The Effects of Representation on the Processing of Probabilistichformation. Printed in The Netherlands Table of Contents. Quarterly Journal of Experimental Psychologll, Treisman, A. Other researchers have not explicitly dealt with this distinction previously but have contributed knowledge to various aspects related to this distinction. The Basic Level in Natural Categories. T h e drawing of her theory shown in Figure 2.
The next section considers the general nature of the segmentation task and the way in which texture can provide the basis for image segmentation. If this is the case, the rapid detection of line relations in the previous experiments should be explicable purely on the basis of topological considerations. For example, in order for even simple spatial relations to be determined between features e. Journal of Experimental Psvchologv: Human Perception and Performance, 12-31. Learning Categories With and Without Trying: Does It Make A Difference? Visual masking and task switching in the attentional blink. Investigations of exemplar and decision bound models in large, ill-defined category structures. In such feature-based studies the texture elements can be delineated syntactically or obtained by convolution masks used to extract preconceived features.
Infant sensitivity to figural coherence in biomechanical motions. Experiment 1 0 Not all trihedral junctions receive equal treatment in early vision. The complexity of the image structure conveying the texture information is illustrated in Figure 1. Influence of scene-based properties on visual search. . The brain is an organ consisting of a large but finite number of neurons. Thus, as Essock points out, the relative lack of constraint in computer vision research may have its advantages, but as far as modeling human perception is concerned, it is also a weakness.
Both of these particular feature maps suggest important discontinuities between roof and sky and between tree and sky. What is the nature of the information from which this choice is made? Prototype and exemplar-based information in natural language concepts. We predicted that 7-month-olds would look longer at inconsistent than at consistent trials. Previous studies suggest an early sensitivity to some dynamic cues to animacy e. Color constancy may also operate- a t these levels Land, 1977. The development of infant sensitivity to biomechanical motions. The most prominent view over the past thirty years is a feature-based view as presented by Beck, Attneave, and others e.
The control condition consisted of scrambled point-light animations, which are perceived as meaningless dot movements rather than human actions for further details on see. Invariably edge contrast, edge brightness level, and brightness of the primitives are confounded in the simple patterns typically utilized. . See for instructions, and for results of the motion rating experiment. The Rapid Recovery of Scene-Based Properties in Early Vision In this section I will summarize a series of studies that have begun to map out the scene-recovery capabilities of early human vision. In the motion control condition, the same concentric rings changed in luminance, but did not move ;. This, of course, has been the approach taken by Beck 1982 , Julesz 1981 , Treisman 1985 and others working within a psychological research framework.
However, solving the task for problem B shows that more is involved. Psychological Review, 111, 309— 332. In addition to being sensitive to specific properties of simple geometric elements, it is sensitive to several complex image properties that correspond to intrinsic properties of the scene. These questions will be addressed in the first section of the chapter. The total duration of the presentation all four trials was about 3.
Evidence for a Structure-Process Link in Incidental Learning E. On Identifying Things: A Case For Context; Part B: Percepts, Concepts, Categories and Development; Chapter 5. The present study demonstrates that these regions do not represent visual-motion properties, but their precise role remains unclear. T h e first is the question of what structure is available to the perceptual system in a given environment. The map indicates where and how many features occur, but not which features exist at a given location.
Blocks were 18-seconds long and were separated by 14 seconds of fixation. All experiments are from Enns and Rensink 1991. Computer Graphics and Image Processing, 8, 313-333. In models that specify only the length and width of blobs Julesz, 1984b , or even area Voorhees and Poggio, 1988 , shape is left only vaguely specified. Mandler indicates that we have not dealt adequately with literature demonstrating what she sees as a trend for infants' first forming abstract categories and later forming more specific ones. It is worth emphasizing how different this is from the conventional view of early vision.
Research programs of many of the contributors have been closely identified within this framework of characterizing and evaluating structure and process in the study of percepts, concepts and categories. High motion words Action Verbs + Animal Nouns did not differ in average pair-wise similarity from low-motion words Mental Verbs + Inanimate Natural Kinds t. Influence of Learning History Factors Upon Representation 401 V1. This interpretation is at present highly speculative. Why is it necessary for every information processing system to be selective? The type of motion pattern shown during the motion scene alternated between the four trials.
This general framework which emphasizes extracted texture features is consistent with viewpoints expressed by Julesz, Treisman, Marr, and Beck, as well as many others. That linear-filter models are that different is questionable for at least two reasons. Condition C showed that the 66 James T. Feature analysis in early vision: Evidence from search asymmetries. . We were somewhat surprised to learn that early vision was sensitive to some line drawings but not to others. They also argue that a second nonlinearity following rectification to suppress weaker responses within a local spatial neighborhood which functions something like a threshold , is optimal.