Social groups and polarization of aesthetic values from symmetry and complexity.

When deciding what images we prefer, our brain must weigh many aesthetic variables, such as symmetry and complexity. To date, aesthetic research has mainly focused on investigating one variable at a time. In this article, we use symmetry and complexity to study the problem of multi aesthetic-variable interactions. For symmetry and complexity, there are two simple interaction hypotheses. The independence hypothesis proposes that the evaluation of aesthetic variables is mutually independent. Meanwhile, Birkhoff's aesthetic-measure hypothesis predicts that people prefer images high in symmetry and low in complexity, and dislike the opposite. To test these hypotheses, we generated images that systematically varied in levels of symmetry and complexity. We then compared the subjects' preference maps to identify regions of likes and dislikes. Unlike the predictions from these hypotheses, we found that most, but not all subjects, formed two distinct natural clusters, termed "islands," in terms of likes and dislikes. We also found that people with more art exposure were less likely to belong to an island. If someone did belong to an island, their gender influenced which cluster they belonged to. We discuss alternate hypotheses, possible mechanisms for the occurrence of islands, and their possible social implications.

Systematic errors in the perception of rhythm.

One hypothesis for why humans enjoy musical rhythms relates to their prediction of when each beat should occur. The ability to predict the timing of an event is important from an evolutionary perspective. Therefore, our brains have evolved internal mechanisms for processing the progression of time. However, due to inherent noise in neural signals, this prediction is not always accurate. Theoretical considerations of optimal estimates suggest the occurrence of certain systematic errors made by the brain when estimating the timing of beats in rhythms. Here, we tested psychophysically whether these systematic errors exist and if so, how they depend on stimulus parameters. Our experimental data revealed two main types of systematic errors. First, observers perceived the time of the last beat of a rhythmic pattern as happening earlier than actual when the inter-beat interval was short. Second, the perceived time of the last beat was later than the actual when the inter-beat interval was long. The magnitude of these systematic errors fell as the number of beats increased. However, with many beats, the errors due to long inter-beat intervals became more apparent. We propose a Bayesian model for these systematic errors. The model fits these data well, allowing us to offer possible explanations for how these errors occurred. For instance, neural processes possibly contributing to the errors include noisy and temporally asymmetric impulse responses, priors preferring certain time intervals, and better-early-than-late loss functions. We finish this article with brief discussions of both the implications of systematic errors for the appreciation of rhythm and the possible compensation by the brain's motor system during a musical performance.

Does Amount of Information Support Aesthetic Values?

Obtaining information from the world is important for survival. The brain, therefore, has special mechanisms to extract as much information as possible from sensory stimuli. Hence, given its importance, the amount of available information may underlie aesthetic values. Such information-based aesthetic values would be significant because they would compete with others to drive decision-making. In this article, we ask, "What is the evidence that amount of information support aesthetic values?" An important concept in the measurement of informational volume is entropy. Research on aesthetic values has thus used Shannon entropy to evaluate the contribution of quantity of information. We review here the concepts of information and aesthetic values, and research on the visual and auditory systems to probe whether the brain uses entropy or other relevant measures, specially, Fisher information, in aesthetic decisions. We conclude that information measures contribute to these decisions in two ways: first, the absolute quantity of information can modulate aesthetic preferences for certain sensory patterns. However, the preference for volume of information is highly individualized, with information-measures competing with organizing principles, such as rhythm and symmetry. In addition, people tend to be resistant to too much entropy, but not necessarily, high amounts of Fisher information. We show that this resistance may stem in part from the distribution of amount of information in natural sensory stimuli. Second, the measurement of entropic-like quantities over time reveal that they can modulate aesthetic decisions by varying degrees of surprise given temporally integrated expectations. We propose that amount of information underpins complex aesthetic values, possibly informing the brain on the allocation of resources or the situational appropriateness of some cognitive models.

