The role of haptic versus visual volume cues in the size-weight illusion

0 This research was supported by a contract to S. J. Ledennan by the Manufacturing Research Corporation of Ontario Centre of Excellence, and by a Postgraduate Level Scholarship from the Natural Sciences and Engineering Research Council of Canada to R. R. Ellis. The research fulfilled part of the requirements for a master's degree (Ellis , 1990). We would particularly like to express our appreciation to the Montreal Association for the Blind (Paul Barber was especially helpful in assist ing us) and to the subjects who participated in Experiment 3. We also thank Cheryl Wilson. Reprint requests should be sent to S. J. Ledennan, Psychology Department, Queen's University , Kingston, Ontario K7L 3N6, Canada Three experiments establish the size-weight illusion as a primarily haptic phenomenon, despite its having been more traditionally considered an example of vision influencing haptic processing. Experiment 1 documents, across a broad range of stimulus weights and volumes, the existence of a purely haptic size-weight illusion, equal in strength to the traditional illusion. Experiment 2 demonstrates that haptic volume cues are both sufficient and necessary for a full-strength illusion. In contrast, visual volume cues are merely sufficient, and produce a relatively weaker effect. Experiment 3 establishes that congenitally blind subjects experience an effect as powerful as that of blindfolded sighted observers, thus demonstrating that visual imagery is also unnecessary for a robust size-weight illusion. The results are discussed in terms of their implications for both sensory and cognitive theories of the size-weight illusion. Applications of this work to a human factors design and to sensor-based systems for robotic manipulation are also briefly considered. - In experimental psychology, weight perception can be traced back to the early experiments of Ernst Weber (1834/1978). His main interest was whether weight per ception resulted more from cutaneous inputs alone or from the muscular sense associated with lifting an object. He found that weight discrimination is more exact if the ob ject is actually lifted rather than simply placed on a hand passively resting on a table. Since these early experiments, weight perception has been and continues to be inves tigated in many psychophysical experiments (see recent review by Jones, 1986). Despite this extensive research, it remains a problem for perceptual theorists. Weight is an important dimension of an object, partic ularly if it has to be moved or manipulated by either a human or a robot. Its assessment presents an interesting paradox to a manipulator-in order to manipulate an ob ject efficiently, its weight must be considered; however, to judge its weight, the object must be lifted. One cau tious solution would be to try to lift the object with a small lifting force. If this force proved to be ineffective, it could be increased slight!y. If this too proved to be unsuccess ful, it could be increased repeatedly until an effective force was found that just lifted the object. A more efficient solution to this paradox would be to use knowledge of particular objects and their properties derived from past experience for an initial weight estimate. This solution is limited to the extent that it would only succeed with objects that the manipulator had previ ously encountered. A third and more universal solution would be to take advantage of the correlation between weight and volume, although no published research on this topic has been lo cated. Large objects (particularly in the natural environ ment) tend to weigh more than small objects. Volumetric information can be quickly and easily processed without physically moving the object. Because a correlation be tween volume and weight is likely, knowing an object's volume provides coarse predictive information about its probable weight. It could, however, provide some sur prising results in situations in which this natural correla tion is violated. As humans move from terrestrial environ ments, where g-forces are quite constant, to either aquatic environments or outer space, where g-forces vary, the relationship between volume and weight becomes more uncertain, although the mass remains constant. Charpentier (1891) first demonstrated that the perceived weight of an object, commonly referred to as its heavi ness, depends not only on its physical weight but also on its size. He presented observers, who were allowed vi sion, with two spheres (40 mm and 100 mm in diameter) of identical weight and had them lift each with the palm of their hand. Hand movements were not specified, but it is reasonable to assume that rather than keeping their palms rigidly flat, the observers cupped their hands to ob tain volumetric information haptically as well as visually. The larger sphere was consistently reported as lighter. This phenomenon became known as the size-weight illu sion. Flournoy's (1894) experiments extended the range over which the illusion occurred and also demonstrated that the illusion was so compelling that it persevered even when the observers were told that all the objects weighed the same. Many theories have attempted to explain the size-weight illusion. Early researchers regarded this phenomenon as a reflection of a density-constancy process (Thouless, 1931). Unfortunately, although an object's density (i.e., mass [or weight, if on earth]/volume) is undoubtedly in volved, this concept of density-constancy does not explain any of the underlying mechanisms responsible for the ob served phenomenon. Another explanation involves the expectation theory (Ross, 1969), which states that prior experience with ob jects leads observers to expect that a larger object will be heavier than a smaller object. The correlation between large volumes and heavy weights would be high for com parisons within object sets that are either all solid or all uniformly hollow and made of the same (or very similar) materials. In cross-set comparisons, or when the objects are made up of radically different materials (i.e., sponge and granite), the correlation would be lower. However, these cases would represent exceptions, and in many in stances would be artificial or man-made. Overall, partic ularly in natural environments, it is reasonable to assume that there is likely a somewhat significant correlation be tween large volumes and heavy weights. This learned cor relation results in an expectation or mental set that could affect the force an observer applies when lifting an object. A series of experiments by Davis and Roberts (1976) supports this cognitive theory. They found that when ob jects of identical weight are lifted, the larger objects are lifted with greater force, and therefore more quickly. Be cause it is assumed that subjects would attempt to lift all objects at the same rate, the greater velocity, accelera tion, and deceleration found during the lift phase proba bly reflect the fact that the ob (...truncated)