal to hold still even as neural activity in motor cortex changes drastically during preparation for the upcoming movement. We found that, contrary to common assumptions, there does not appear to be a 'gate' comprised of high inhibition during preparation. Instead, using the dynamical systems perspective, we found that preparatory activity has a special structure such that it remains in intrinsically muscle neutral, 'iso-force' patterns. Secondly, we searched for coherent dynamics in the movement-time activity of motor cortex. We found that motor cortex appears to obey a relatively simple set of dynamics, dominated by oscillatory patterns. Moreover, the exact neural trajectory is heavily determined ('seeded') by the immediately preceding preparatory activity. In order to causally perturb these dynamics with patterned stimulation and cell-type specificity, we then developed a set of optogenetic techniques for use in primates. Finally, we investigated how the dynamics of the decision-making process are reflected in motor cortex. To do so, we combined a novel decision-making paradigm, many simultaneous neural recordings, and single-trial analytical techniques. Preliminary results are given for this final section, demonstrating the presence of vacillation in monkey decision-making. In summary, we found that preparation and movement can be understood as an oscillatory dynamical system seeded by preparatory activity that lives in an iso-force space, that inhibitory and excitatory neurons seem to play more similar roles in the dynamical system than might be expected, and that moment-by-moment processes of motor decision-making can be seen in motor cortex.