Stochasticity, Nonlinear Value Functions, and Update Rules in Learning Aesthetic Biases.

A theoretical framework for the reinforcement learning of aesthetic biases was recently proposed based on brain circuitries revealed by neuroimaging. A model grounded on that framework accounted for interesting features of human aesthetic biases. These features included individuality, cultural predispositions, stochastic dynamics of learning and aesthetic biases, and the peak-shift effect. However, despite the success in explaining these features, a potential weakness was the linearity of the value function used to predict reward. This linearity meant that the learning process employed a value function that assumed a linear relationship between reward and sensory stimuli. Linearity is common in reinforcement learning in neuroscience. However, linearity can be problematic because neural mechanisms and the dependence of reward on sensory stimuli were typically nonlinear. Here, we analyze the learning performance with models including optimal nonlinear value functions. We also compare updating the free parameters of the value functions with the delta rule, which neuroscience models use frequently, vs. updating with a new Phi rule that considers the structure of the nonlinearities. Our computer simulations showed that optimal nonlinear value functions resulted in improvements of learning errors when the reward models were nonlinear. Similarly, the new Phi rule led to improvements in these errors. These improvements were accompanied by the straightening of the trajectories of the vector of free parameters in its phase space. This straightening meant that the process became more efficient in learning the prediction of reward. Surprisingly, however, this improved efficiency had a complex relationship with the rate of learning. Finally, the stochasticity arising from the probabilistic sampling of sensory stimuli, rewards, and motivations helped the learning process narrow the range of free parameters to nearly optimal outcomes. Therefore, we suggest that value functions and update rules optimized for social and ecological constraints are ideal for learning aesthetic biases.

Evolution of Neuroaesthetic Variables in Portrait Paintings throughout the Renaissance.

To compose art, artists rely on a set of sensory evaluations performed fluently by the brain. The outcome of these evaluations, which we call neuroaesthetic variables, helps to compose art with high aesthetic value. In this study, we probed whether these variables varied across art periods despite relatively unvaried neural function. We measured several neuroaesthetic variables in portrait paintings from the Early and High Renaissance, and from Mannerism. The variables included symmetry, balance, and contrast (chiaroscuro), as well as intensity and spatial complexities measured by two forms of normalized entropy. The results showed that the degree of symmetry remained relatively constant during the Renaissance. However, the balance of portraits decayed abruptly at the end of the Early Renaissance, that is, at the closing of the 15th century. Intensity and spatial complexities, and thus entropies, of portraits also fell in such manner around the same time. Our data also showed that the decline of complexity and entropy could be attributed to the rise of chiaroscuro. With few exceptions, the values of aesthetic variables from the top of artists of the Renaissance resembled those of their peers. We conclude that neuroaesthetic variables have flexibility to change in brains of artists (and observers).

A Theoretical Framework for How We Learn Aesthetic Values.

How do we come to like the things that we do? Each one of us starts from a relatively similar state at birth, yet we end up with vastly different sets of aesthetic preferences. These preferences go on to define us both as individuals and as members of our cultures. Therefore, it is important to understand how aesthetic preferences form over our lifetimes. This poses a challenging problem: to understand this process, one must account for the many factors at play in the formation of aesthetic values and how these factors influence each other over time. A general framework based on basic neuroscientific principles that can also account for this process is needed. Here, we present such a framework and illustrate it through a model that accounts for the trajectories of aesthetic values over time. Our framework is inspired by meta-analytic data of neuroimaging studies of aesthetic appraisal. This framework incorporates effects of sensory inputs, rewards, and motivational states. Crucially, each one of these effects is probabilistic. We model their interactions under a reinforcement-learning circuitry. Simulations of this model and mathematical analysis of the framework lead to three main findings. First, different people may develop distinct weighing of aesthetic variables because of individual variability in motivation. Second, individuals from different cultures and environments may develop different aesthetic values because of unique sensory inputs and social rewards. Third, because learning is stochastic, stemming from probabilistic sensory inputs, motivations, and rewards, aesthetic values vary in time. These three theoretical findings account for different lines of empirical research. Through our study, we hope to provide a general and unifying framework for understanding the various aspects involved in the formation of aesthetic values over time.

