Notes and Comment Resolution in one dimension with random variations in background dimensions

Abstract

In the usual one-dimensional resolution (discrimination or identification) experiment, the stimuli vary along one dimension, the target dimension, and the observer is required to respond differentially on the basis of the stimulus value along that dimension. In this note, we focus on experiments in which the observer's task remains the same, but in which the other parameters that define the stimuli, the background parameters, are varied randomly over specified ranges of these parameters. Thus, for example, the stimulus set might consist of tone pulses of fixed duration, the target parameter T might be the carrier frequency of the tone pulse, and the background parameter B that is randomly varied might be the amplitude of the tone pulse. The terms FIXED or CERTAIN are used for situations of the first type, whereas ROVED or UNCERTAIN are used for situations of the second type. Obviously, paradigms of the second type converge to those of the first type as the range of variation for the background parameters goes to zero. In this sense, FIXED paradigms can be regarded as a special case of ROVED paradigms.

Application of Psychophysical Techniques to Haptic Research

“…The 1D IT s so obtained are, therefore, appropriately reduced to account for the perceptual interaction between F and A. This general additivity law has been empirically verified in several studies [75], [80], [81], [82]. It can be easily extended to more than two dimensions in the form: IT ðA; B; C; .…”

“…is limited to 2.3 to 3.2 bits; or equivalently, 5-7 perfectly identifiable levels. In haptics research, the unidimensional channel capacity appears to be lower: up to four levels for finger-span length and close to three levels for force magnitude and stiffness identification [70], [75]. Experimentally, as the number of alternatives in a stimulus set increases, estimated information transfer will initially increase linearly and then reach an asymptotic value that represents the maximum information transfer that can be obtained with the stimulus set, that is channel capacity (see [76] for data on size identification).…”

“…This required a total of 78,125 trials according to the 5K 2 rule! Durlach et al [75] proposed a general additivity law that allows the prediction of multidimensional IT from IT s estimated with unidimensional stimuli. For example, let us consider the case of a 2D experiment where the stimuli vary along two dimensions (e.g., frequency and amplitude of a vibration).…”

“…For example, within a force range of 0.1-5.0 N, there are 41 JNDs (assuming a force-magnitude JND of 10 percent) but the information transfer is 1.54 bits (or equivalently, only three perfectly identifiable force magnitude levels) [70]. This is explained by a decision model developed by Durlach et al [75] stipulating that whereas the JND is limited by the "sensory noise" in our peripheral sensory systems, IT is further limited by a "memory noise" due to the need to hold the absolute rather than relative force-magnitude percepts in our memory. Therefore, it is much easier to discriminate the relative magnitudes of two forces than to identify the absolute magnitude of a force in isolation, or to discriminate the relative locations of two tactors on a belt than to identify the absolute location of a single tactor on the belt.…”

Perceptual limits for a robotic rehabilitation environment using visual feedback distortion

“…Most importantly, all of the JNDs cited earlier were measured with all physical dimensions other than the target dimension fixed. These other physical dimensions are called background dimensions [20], and the most common way to measure kinesthetic JNDs is to keep the background dimensions constant. Our force JND experiment, however, has two unfixed background dimensions that are related: distance and the stiffness of the virtual spring.…”