Hypokinesia is one of the most disabling movement abnormalities caused by Parkinson's disease (PD). It is a clinical term that refers to the general reduction of mobility experienced by people with Parkinson's disease (PWPD). Hypokinesia affects movements in three ways: 1) movements are slow 2) movements are performed with reduced amplitude compared to what is required for the task and 3) it takes longer for PWPD to initiate movements. All of these impairments affect the ability of PWPD to perform common daily activities and can cause feelings of frustration and anger that may lead to altered self esteem, depression, and a reduction in their quality of life. Although effective treatments do exist for well-selected patients, there are many avenues open for improving current treatments, identifying new treatments, and understanding mechanisms of hypokinesia that could potentially lead to more effective therapies and greatly impact the lives of PWPD. In this dissertation we sought to improve current treatments and provide opportunities for future treatments for hypokinesia by addressing three important clinical questions relating to hypokinesia. First, are there immediate improvements in hypokinesia from a surgical procedure called deep brain stimulation (DBS) that can be measured during surgery and used to help guide surgical decisions and optimize clinical outcomes? Second, can we determine what types of movements are affected by hypokinesia in early stage, untreated PD in order to provide an objective metric used to assess an emerging treatment for PD? Third, might perceptual deficits that are linked to sensory processing impairments play a role in the manifestation of hypokinesia? If so, targeting these deficits may provide new and better therapies. We addressed these questions by using the quantitative and computational techniques outlined in the next three paragraphs. A common treatment for advanced PD is a surgical procedure called DBS. The surgery is performed on awake patients, and it entails surgically implanting electrodes (leads) that provide chronic stimulation to the affected brain area that controls movement. Although the treatment is almost always effective, the degree of improvement in hypokinesia varies among patients. The accuracy of the placement of the DBS lead in the affected area is believed to have the most effect on the improvement of hypokinesia. We suggest that using quantitative measurements of hypokinesia to evaluate the efficacy of the location of the DBS lead in improving hypokinesia during the surgical procedure might therefore improve the overall clinical outcomes. The surgical team would then have objective, accurate measurements of the degree of improvement in hypokinesia during the surgery, when the lead's position in the brain could be modified to achieve optimal results. Therefore, we designed a prospective study to measure upper extremity hypokinesia using a quantitative measure of angular velocity. Analysis of 98 DBS procedures performed on 61 patients showed that on average there was an 81% improvement in quantitative measures of hypokinesia from implanting and activating the DBS lead (p< 0.03). This study demonstrated that objective, high-resolution, accurate measurements of improvements in hypokinesia from intra-operative DBS are possible in this highly constrained environment and could therefore be used to help guide surgical decisions and optimize clinical outcomes. PD has no cure, but treatments in the near future may include disease-slowing medications. Although few studies have characterized the motor control abnormalities of very early stage PD, when symptoms are mild and usually unilateral, this group is the most targeted for potential disease-modifying therapeutics. In this study, we asked if quantitative measures of finger, limb, and postural movement velocity could detect hypokinesia in 20 patients with very early stage, untreated PD. The results revealed evidence of significant finger and limb hypokinesia of the patient group's more affected side when compared to the non-dominant side of 19 age-matched healthy adults (HAs) (p=0.001 and p< 0.001, respectively). Furthermore, the patient group's limb movement velocity on the more affected side was significantly slower than their less affected side (p=0.005), highlighting the importance of using an outcome measure that is lateralized in studies of very early stage PD. In contrast to our previous study that revealed significant postural hypokinesia in patients with advanced PD, we did not detect postural hypokinesia in patients with very early stage, untreated PD. Based on these findings, we suggest that the use of quantitative lateralized measures of hypokinesia would be useful in neuroprotective clinical studies of very early stage, untreated PD and may improve the chances of detecting a disease-modifying effect of potential neuroprotective therapeutics. Detecting such a therapy would have a large impact by improving the lives of PWPD. Although hypokinesia is considered a movement abnormality, new research is suggesting that perceptual deficits may play a role in the manifestation of abnormal movements in PWPD. Motor control theory posits that a sensorimotor integration process (SIP) is used by the central nervous system to perceive and control movement by combining internally generated predictions of movement parameters with the processing of sensory feedback. A previous study examining the SIP demonstrated that HAs overestimated their limb position in the direction of movement, and that the error and its variance (VOE) depended on movement duration. Using quantitative measures of hypokinesia, we asked if PWPD showed errors in perceived limb position and if the dependence on movement duration was different from HAs. We used an established computational model of the SIP to explore mechanisms for the error and VOE as a function of movement duration. Twenty PWPD, off medication, and 20 age-matched HAs were asked to estimate the position of their hand after performing 50, slow, non-visually guided wrist flexion or extension movements for a random period of time (< 4.0 sec). Both groups overestimated the amount they moved; however, the PWPD's error and VOE were larger (p< 0.001). More specifically, HAs exhibited increasing error/VOE for small movement durations that reduced/stabilized for longer movement durations. PWPD, however, showed increasing error/VOE with increasing movement duration that did not significantly improve/stabilize. The results from the model revealed an 88% increase in the variance (noise) in the sensory feedback parameter in PWPD compared to HAs, which suggests the PWPD's SIP could no longer effectively access sensory feedback information to correct errors in other components of the SIP due to the large amount of noise in this signal. This study provides experimental evidence that perceptual deficits may play a role in hypokinesia and computational evidence that abnormal processing of sensory feedback in PWPD's SIP could contribute to increased perceptual error in limb position after non-visually guided movements. The work in this dissertation quantified the immediate improvements in hypokinesia from intra-operative DBS, the presences of hypokinesia in early stage, untreated PD, and the degree of perceptual deficits and their dependency on movement duration in PD. Furthermore, this research has provided evidence for possible mechanisms for hypokinesia. Taken together, this work has the possibly to provide immediate improvements for current treatments and provides several platforms for future therapies to treat hypokinesia and improve the lives of PWPD.

Neural recording systems are fundamental to the advancement of brain-machine interfaces that can significantly improve the quality of lives of patients with neurological diseases, such as spinal cord injuries or quadriplegia. This thesis presents two newly developed wireless neural recording systems that are able to provide a high degree of usability and neural decoding accuracy. They are capable of simultaneously transmitting 32 to 96 channels of neural signals detected by an implanted neural sensor array. This work was carried out within the framework of the Hermes project and its technical design challenges will be addressed. The Hermes project is aimed at primarily developing hardware and software tools that extract neural information from the motor cortex. Those tools can enable practical prosthetic devices used to significantly ameliorate the life of patients with neurological impairments that directly affect motor functions. The first developed system, HermesD, is a 32-channel broadband transmission system using an FSK modulated carrier at 24 Mbit/s in the 3.7-4.1 GHz band. The link range extends beyond 20 m and the total power consumption is 142 mW. The HermesD system uses only COTS components and can be easily replicated. HermesD is fully operational and is currently used to transmit broadband neural data for neuroscience research in the Neural Prosthetic Systems Laboratory (NPSL) at Stanford University. HermesD is also planned as the base platform for future human trials to take place in the same laboratory. The second system that represents the next Hermes generation, HermesE, uses a novel UWB transmitter architecture implemented in a custom IC in the 65-nm CMOS technology. The transmitted signal bandwidth covers the 3.6 to 7.5 GHz frequency range. The time domain waveform is digitally programmable, allowing a very flexible control of the output spectrum to avoid interference and to allow multi-band operation. The UWB transmitter chip is part of a 96-channel broadband recording system delivering 40 Mbit/s. Its power consumption is 230 uW for a communication range of about 5 m. The antenna subsystems for these wireless recording devices presented a design challenge given the requirements for small size, large bandwidth and high efficiency. While HermesD has an operating FBW of 10%, HermesE is much more demanding in this respect, with 70% FBW, requiring unconventional antenna structures. The design techniques and performance of the antennas required to meet the specifications of both systems are also addressed in this work.