A Statistical Analysis of the Relationship between Harmonic Surprise and Preference in Popular Music.

Studies have shown that some musical pieces may preferentially activate reward centers in the brain. Less is known, however, about the structural aspects of music that are associated with this activation. Based on the music cognition literature, we propose two hypotheses for why some musical pieces are preferred over others. The first, the Absolute-Surprise Hypothesis, states that unexpected events in music directly lead to pleasure. The second, the Contrastive-Surprise Hypothesis, proposes that the juxtaposition of unexpected events and subsequent expected events leads to an overall rewarding response. We tested these hypotheses within the framework of information theory, using the measure of "surprise." This information-theoretic variable mathematically describes how improbable an event is given a known distribution. We performed a statistical investigation of surprise in the harmonic structure of songs within a representative corpus of Western popular music, namely, the McGill Billboard Project corpus. We found that chords of songs in the top quartile of the Billboard chart showed greater average surprise than those in the bottom quartile. We also found that the different sections within top-quartile songs varied more in their average surprise than the sections within bottom-quartile songs. The results of this study are consistent with both the Absolute- and Contrastive-Surprise Hypotheses. Although these hypotheses seem contradictory to one another, we cannot yet discard the possibility that both absolute and contrastive types of surprise play roles in the enjoyment of popular music. We call this possibility the Hybrid-Surprise Hypothesis. The results of this statistical investigation have implications for both music cognition and the human neural mechanisms of esthetic judgments.

Cell type-specific changes in retinal ganglion cell function induced by rod death and cone reorganization in rats.

We have determined the impact of rod death and cone reorganization on the spatiotemporal receptive fields (RFs) and spontaneous activity of distinct retinal ganglion cell (RGC) types. We compared RGC function between healthy and retinitis pigmentosa (RP) model rats (S334ter-3) at a time when nearly all rods were lost but cones remained. This allowed us to determine the impact of rod death on cone-mediated visual signaling, a relevant time point because the diagnosis of RP frequently occurs when patients are nightblind but daytime vision persists. Following rod death, functionally distinct RGC types persisted; this indicates that parallel processing of visual input remained largely intact. However, some properties of cone-mediated responses were altered ubiquitously across RGC types, such as prolonged temporal integration and reduced spatial RF area. Other properties changed in a cell type-specific manner, such as temporal RF shape (dynamics), spontaneous activity, and direction selectivity. These observations identify the extent of functional remodeling in the retina following rod death but before cone loss. They also indicate new potential challenges to restoring normal vision by replacing lost rod photoreceptors.NEW & NOTEWORTHY This study provides novel and therapeutically relevant insights to retinal function following rod death but before cone death. To determine changes in retinal output, we used a large-scale multielectrode array to simultaneously record from hundreds of retinal ganglion cells (RGCs). These recordings of large-scale neural activity revealed that following the death of all rods, functionally distinct RGCs remain. However, the receptive field properties and spontaneous activity of these RGCs are altered in a cell type-specific manner.

Inferring Master Painters' Esthetic Biases from the Statistics of Portraits.

