How do people perceive the size of objects visually? Since the size of the visual image of an object varies depending on the distance of the object from the observer, this has always been a problem for perceptual theorists. One possibility is that people know the size of familiar objects that they can recognize from their shapes. An important category of recognizable objects consists of biological objects whose sizes vary with growth. Physical laws that constrain growth processes produce subtle changes in the shape of biological objects that correspond to changes in their sizes. Specific shapes, therefore, correspond to specific sizes. Is it possible that observers can discriminate these subtle variations in shape and use them as visual information about size? This research will investigate two physically determined aspects of tree shape as information about tree height. Observers will judge computer-generated tree silhouettes appearing without any other visual information about size or distance. All silhouettes will be of the same image size. The research will use a computer model derived from biophysical studies of tree growth. The research will compare results with the computer-generated tree silhouettes to perceptual results with silhouettes of real trees and with judgments of real trees in natural viewing conditions. Tree motions also vary in ways specific to tree size, due to some of the same physical constraints determining tree shape. The research will isolate these motions in patch-light displays, in which only the tips of the branches can be seen moving, and test their role as visual information about size. Results of these studies will contribute to an understanding of scaling in space perception, especially in circumstances where distance information is perturbed, as in looking through telescopes or other imaging apparatus. Previous research has shown that motions from a variety of simple physical events appearing in patch-light displays can enable visual recognition of the events, while static images from the displays are unrecognizable. What characteristics of motions can observers discriminate and use as information about events? The second part of this research will manipulate motions in computer- generated displays via dynamical models of these simple events. The importance of using dynamical models in this way is that they capture the physical properties of events that determine the motions specific to an event. The simulations will test the ability of observers to discriminate different types of motions and to recognize different events and event properties. The results will be especially relevant to understanding visual recognition in situations where only low frequency spatial patterns are available, as in vision with cataracts or with night vision goggles.
