Perception & Psychophysics 1989, 46 (I), 29-38 Manual discrimination and identification of length by the finger-span method N. I. DURLACH, L. A. DELHORNE, A. WONG, W. Y. KO, W. M. RABINOWITZ, and J. HOLLERBACH Massachusetts Institute of Technology, Cambridge, Massachusetts Experiments were conducted on length resolution for objects held between the thumb and fore- finger. The just noticeable difference in length measured in discrimination experiments is roughly 1 mm for reference lengths of 10 to 20 mm. It increases monotonically with reference length but violates Weber's law. Also, it decreases when the subject is permitted to maintain a constant finger span between trials; however, it tends to increase when the nondominant hand is used. As would be expected from studies of other stimulus dimensions in other sense modalities, reso- lution is considerably poorer in identification experiments than in discrimination experiments. For stimulus sets that cover a broad range (90 mm), the total information transfer is roughly 2 bits; for those that cover a relatively small range (18 mm), it is roughly 1 bit. The data are analyzed and interpreted using analysis techniques and models that have been used previously in studies of audition (e.g., Durlach & Braida, 1969). This paper concerns the ability of humans to dis- criminate and identify the extent of an object by holding it between the thumb and forefinger of a single hand (the finger-span method of length estimation). We are in- terested in the ability to resolve length in this fashion for a number of reasons. First, we want to characterize, understand, and model the human's ability to recognize objects manually. We want to achieve this goal both as an end in itself and as background for the design of improved robots. Although some previous work has been conducted in this area (e.g., Lederman & Klatzky, 1987), the results are still very limited. To understand object recognition of any kind, it is necessary to (1) determine the dimensions (first in phys- ical space, then in perceptual space) that are relevant to the given recognition task, (2) measure the ability to resolve differences along these dimensions, and (3) construct a model of how the sensed values along the different dimensions are combined and used to select a specific recognition response (a model that must take into account a priori information, payoffs, decision making, short-term memory, etc.). The experiments reported in this paper are addressed to requirement 2. Although the specific dimensions that are important in a given recog- nition task will obviously depend on the collection of ob- This work was supported by NSF Grants DMC 83-52460 and BNS 84-11811, Nlli Grant 5 ROI NS 14092, and ONR Grant NOOOI4-88- K0338. The authors are indebted to R. Uchanski for her help on some statistical questions and to C. Sherrick and M. Teghtsoonian for many useful comments on the original manuscript. Correspondence may be addressed to Nat Durlach, Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA 02139. 29 jects in the stimulus set, many practical recognition tasks involve object extent as a significant parameter.' A second motivation for studying length resolution stems from our previous research on tactual communication of speech by the deaf-blind (e.g., Norton et al., 1977; Reed, Rabinowitz, Durlach, & Braida, 1985). To understand the remarkable performance achieved by these individuals, we need an improved understanding of the hand's ability to discriminate and identify patterns of stimulation that involve changes not only in vibration and airflow, but also in shape and compliance. 2 Third, we are interested in man-machine interfaces for teleoperator and virtual-environment systems. Knowledge of the operator's resolution limits is essential to the de- sign of efficient interfaces; the resolution of the interface should be appropriately matched to that of the operator. Finally, there is the continuing uncertainty (and con- troversy) over the physiological mechanisms that under- lie the sense of finger position. Although information on finger position is available from cutaneous and muscle receptors and also, to some extent, from joint receptors, the relative contribution of each of these mechanisms to various systems under various conditions is not yet well understood (e.g., see the review by Clark & Horch, 1986). Since measurement of object length by the finger- span method involves the sensing of differential finger p0- sition, a comprehensive and accurate characterization of finger-span length resolution could provide useful data for the evaluation of hypotheses concerning the mechanisms that underlie the sense of finger position. In general, there are several methods of estimating length manually: (1) by the finger-span method considered in this paper (or one in which the two fingers are on differ- ent hands); (2) by the temporal-sweep method, in which a fixed area of skin (e.g., a finger pad) is swept across Copyright 1989 Psychonomic Society, Inc.