The Processing Fluency Theory posits that the ease of sensory information processing in the brain facilitates esthetic pleasure. Accordingly, the theory would predict that master painters should display biases toward visual properties such as symmetry, balance, and moderate complexity. Have these biases been occurring and if so, have painters been optimizing these properties (fluency variables)? Here, we address these questions with statistics of portrait paintings from the Early Renaissance period. To do this, we first developed different computational measures for each of the aforementioned fluency variables. Then, we measured their statistics in 153 portraits from 26 master painters, in 27 photographs of people in three controlled poses, and in 38 quickly snapped photographs of individual persons. A statistical comparison between Early Renaissance portraits and quickly snapped photographs revealed that painters showed a bias toward balance, symmetry, and moderate complexity. However, a comparison between portraits and controlled-pose photographs showed that painters did not optimize each of these properties. Instead, different painters presented biases toward different, narrow ranges of fluency variables. Further analysis suggested that the painters' individuality stemmed in part from having to resolve the tension between complexity vs. symmetry and balance. We additionally found that constraints on the use of different painting materials by distinct painters modulated these fluency variables systematically. In conclusion, the Processing Fluency Theory of Esthetic Pleasure would need expansion if we were to apply it to the history of visual art since it cannot explain the lack of optimization of each fluency variables. To expand the theory, we propose the existence of a Neuroesthetic Space, which encompasses the possible values that each of the fluency variables can reach in any given art period. We discuss the neural mechanisms of this Space and propose that it has a distributed representation in the human brain. We further propose that different artists reside in different, small sub-regions of the Space. This Neuroesthetic-Space hypothesis raises the question of how painters and their paintings evolve across art periods.

Reshaping the Cone-Mosaic in a Rat Model of Retinitis Pigmentosa: Modulatory Role of ZO-1 Expression in DL-Alpha-Aminoadipic Acid Reshaping.

In S334ter-line-3 rat model of Retinitis Pigmentosa (RP), rod cell death induces the rearrangement of cones into mosaics of rings while the fibrotic processes of Müller cells remodel to fill the center of the rings. In contrast, previous work established that DL-alpha-aminoadipic-acid (AAA), a compound that transiently blocks Müller cell metabolism, abolishes these highly structured cone rings. Simultaneously, adherens-junction associated protein, Zonula occludens-1 (ZO-1) expression forms in a network between the photoreceptor segments and Müller cells processes. Thus, we hypothesized that AAA treatment alters the cone mosaic rings by disrupting the distal sealing formed by these fibrotic processes, either directly or indirectly, by down regulating the expression of ZO-1. Therefore, we examined these processes and ZO-1 expression at the outer retina after intravitreal injection of AAA and observed that AAA treatment transiently disrupts the distal glial sealing in RP retina, plus induces cones in rings to become more homogeneous. Moreover, ZO-1 expression is actively suppressed after 3 days of AAA treatment, which coincided with cone ring disruption. Similar modifications of glial sealing and cone distribution were observed after injection of siRNA to inhibit ZO-1 expression. These findings support our hypothesis and provide additional information about the critical role played by ZO-1 in glial sealing and shaping the ring mosaic in RP retina. These studies represent important advancements in the understanding of retinal degeneration's etiology and pathophysiology.

TIMP-1 affects the spatial distribution of dendritic processes of second-order neurons in a rat model of Retinitis Pigmentosa.

Retinitis Pigmentosa (RP) is an inherited disorder that may lead to blindness. In the rhodopsin S334ter-line-3 rat model of RP, the death of rods induces spatial rearrangement of cones into regular ring mosaics. Using this model, we discovered that the ring mosaics are restored to a homogeneous distribution upon application of tissue inhibitor of metalloproteinase-1 (TIMP-1). In this study, we further investigated the cone migration and spatial distribution of second-order neurons and their connections to cones in the presence or absence of TIMP-1 using immunohistochemistry to identify retinal neurons and their connections with cones. M-opsin cell bodies and their outer segments were evaluated to determine whether TIMP-1 delays the degeneration of outer segments of cones. We observed that during cone rearrangement into ring mosaics in RP retina, dendritic processes of second-order neurons undergo remodeling to maintain their synaptic connections with the cones in the rings. TIMP-1 treatment induced the cones to rearrange and dendritic processes of second-order neurons to return to a more homogeneous spatial distribution. In addition, TIMP-1 treatment protected the outer segments of cones at later stages of retinal degeneration. Our findings clearly demonstrate that despite their dramatic spatial rearrangement, cones and second-order neuron processes maintain their synaptic connections before and after TIMP-1 treatment.