Optogenetic neuromodulation is giving scientists an unprecedented ability to modulate neural circuits, providing specificity with regards to location, cell type, as well as neuromodulatory effect. On the other hand, electrical stimulation and lesions, methods commonly used to study neural circuits, are lacking in specificity, often affecting both local cells and passing axons, as well as multiple cell types. Our laboratory has been at the forefront of the field of optogenetics, having developed, for use in mammalian systems, Channelrhodopsin-2 (ChR2), an algal light-activated cation channel that depolarizes neurons in response to blue light, and Natronomonas pharaonis halorhodopsin (eNpHR), a chloride pump that hyperpolarizes neurons in response to amber light. These proteins can control neuronal activity with millisecond timescale precision, and through promoters, they can be targeted to specific cell-types in heterogeneous tissue. Along with its specificity, light stimulation with optogenetic tools often allows the recording of neural activity without the artifact that obfuscates recordings with electrical stimulation. The advantages provided by optogenetics allowed us to make a breakthrough in determining the therapeutic mechanism of deep brain stimulation, a robust treatment for Parkinson's disease in which stimulating electrodes are implanted deep in the brain. Using optogenetics, we replicated the effect of deep brain stimulation by modulating cortical inputs into the region where the stimulating electrode is normally placed. Combined with other corroborating publications, a hypothesis is emerging that electrical stimulation deep in the brain actually produces its effect by modulating cortical projections to the deep brain region. Based on this concept, several clinical studies are being done to better understand the cortical role in Parkinson's disease and determine whether cortical stimulation (potentially non-invasive), could be an alternative to the invasive implants currently used. In order to perform these experiments, we studied the transmission of visible light in brain tissue to estimate the volume of activation produced by optogenetic stimulation and developed a device to measure fluorescence in awake, behaving animals, allowing the quantification of virally transfected gene expression over time, as well as the localization of expression along axon bundles. The knowledge gained from these experiments will have a significant impact on future experiments in the broader field of optogenetics.

Neural prostheses translate signals from the brain into useful control signals, manipulating end-effectors such as computer cursors or robotic arms. Their aim is to offer greater interaction with the world for patients suffering from limb dysfunction due to spinal cord injury, neurodegenerative disease, and other conditions leading to limb paralysis. Prior intracortical electrode neural prosthesis studies have demonstrated compelling proof-of-concept systems, but barriers to successful clinical translation still remain, such as performance and robustness. Measures of performance include the speed, accuracy, and bitrate of the system. Robustness refers to the sustained performance of the system within a day and across days. The work presented here demonstrates algorithms and advances for neural prostheses that increase both performance and robustness. The recalibrated feedback intention trained Kalman filter (ReFIT-KF) increased performance by at least twofold compared to previously reported decoders, approaching the speed of natural arm movements. It achieved bitrates of up to 4.5 bits per second (bps) and communication rates of up to 10 words per minute (wpm) when used on a typing task. These results were reliable and repeatable for hours at a time across 4 array-years between two subjects. Utilizing neural spike threshold crossings as a signal source, the ReFIT-KF algorithm also demonstrated sustainable performance without any changes to decoder weights for one year with a degradation rate of 0.05 bps per month. Performance further increased when the ReFIT-KF was combined with an HMM state decoder for the detection of clicks, eliminating the need for hold periods. This combined ReFIT-KF and HMM decoder achieved bitrates of up to 6.5 bps and 15 wpm. Taken together, these findings may help advance neural prostheses closer to clinical viability.

