Digital image-processing systems combined with color raster television technology are opening new worlds for the visual psychophysicist. Some of the inherent limitations and practical hazards of these technologies are explored. Techniques for circumventing some of the current limitations are presented, especially with regard to the issue of real-time variation of visual images. This paper is divided into two parts. In the first part, I examine the question: To what extent can we generate arbitrary visual stimuli using current computer imageprocessing systems? Thisquestion can bestbe approached by making analogies between vision and audition. For both vision and audition, the problem is divided (somewhat idiosyncratically) into four issues: (1) abstract representation of the stimulus (e.g., an equation or algorithm); (2) explicit representation of the stimulus (e.g., pixelvalues); (3) communication of the explicit representation to a physical transducer; and (4) physical transduction. Limitations of current theory or hardware are pointed out for each issue.In the second part of the paper, I describe several instances (fromactual practice) in which varioustechniques wereusedto generate stimuli thatwerenot whattheunderlying hardware devices were designed to generate; in short, I will document various tricks used to get around some of the limitations. Attention will be focused on a specific image-processing computer system, but the ideas are of wide applicability, given the current generationof imageprocessors. At somepoint in the future, there may be image processors and displays that are designed explicitly for use in psychophysical vision research; in the meantime, however, the experimenter can generate a surprisingly general collection of stimuli despite hardware constraints.
The Dream of Psychophysical OmnipotenceWhat is needed to be able to create, in a laboratory, any visual stimulus? What information rate is required? How can the information be stored, communicated, and transformed into light? In addressing these questions, it is helpful to first considerthe analogous questions for audition, using data summarized by Green (1976 An arbitrary auditory stimulus has two dimensions: intensity and time. For most purposes in psychophysics (i.e., ignoring such things as bone conduction), if we specify intensity as a function of time at two points in space (one for each ear), we have specifiedthe auditory stimulus completely. Furthermore, the required bandwidth for an auditory stimulus is 20 kHz. If we were to specify an auditory stimulusat discrete points in time separated by, for example, %0,000 sec (twice the maximum frequency), then any errors introducedby temporal quantization would be irrelevant, because the auditory system cannot respond to them. (This is not the whole story, as temporal differences in phase between the two ears can be detectedat finer temporal resolution.) Audible sound pressurescover a range of 10 log units, and intensity discrimination is on the order of I %. Ifwe were to represent intensit...