The effect of TIMP-1 on the cone mosaic in the retina of the rat model of retinitis pigmentosa.

PURPOSE: The array of photoreceptors found in normal retinas provides uniform and regular sampling of the visual space. In contrast, cones in retinas of the S334ter-line-3 rat model for RP migrate to form a mosaic of rings, leaving large holes with few or no photoreceptors. Similar mosaics appear in human patients with other forms of retinal dystrophy. In the current study, we aimed to investigate the effect of tissue inhibitor of metalloproteinase-1 (TIMP-1) on the mosaic of cones in S334ter-line-3 rat retinas. We focused on TIMP-1 because it is one of the regulators of the extracellular matrix important for cellular migration. METHODS: Immunohistochemistry was performed to reveal M-opsin cone cells (M-cone) and the results were quantified to test statistically whether or not TIMP-1 restores the mosaics to normal. In particular, the tests focused on the Voronoi and nearest-neighbor distance analyses. RESULTS: Our tests indicated that TIMP-1 led to significant disruption of the M-opsin cone rings in S334ter-line-3 rat retinas and resulted in almost complete homogeneous mosaics. In addition, TIMP-1 induced the M-cone spatial distribution to become closer to random with decreased regularity in S334ter-line-3 rat retinas. CONCLUSIONS: These findings confirm that TIMP-1 induced M-cone mosaics in S334ter-line-3 to gain homogeneity without reaching the degree of regularity seen in normal retinal mosaics. Even if TIMP-1 fails to promote regularity, the effects of this drug on homogeneity appear to be so dramatic that TIMP-1 may be a potential therapeutic agent. TIMP-1 improves sampling of the visual field simply by causing homogeneity.

Spatiotemporal pattern of rod degeneration in the S334ter-line-3 rat model of retinitis pigmentosa.

We have recently described the surviving cones and Müller-glia process remodeling in retinitis pigmentosa (RP) and shown that rod degeneration triggers the reorganization of the cone mosaic into an orderly array of rings. Within these rings, remodeled Müller-glia processes envelope cones. Here, we report the spatiotemporal pattern of healthy rods, their relationship with dying rods and the way that rod death stimulates the modification of cone spatial-distribution patterns and Müller-glia processes in the S334ter-line-3 rat, a transgenic model expressing a rhodopsin mutation that causes RP. The spatial patterns of rods, cones, microglial and Müller cells were labeled by immunocytochemistry with cell-type-specific markers at various stages of deveopment in rat whole-mount retinas. Spatial patterns of dying cells were examined by TUNEL staining. The S334ter rod mosaic began to develop small holes around postnatal day 10. These hot-spots of cell death progressively increased in size, leaving larger rod-less holes behind. The holes were temporarily occupied by active microglial cells, before being replaced by remodeled Müller-cell processes. Our data suggest that the hot spots of rod death create holes in the rod mosaic early in retinal degeneration and that the resulting pattern triggers the modification of the spatial-distribution patterns of cones and glia cells.

Complex biomedical systems: from basic science to translation.

The Department of Biomedical Engineering (BME) of the University of Southern California (BME@USC) has a longstanding tradition of advancing biomedicine through the development and application of novel engineering ideas. More than 80 primary and affiliated faculty members conduct cutting-edge research in a wide variety of areas, such as neuroengineering, biosystems and biosignal analysis, medical devices (including biomicroelectromechanical systems (bioMEMS) and bionanotechnology), biomechanics, bioimaging, and imaging informatics. Currently, the department hosts six internationally recognized research centers: the Biomimetic MicroElectronic Systems Engineering Research Center (funded by the National Science Foundation), the Biomedical Simulations Resource [funded by the National Institutes of Health (NIH)], the Medical Ultrasonic Transducer Center (funded by NIH), the Center for Neural Engineering, the Center for Vision Science and Technology (funded by an NIH Bioengineering Research Partnership Grant), and the Center for Genomic and Phenomic Studies in Autism (funded by NIH). BME@USC ranks in the top tier of all U.S. BME departments in terms of research funding per faculty.