Neurons have a limited dynamic range. To more efficiently encode the large range of natural inputs, neural circuits adapt by dynamically changing their output range as a function of the input statistics. Variance adaptation provides an informative example of this process, whereby neurons change their response characteristics as a function of variance of their input. When their input distribution changes, sensory systems shift and scale their response curves to efficiently cover the new range of input values and they focus on different segments of the frequency spectrum, for example by choosing to average out the noise in a low signal-to-noise ratio environment by low-pass filtering their input and sacrificing resolution. In multiple sensory systems, adaptation to the variance of a sensory input changes the sensitivity, kinetics and average response over timescales ranging from < 100 ms to tens of seconds. Here we present a simple biophysically relevant model of retinal contrast adaptation that accurately captures both the membrane potential response and all adaptive properties. The adaptive component of this model is a first-order kinetic process of the type used to describe ion channel gating and synaptic transmission. We conclude that all adaptive dynamics can be accounted for by depletion of a signaling mechanism, and that contrast adaptation can be explained as adaptation to the mean of a thresholded signal. A diverse set of adaptive properties that implement theoretical principles of efficient coding can be generated by a single type of molecule or synapse with just a few microscopic states. The LNK model helps to highlight important aspects of adaptation by letting us focus on individual computational blocks separately. By using the LNK model, we investigate the source of the adaptive process in On-Off retinal ganglion cells, which show strong changes in their kinetics as a function of contrast. By analyzing properties of the LNK model, we conclude that most of the adaptive effect is due to differences in the threshold of the two pathways, with a smaller contribution from different adaptive kinetics. Adaptive temporal decorrelation in the retina arises due to differential thresholding in two parallel neural pathways.

Any time we move, our brains solve the difficult problem of translating our motor intentions to muscle commands. Understanding how this computation takes place, and in particular, what role the motor cortex plays in movement generation, has been a central issue in systems neuroscience that remains unresolved. In this thesis, we took an unconventional approach to the analysis of cortical neural activity and its relationship to executed movements. We used dimensionality reduction to extract the salient patterns of neural population activity, and related those to the muscle activity patterns generated during arm reaches to a grid of targets. We found that salient neural activity patterns appeared to tightly reflect muscle activity patterns with a biologically-plausible lag. We also applied our analyses to movements that were planned before being executed, and found that a muscle-framework view of the cortical activity was consistent with previously-described predictions of movement kinematics based on the state of the cortical population activity. Overall, our results elucidate remarkable simplicity of the motor-cortical activity at the population level, despite the complexity and heterogeneity of individual cell's activities.

The work presented in this dissertation primarily focuses on decision-related activity in the lateral intraparietal area (LIP) and, secondarily, the dorsolateral prefrontal cortex (DLPFC). In Chapter 1 we review the previous independent investigations indicating that these areas are separately modulated by sensory information, value information and choice appropriate to represent decisions. We argue that when both sensory and value information must be simultaneously integrated to make choices, it is unknown, if, how and when these areas integrate these factors. We present a behavioral paradigm in which animal subjects must combine sensory and value information, on a trial-to-trial basis, to make optimal choices. This paradigm is based on a well-known motion discrimination task; however, in our task the magnitude of the reward associated with each option varies from trial to trial. On some trials both options are worth equally large or small rewards. On other trials one option's reward is greater than that of the other. In Chapter 2, we demonstrate that in the unequal reward conditions subjects' choices are consistently biased towards the greater magnitude option. Additionally, we will show that this bias is independent of the motion stimulus strength and its magnitude is nearly optimal. In Chapter 3, we observe that single neurons in cortical area LIP consistently, simultaneously and dynamically represent both sensory and value information. We will argue that this representation supports an integrator model of decision making, in which sensory information is accumulated until the decision is resolved by a threshold crossing. Our results support an interpretation of this model in which value information adjusts the likelihood of a threshold crossing by iv raising or lowering the accumulator's initial state. In Chapter 4, we present a preliminary comparison between LIP and DLPFC activity, under identical conditions, suggesting they play fundamentally different roles in decision making. In Chapter 5, we discuss future lines of research.