Nicole C. Rust

Do We Know What the Early Visual System Does

We can claim that we know what the visual system does once we can predict neural responses to arbitrary stimuli, including those seen in nature. In the early visual system, models based on one or more linear receptive fields hold promise to achieve this goal as long as the models include nonlinear mechanisms that control responsiveness, based on stimulus context and history, and take into account the nonlinearity of spike generation. These linear and nonlinear mechanisms might be the only essential determinants of the response, or alternatively, there may be additional fundamental determinants yet to be identified. Research is progressing with the goals of defining a single “standard model” for each stage of the visual pathway and testing the predictive power of these models on the responses to movies of natural scenes. These predictive models represent, at a given stage of the visual pathway, a compact description of visual computation. They would be an invaluable guide for understanding the underlying biophysical and anatomical mechanisms and relating neural responses to visual perception.

How MT cells analyze the motion of visual patterns

Neurons in area MT (V5) are selective for the direction of visual motion. In addition, many are selective for the motion of complex patterns independent of the orientation of their components, a behavior not seen in earlier visual areas. We show that the responses of MT cells can be captured by a linear-nonlinear model that operates not on the visual stimulus, but on the afferent responses of a population of nonlinear V1 cells. We fit this cascade model to responses of individual MT neurons and show that it robustly predicts the separately measured responses to gratings and plaids. The model captures the full range of pattern motion selectivity found in MT. Cells that signal pattern motion are distinguished by having convergent excitatory input from V1 cells with a wide range of preferred directions, strong motion opponent suppression and a tuned normalization that may reflect suppressive input from the surround of V1 cells.

Spatiotemporal Elements of Macaque V1 Receptive Fields

Neurons in primary visual cortex (V1) are commonly classified as simple or complex based upon their sensitivity to the sign of stimulus contrast. The responses of both cell types can be described by a general model in which the outputs of a set of linear filters are nonlinearly combined. We estimated the model for a population of V1 neurons by analyzing the mean and covariance of the spatiotemporal distribution of random bar stimuli that were associated with spikes. This analysis reveals an unsuspected richness of neuronal computation within V1. Specifically, simple and complex cell responses are best described using more linear filters than the one or two found in standard models. Many filters revealed by the model contribute suppressive signals that appear to have a predominantly divisive influence on neuronal firing. Suppressive signals are especially potent in direction-selective cells, where they reduce responses to stimuli moving in the nonpreferred direction.

Spike-triggered neural characterization

Response properties of sensory neurons are commonly described using receptive fields. This description may be formalized in a model that operates with a small set of linear filters whose outputs are nonlinearly combined to determine the instantaneous firing rate. Spike-triggered average and covariance analyses can be used to estimate the filters and nonlinear combination rule from extracellular experimental data. We describe this methodology, demonstrating it with simulated model neuron examples that emphasize practical issues that arise in experimental situations.

Analyzing Neural Responses to Natural Signals: Maximally Informative Dimensions

We propose a method that allows for a rigorous statistical analysis of neural responses to natural stimuli that are nongaussian and exhibit strong correlations. We have in mind a model in which neurons are selective for a small number of stimulus dimensions out of a high-dimensional stimulus space, but within this subspace the responses can be arbitrarily nonlinear. Existing analysis methods are based on correlation functions between stimuli and responses, but these methods are guaranteed to work only in the case of gaussian stimulus ensembles. As an alternative to correlation functions, we maximize the mutual information between the neural responses and projections of the stimulus onto low-dimensional subspaces. The procedure can be done iteratively by increasing the dimensionality of this subspace. Those dimensions that allow the recovery of all of the information between spikes and the full unprojected stimuli describe the relevant subspace. If the dimensionality of the relevant subspace indeed is small, it becomes feasible to map the neurons input-output function even under fully natural stimulus conditions. These ideas are illustrated in simulations on model visual and auditory neurons responding to natural scenes and sounds, respectively.