Rearrangement of the cone mosaic in the retina of the rat model of retinitis pigmentosa.

In retinitis pigmentosa (RP), the death of cones normally follows some time after the degeneration of rods. Recently, surviving cones in RP have been studied and reported in detail. These cones undergo extensive remodeling in their morphology. Here we report an extension of the remodeling study to consider possible modifications of spatial-distribution patterns. For this purpose we used S334ter-line-3 transgenic rats, a transgenic model developed to express a rhodopsin mutation causing RP. In this study, retinas were collected at postnatal (P) days P5-30, 90, 180, and P600. We then immunostained the retinas to examine the morphology and distribution of cones and to quantify the total cone numbers. Our results indicate that cones undergo extensive changes in their spatial distribution to give rise to a mosaic comprising an orderly array of rings. These rings first begin to appear at P15 at random regions of the retina and become ubiquitous throughout the entire tissue by P90. Such distribution pattern loses its clarity by P180 and mostly disappears at P600, at which time the cones are almost all dead. In contrast, the numbers of cones in RP and normal conditions do not show significant differences at stages as late as P180. Therefore, rings do not form by cell death at their centers, but by cone migration. We discuss its possible mechanisms and suggest a role for hot spots of rod death and the remodeling of Müller cell process into zones of low density of photoreceptors.

Aquaporin-4 immunoreactivity in Müller and amacrine cells of marine teleost fish retina.

Aquaporins (AQPs) are membrane proteins that facilitate water transport across biological membranes and are essential for the proper function of neural tissue. Although AQPs have been extensively studied in mammalian retina, their presence in lower vertebrate retina is less frequently characterized. AQP4 expressed in mammalian and chick Müller cells plays a major part in maintaining retinal homeostasis. In this study, we examined the immunoreactivity of AQP4 in the adult retina of gilthead sea bream (Sparus aurata-teleost fish), during light and dark adaptation. The AQP4 expression was detected in Müller cell somas at the inner nuclear layer and in the end-feet processes near the vitreoretinal border. Moreover, AQP4 was also evident in cone photoreceptor cells and in a GABAergic subpopulation of amacrine cells (AQP4-ACs). Four different types of AQP4-ACs were characterized based on their morphology and dendrite stratification. Interestingly, a stronger AQP4 immunoreactivity was observed in the inner nuclear layer during dark adaptation, accompanied by a significant increment in AQP4-ACs cell size. Hence, AQP4 may play an important role in water distribution in the teleost fish retina.

Dependence of the retinal Ganglion cell's responses on local textures of natural scenes.

An important property of natural images contributing to a retinal ganglion cell's (RGC) responses is the temporal modulation of mean intensity (contrast) in the receptive field (RF) center. However, these responses exhibit a significant amount of variability. This variability could arise in part from responses to the spatial intensity variation of the natural images in the RF center, i.e., their local intensity distribution or their local visual texture. We tested five predictions derived from this hypothesis: First, responses tend to increase with the variance of the local visual texture of natural images. Second, the skewed distribution of intensities in natural images leads to asymmetric responses to their onset and offset to and from a gray background of the same mean intensity. Third, repeating this experiment with the negative of natural images inverts the asymmetry. Fourth, performing an intensity histogram equalization of the images eliminates the asymmetry. Fifth, RGCs' responses increase with the spatial contrast of artificial plaids. The hypothesis passed all five tests, which indicate that responses to natural images increase with the variance of their visual texture. To account for this texture sensitivity, we propose a model in which the RFs of most RGCs of the rabbit have multiple nonlinear subunits.

Role of Müller cells in cone mosaic rearrangement in a rat model of retinitis pigmentosa.

