this article.

frequency, natural image Humans can recognize the gist of a novel image in a single glance, independent of its complexity. How is this remarkable feat accomplished? Based on behavioral and computational evidence, this paper describes a formal approach to the representation and the mechanism of scene gist understanding, based on scene-centered, rather than objectcentered primitives. We show that the structure of a scene image can be estimated by the mean of global image features, providing a statistical summary of the spatial layout properties (Spatial Envelope representation) of the scene. Global features are based on configurations of spatial scales and are estimated without invoking segmentation or grouping operations. The scene-centered approach is not an alternative to local image analysis but would serve as a feed-forward and parallel pathway of visual processing, able to quickly constrain local feature analysis and enhance object recognition in cluttered natural scenes. 1

Many experiments have shown that the human visual system makes extensive use of contextual information for facilitating object search in natural scenes. However, the question of how to formally model contextual influences is still open. On the basis of a Bayesian framework, the authors present an original approach of attentional guidance by global scene context. The model comprises 2 parallel pathways; one pathway computes local features (saliency) and the other computes global (scenecentered) features. The contextual guidance model of attention combines bottom-up saliency, scene context, and top-down mechanisms at an early stage of visual processing and predicts the image regions likely to be fixated by human observers performing natural search tasks in real-world scenes.

Many researchers interested in connectionist models accept that such models are "neurally inspired " but do not worry too much about whether their models are biologically realistic. While such a position may be perfectly justifiable, the present paper attempts to illustrate how biological information can be used to constrain connectionist models. Two particular areas are discussed. The first section deals with visual information processing in the primate and human visual system. It is argued that speed with which visual information is processed imposes major constraints on the architecture and operation of the visual system. In particular, it seems that a great deal of processing must depend on a single bottum-up pass. The second section deals with biological aspects of learning algorithms. It is argued that although there is good evidence for certain coactivation related synaptic modification schemes, other learning mechanisms, including back-propagation, are not currently supported by experimental data.

When an observer detects a target in a rapid stream of visual stimuli, there is a brief period of time during which the detection of subsequent targets is impaired. In this study, event-related potentials (ERPs) were recorded from normal adult observers to determine whether this "attentional blink " reflects a suppression of perceptual processes or an impairment in postperceptual processes. No suppression was observed during the attentional blink interval for ERP components corresponding to sensory processing (the P1 and N1 components) or semantic analysis (the N400 component). However, complete suppression was observed for an ERP component that has been hypothesized to reflect the updating of working memory (the P3 component). Results indicate that the attentional blink reflects an impairment in a postperceptual stage of processing. Over the past several decades, the vast majority of studies of visual attention have examined the operation of attention across space. In the visual search task, for example, a target item must be detected within an array of distractor items that are presented at different locations from the target. In recent

Our intuition that we richly represent the visual details of our environment is illusory. When viewing a scene, we seem to use detailed representations of object properties and interobject relations to achieve a sense of continuity across views. Yet, several recent studies show that human observers fail to detect changes to objects and object properties when localized retinal information signaling a change is masked or eliminated (e.g., by eye movements). However, these studies changed arbitrarily chosen objects which may have been outside the focus of attention. We draw on previous research showing the importance of spatiotemporal information for tracking objects by creating short motion pictures in which objects in both arbitrary locations and the very center of attention were changed. Adult observers failed to notice changes in both cases, even when the sole actor in a scene transformed into another person across an instantaneous change in camera angle (or "cut").

Efficient categorizations of complex visual stimuli require effective encodings of their distinctive properties. However, the question remains of how processes of object and scene categorization use the information associated with different perceptual spatial scales. The psychophysics of scale perception suggests that recognition uses coarse blobs before fine scale edges, because the former is perceptually available before the latter. Although possible, this perceptually determined scenario neglects the nature of the task the recognition system must solve. If different spatial scales transmit different information about the input, an identical scene might be flexibly encoded and perceived at the scale that optimizes information for the considered task—i.e., the diagnostic scale. This paper tests the hypothesis that scale diagnosticity can determine scale selection for recognition. Experiment 1 tested whether coarse and fine spatial scales were both available at the onset of scene categorization. The second experiment tested that the selection of one scale could change depending on the diagnostic information present at this scale. The third and fourth experiments investigated whether scalespecific cues were independently processed, or whether they perceptually cooperated in the recognition of the input scene. Results suggest that a mandatory low-level registration of multiple spatial scales promotes flexible scene encodings, perceptions, and categorizations.

Many theories have been proposed to explain difficulty with center embedded constructions, most attributing the problem to some kind of limited capacity short-term memory. However, these theories have developed for the most part independently of more traditional memory research, which has focused on uncovering general principles such as chunking and interference. This article attempts to gain some unification with this research by suggesting that an interesting range of core sentence processing phenomena can be explained as interference effects in a sharply limited syntactic working memory. These include difficult and acceptable embeddings, as well as certain limitations on ambiguity resolution, length effects in garden path structures, and the requirement for locality in syntactic structure. The theory takes the form of an architecture for parsing which can index no more than two constituents under the same syntactic relation. A limitation of two or three items shows up in a variety o...

The authors present a theory of how relational inference and generalization can be accomplished within a cognitive architecture that is psychologically and neurally realistic. Their proposal is a form of symbolic connectionism: a connectionist system based on distributed representations of concept meanings, using temporal synchrony to bind fillers and roles into relational structures. The authors present a specific instantiation of their theory in the form of a computer simulation model, Learning and Inference with Schemas and Analogies (LISA). By using a kind of self-supervised learning, LISA can make specific inferences and form new relational generalizations and can hence acquire new schemas by induction from examples. The authors demonstrate the sufficiency of the model by using it to simulate a body of empirical phenomena concerning analogical inference and relational generalization. A fundamental aspect of human intelligence is the ability to form and manipulate relational representations. Examples of relational thinking include the ability to appreciate analogies between seemingly different objects or events (Gentner, 1983; Holyoak & Thagard, 1995), the ability to apply abstract rules in novel situations (e.g., Smith, Langston, & Nisbett, 1992), the ability to understand and learn language (e.g., Kim, Pinker, Prince, & Prasada, 1991), and even the ability to appreciate perceptual similarities

In this research, we aim to ground scene recognition on information other than the identity of component objects. Specifically we seek to understand the structure of color cues that allows the express recognition of scene gists. Using the L*a*b* color space we examined the conditions under which chromatic cues concur with brightness to allow a viewer to recognize scenes at a glance. Using different methods, Experiments 1 and 2 tested the hypothesis that colors do contribute when they are diagnostic (i.e., predictive) of a scene category. Experiment 3 examined the structure of colored cues at different spatial scales that are responsible for the effects of color diagnosticity reported in Experiments 1 and 2. Together, the results suggest that colored blobs at a coarse spatial scale concur with luminance cues to form the relevant spatial layout that mediates express scene recognition. © 2000 Academic Press Key Words: scene; color; diagnostic information; recognition; categorization; spatial scale; L*a*b*; spatial layout. In Potter’s (1975) classical scene-recognition experiment, subjects faced a screen on which slides of real-world scenes appeared in rapid succession (at a rate of 125 ms/slide). Their task was to press a button as soon as they detected, e.g., a beach. Subjects ’ efficiency was very high and this presents a puzzling problem for scene analysis: how can a scene be so rapidly recognized despite its variability, large number of component objects, and multiple sources of interfering factors? Following Marr’s (1982) influential conception, scene recognition has