The primary visual cortex (V1) is the ¢rst cortical area to receive visual input, and inferior temporal (IT) areas are among the last along the ventral visual pathway. We recorded, in area V1 of anaesthetized cats and area ITof awake macaque monkeys, responses of neurons to videos of natural scenes. Responses were analysed to test various hypotheses concerning the nature of neural coding in these two regions. A variety of spike-train statistics were measured including spike-count distributions, interspike interval distributions, coe¤cients of variation, power spectra, Fano factors and di¡erent sparseness measures. All statistics showed non-Poisson characteristics and several revealed self-similarity of the spike trains. Spike-count distributions were approximately exponential in both visual areas for eight di¡erent videos and for counting windows ranging from 50 ms to 5 seconds. The results suggest that the neurons maximize their information carrying capacity while maintaining a ¢xed long-term-average ¢ring rate, or equivalently, minimize their average ¢ring rate for a ¢xed information carrying capacity. I N T RO DUC T IONIt has been suggested that visual representations are optimized to transmit the maximum information about the images encountered in everyday life (Uttley 1973;Linsker 1987;Barlow 1989). This simple assumption has proven su¤cient to account for the characteristics of large monopolar cells in the £y (Srinivasan et al. 1982;Van Hateren 1992;Laughlin 1981), the temporal characteristics of retinal ganglion cells (Dong & Atick 1995), human spatial frequency thresholds (Atick & Redlich 1992;Van Hateren 1993), and the psychophysics of orientation perception for short presentation times (Baddeley & Hancock 1991).Maximization of information is a powerful theoretical principle that leads to testable predictions about the ¢ring patterns of neurons. However, to generate speci¢c predictions we must make some assumptions about the nature of the neural code and the type of constraint that limits its information carrying capacity. To apply information maximization to neuronal spike trains, we must identify which of their characteristics carry information. In our analysis, we will consider two possibilities: that ¢ring rates, or more precisely, spike counts over discrete intervals of time, are the information carrying elements; or that interspike intervals play this role. Without any constraints on the rate or precision of neuronal spiking, the information carrying capacity of a spike train is in¢nite. Thus, constraints play a crucial role in any information maximization procedure. We will consider three possibilities, constraints on the maximum ¢ring rate, the average ¢ring rate, or a quantity known as the sparseness of the ¢ring-rate distribution. Identifying the nature of the constraint that limits information carrying capacity has important implications for the biophysical mechanisms that underlie neural coding.Assuming the ¢ring rates carry information, Laughlin (1981) proposed a constraint on the maximum ¢...

It has been suggested that the primate perirhinal cortex contributes exclusively to memory. However, recent studies in macaque monkeys have implied that the perirhinal cortex may also contribute to object perception. To investigate whether the perirhinal cortex does contribute to perception, we devised several perceptual oddity tasks in which monkeys had to choose which stimulus of several presented concurrently on a touch screen was different. Macaques with bilateral perirhinal cortex ablations were selectively impaired relative to controls at perceptually discriminating the odd stimulus when the odd stimulus was a different object and when the discrimination could not be done on the basis of simple differences in features between the stimuli. They remained unimpaired relative to controls on discriminating the odd stimulus when the odd stimulus was a different color, a different shape, or a different size even when these discriminations were extremely difficult. They were also impaired on human and monkey face oddity tasks and oddity tasks with scenes containing objects. Therefore, we reject the notion that the macaque perirhinal cortex has a role exclusive to memory and conclude that the macaque perirhinal cortex does contribute to perception. We argue that the perirhinal cortex is neither specialized for perception nor memory processes alone, but rather, is specialized for processing stimuli that require processing at a more abstract level such as at the level of an object and that the perirhinal cortex contributes to both memory and perception of such stimuli.

A view-invariant representation of objects in the brain would have many computational advantages. Here we describe a population of single neurons in the temporal visual cortex (IT) that have view-invariant representations of familiar objects. Ten real plastic objects were placed in the monkeys' home cages for a period of time before neurophysiological experiments in which neuronal responses were measured to four views of each object. The macaques performed a visual fixation task, and had never been trained in object discrimination. The majority of the visual neurons recorded were responsive to some views of some objects and/or to the control stimuli, as would be expected from previous studies. However, a small subset of these neurons were responsive to all views of one or more of the objects, providing evidence that these neurons were coding for objects, rather than simply for individual views or visual features within the image. This result was confirmed by information theoretic analyses, which showed that the neurons provided information about which object was being seen, independently of the view. The coding scheme was shown to be sparse distributed, with relatively independent information being provided by the different neurons. Hypotheses about how these view-invariant cells are formed are described.

A degraded image of an object or face, which appears meaningless when seen for the first time, is easily recognizable after viewing an undegraded version of the same image. The neural mechanisms by which this form of rapid perceptual learning facilitates perception are not well understood. Psychological theory suggests the involvement of systems for processing stimulus attributes, spatial attention and feature binding, as well as those involved in visual imagery. Here we investigate where and how this rapid perceptual learning is expressed in the human brain by using functional neuroimaging to measure brain activity during exposure to degraded images before and after exposure to the corresponding undegraded versions. Perceptual learning of faces or objects enhanced the activity of inferior temporal regions known to be involved in face and object recognition respectively. In addition, both face and object learning led to increased activity in medial and lateral parietal regions that have been implicated in attention and visual imagery. We observed a strong coupling between the temporal face area and the medial parietal cortex when, and only when, faces were perceived. This suggests that perceptual learning involves direct interactions between areas involved in face recognition and those involved in spatial attention, feature binding and memory recall.

The distribution of responses of sensory neurons to ecological stimulation has been proposed to be designed to maximize information transmission, which according to a simple model would imply an exponential distribution of spike counts in a given time window. We have used recordings from inferior temporal cortex neurons responding to quasi-natural visual stimulation (presented using a video of everyday lab scenes and a large number of static images of faces and natural scenes) to assess the validity of this exponential model and to develop an alternative simple model of spike count distributions. We find that the exponential model has to be rejected in 84% of cases (at the p < 0.01 level). A new model, which accounts for the firing rate distribution found in terms of slow and fast variability in the inputs that produce neuronal activation, is rejected statistically in only 16% of cases. Finally, we show that the neurons are moderately efficient at transmitting information but not optimally efficient.