50 Years of Research into Haptic Perception

Gibson’s 1966 book The Senses Considered as Perceptual Systems recently turned 50. Two issues of the journal Ecological Psychology commemorated that event (here, and here). This is the third in a series of posts reviewing those contributions. It covers Carello & Turvey's Useful Dimensions of Haptic Perception: 50 Years After The Senses Considered as Perceptual Systems.

Haptic perception is extremely neglected relative to visual and auditory perception. The term could refer simply to feeling things by touch, but in the context of EcoPsych is more likely to refer to perception as the result of manipulating objects, i.e., picking things up and moving them around. The Senses Considered included chapters about the haptic system, but offered only a cursorily outline of what an improved study of the haptic system would look like. Some the first wave of Gibson-inspired researchers latched onto those chapters, and created some of the more notable research triumphs of the field. Carello and Turvey performed, or supervised people who were performing, much of that work. Given that several good summaries of the research exist, they choose to focus instead on showing how the haptic research has been a uniquely suited context for exploring the novel implications of an ecological approach to perception.



First, a contrast must be made between what a perceptual system can do, and what its function is. The function of a perceptual system is to put the organism in touch with its environment, on a scale relevant for action. The Cartesian starting point for most research imagines that the organism is trying to create an 'internal world' and the Marr starting point imagines that the organism is motivated to process information, but the ecological starting point is that the organism is trying to act in a world. In contrast with the large psychophysics literature on "passive touch" (in which experimenters poked the body in various ways and asked for self-reported experience), the ecological approach starts with "active touch" (in which the organism's movements create feedback loops). There is much research showing that the ways in which people touch and manipulate objects varies depending on what they will need to use the objects for. However, ecological psychology pushes even a step beyond that, in addition to knowing how people move objects in order to determine different uses the objects have, ecological psychologists want to characterize the variables detected by such movements. Clipping a bit from the article:

"The research and theorizing inspired by the Gibson chapters respect the proper function of touch. Many systems can do all sorts of activities that do not necessarily reflect their raison d’etre—the core capabilities that have warranted their persistence as systems. Although one might be able to localize poking from a stiff bristle somewhere on the skin surface or appreciate that there are two bristles rather than one [as the classical psychophysicists studied diligently], a psychology built on skin localization and 2-point thresholds will never approach an understanding of sensitivities that matter to effective behavior. Effective behavior is made possible by obtaining information about the environment, about oneself, about one’s relation to the environment, and about the environment’s relation to oneself. The haptic perceptual system must underwrite exteroception, proprioception, exproprioception, and proexteroception, respectively."

That is, wielding an object properly can tell you something about the object, about your self, about yourself in relation to the object, and about what relations you might have with the object in the future.

The most noteworthy of these efforts build off of Solomon & Turvey's (1988) work on perceiving the length of a wielded rod, where "length" is shorthand for the functional judgement of how far away one could poke an object with the rod. Hundreds of projects out of the University of Connecticut and its allied schools study this problem. What is eventually revealed and repeatedly confirmed is that:

1. The "intertia tensor" or "2nd moment of momentum" or, in something closer to layman terms, the amount of resistance to rotational movement, specifies the length of the rod,
2. that people tasked with judging length move the rod so as to reveal that variable, and 
3. (discussed towards the end) this works whether the rod is held normally in the hand, taped to the wrist or feet, mounted on the back, held under water, and countless other variations. 

If you give people more complex objects, the relative ratios of the moments of inertia and the products of inertia become more useful (illustrations and detailed discussions for all these points is included). Though it is typical to think of these movements in a Cartesian frame with x, y, and z dimensions oriented along with the floor and right-angled walls, there are a wide (but not unlimited) range of dimensional orientations that would yield the same results. Generalizing, one could say that:

"Experiments have shown, in brief, that perceivedmagnitudes (e.g., height, width, weight) are specified by eigenvalues, perceived directions (e.g., object orientation) are specified by eigenvectors, and perceived magnitudes in a direction (e.g., length in front of the grip) are specified by eigenvalues and eigenvectors together."

Clever experiments have also revealed that the numerous specifiable object properties can be detected independently of each other. That is: "Contrary to classical computational views, one does not perceive one property and use it to infer another."

The bulk of the rest of the paper - very math intensive but supported by good conceptual discussion - pushes the final step of the ecological program, showing:

1. That the variables being described can be transformed a bit more to yield useful dimensions of the object; that is, they specify affordances best understood in terms of what the organism can do with the object;
2. that the perceptions in question are not dependent upon particular sensations (that is, on particular patterns of firing in particular sensory neurons), as the larger haptic system can pick key variables in a variety of ways; and
3. that the haptic system is well designed to take advantage of the redundancy in ways of obtaining key information by being highly redundant and interdependent itself. 

The idea is hard to present elegantly, Turvey and Carello give a jargony summary, but I don't think I can improve upon it at this time:

"What we have come to appreciate is that the body understood as a muscular-connective tissue-skeletal system and interpreted as a tensegrity architecture—continuous tension elements and intermittent compression elements at all of its scales—promises the requisite homogeneity and isotropy for the medium of haptic perception."

One benefit of thinking of the muscular-connectiveTissue-skeletal system as a giant tension array is that the state of that array is specific to both the where the body is, what it is doing, and how "external" objects and events are affecting the array.

The concluding sections of the paper - quite elegant - draw the parallels between the haptic system and the optic system which are often extremely difficult to appreciate. One section is even titled "Proprioception is not special"!

While this paper is not as approachable as many of the others in the Special Issues, its technical prowess and evidentiary weight is impressive. For those not familiar with this impressive body of work, it will be a very good place to start, and more importantly (in this context), it will start the reader with an excellent grounding of how this work has built out of, and developed ideas that Gibson innovated in those few daring chapters, 50 years ago.