Retinitis pigmentosa (RP) is a type of inherited retinal degenerative disease, which leads to blindness. The primary pathological event of this disease is the death of rods because of genetic mutations. The S334ter-line-3 rat is a transgenic model developed to express a rhodopsin mutation similar to that found in RP. In this study, the rod's death triggered are organization of the cone mosaic into an orderly array of rings. Four observations were relevant to understand this reorganization. First, rods died in hot spots, which progressively increased as circular waves, leaving rod-less zones behind. Second, rings of cones formed around these zones. Third, remodeled Müller glia processes enveloped cones and filled the center of their rings. Zonula occludens-1 located between the photoreceptor inner segments and the apical processes of Müller cells showed in the rings. Fourth, these rings were formed before the onset of cone cell deaths and were maintained until late stages of RP. From these observations,we hypothesize that cone-Müller-cell interactions mediate and maintain the rings. A test of this hypothesis can be performed by injecting DL-a-aminoadipic acid (AAA), which is known to disrupt Müller cell metabolism. A single intravitreal injection of AAA at P50 disrupted the rings of cones 3 days after the injection. These findings indicate that the rearrangement of cones in rings is modulated by Müller cells in RP. Thus, if the relationship between photoreceptors and Müller glia is better understood, the latter could potentially be manipulated for effective neuroprotection or the restoration of normal cone arrays.

Separability of stimulus parameter encoding by on-off directionally selective rabbit retinal ganglion cells.

The ganglion cell output of the retina constitutes a bottleneck in sensory processing in that ganglion cells must encode multiple stimulus parameters in their responses. Here we investigate encoding strategies of On-Off directionally selective retinal ganglion cells (On-Off DS RGCs) in rabbits, a class of cells dedicated to representing motion. The exquisite axial discrimination of these cells to preferred vs. null direction motion is well documented: it is invariant with respect to speed, contrast, spatial configuration, spatial frequency, and motion extent. However, these cells have broad direction tuning curves and their responses also vary as a function of other parameters such as speed and contrast. In this study, we examined whether the variation in responses across multiple stimulus parameters is systematic, that is the same for all cells, and separable, such that the response to a stimulus is a product of the effects of each stimulus parameter alone. We extracellularly recorded single On-Off DS RGCs in a superfused eyecup preparation while stimulating them with moving bars. We found that spike count responses of these cells scaled as independent functions of direction, speed, and luminance. Moreover, the speed and luminance functions were common across the whole sample of cells. Based on these findings, we developed a model that accurately predicted responses of On-Off DS RGCs as products of separable functions of direction, speed, and luminance (r = 0.98; P < 0.0001). Such a multiplicatively separable encoding strategy may simplify the decoding of these cells' outputs by the higher visual centers.

ePPR: a new strategy for the characterization of sensory cells from input/output data.

A central goal of systems neuroscience is to characterize the transformation of sensory input to spiking output in single neurons. This problem is complicated by the large dimensionality of the inputs. To cope with this problem, previous methods have estimated simplified versions of a generic linear-nonlinear (LN) model and required, in most cases, stimuli with constrained statistics. Here we develop the extended Projection Pursuit Regression (ePPR) algorithm that allows the estimation of all of the parameters, in space and time, of a generic LN model using arbitrary stimuli. We first prove that ePPR models can uniformly approximate, to an arbitrary degree of precision, any continuous function. To test this generality empirically, we use ePPR to recover the parameters of models of cortical cells that cannot be represented exactly with an ePPR model. Next we evaluate ePPR with physiological data from primary visual cortex, and show that it can characterize both simple and complex cells, from their responses to both natural and random stimuli. For both simulated and physiological data, we show that ePPR compares favorably to spike-triggered and information-theoretic techniques. To the best of our knowledge, this article contains the first demonstration of a method that allows the estimation of an LN model of visual cells, containing multiple spatio-temporal filters, from their responses to natural stimuli.