Selectivity and Tolerance (“Invariance”) Both Increase as Visual Information Propagates from Cortical Area V4 to IT

Our ability to recognize objects despite large changes in position, size, and context is achieved through computations that are thought to increase both the shape selectivity and the tolerance (“invariance”) of the visual representation at successive stages of the ventral pathway [visual cortical areas V1, V2, and V4 and inferior temporal cortex (IT)]. However, these ideas have proven difficult to test. Here, we consider how well population activity patterns at two stages of the ventral stream (V4 and IT) discriminate between, and generalize across, different images. We found that both V4 and IT encode natural images with similar fidelity, whereas the IT population is much more sensitive to controlled, statistical scrambling of those images. Scrambling sensitivity was proportional to receptive field (RF) size in both V4 and IT, suggesting that, on average, the number of visual feature conjunctions implemented by a V4 or IT neuron is directly related to its RF size. We also found that the IT population could better discriminate between objects across changes in position, scale, and context, thus directly demonstrating a V4-to-IT gain in tolerance. This tolerance gain could be accounted for by both a decrease in single-unit sensitivity to identity-preserving transformations (e.g., an increase in RF size) and an increase in the maintenance of rank-order object selectivity within the RF. These results demonstrate that, as visual information travels from V4 to IT, the population representation is reformatted to become more selective for feature conjunctions and more tolerant to identity preserving transformations, and they reveal the single-unit response properties that underlie that reformatting.

In praise of artifice

The visual system evolved to process natural images, and the goal of visual neuroscience is to understand the computations it uses to do this. Indeed the goal of any theory of visual function is a model that will predict responses to any stimulus, including natural scenes. It has, however, recently become common to take this fundamental principle one step further: trying to use photographic or cinematographic representations of natural scenes (natural stimuli) as primary probes to explore visual computations. This approach is both challenging and controversial, and we argue that this use of natural images is so fraught with difficulty that it is not useful. Traditional methods for exploring visual computations that use artificial stimuli with carefully selected properties have been and continue to be the most effective tools for visual neuroscience. The proper use of natural stimuli is to test models based on responses to these synthetic stimuli, not to replace them.

Signals in inferotemporal and perirhinal cortex suggest an “untangling” of visual target information

Finding sought visual targets requires our brains to flexibly combine working memory information about what we are looking for with visual information about what we are looking at. To investigate the neural computations involved in finding visual targets, we recorded neural responses in inferotemporal cortex (IT) and perirhinal cortex (PRH) as macaque monkeys performed a task that required them to find targets in sequences of distractors. We found similar amounts of total task-specific information in both areas; however, information about whether a target was in view was more accessible using a linear read-out or, equivalently, was more untangled in PRH. Consistent with the flow of information from IT to PRH, we also found that task-relevant information arrived earlier in IT. PRH responses were well-described by a functional model in which computations in PRH untangle input from IT by combining neurons with asymmetric tuning correlations for target matches and distractors.

Spike-triggered characterization of excitatory and suppressive stimulus dimensions in monkey V1

Neurons in primary visual cortex are commonly characterized using linear models, or simple extensions of linear models. Specifically, V1 simple cell responses are often characterized using a rectified linear receptive field, and complex cell responses are often described as the sum of squared responses of two linear subunits. We examined this class of model directly by applying spike-triggered covariance analysis to responses of monkey V1 neurons under binary white noise stimulation. The analysis extracts a low-dimensional subspace of the full stimulus space that is primarily responsible for generation of the neural response, including both excitatory and suppressive components. We found no fewer than two excitatory dimensions in simple cells, and as many as seven dimensions in complex cells. For all cells, we also found suppressive dimensions that were at least equal in number to the excitatory dimensions. These results suggest that extensions to standard models are required to fully describe the response properties of cells in V1.

Dissociation of Neuronal and Psychophysical Responses to Local and Global Motion

Most neurons in cortical area MT (V5) are strongly direction selective, and their activity is closely associated with the perception of visual motion. These neurons have large receptive fields built by combining inputs with smaller receptive fields that respond to local motion. Humans integrate motion over large areas and can perceive what has been referred to as global motion. The large size and direction selectivity of MT receptive fields suggests that MT neurons may represent global motion. We have explored this possibility by measuring responses to a stimulus in which the directions of simultaneously presented local and global motion are independently controlled. Surprisingly, MT responses depended only on the local motion and were unaffected by the global motion. Yet, under similar conditions, human observers perceive global motion and are impaired in discriminating local motion. Although local motion perception might depend on MT signals, global motion perception depends on mechanisms qualitatively different from those in MT. Motion perception therefore does not depend on a single cortical area but reflects the action and interaction of multiple brain systems.