Touch and the Observer's Vantage Point

John M. Kennedy
University of Toronto

     On a clear day, when you seem to see forever as you stand spellbound before a vista of distant mountains, you have an impression of space, but you also have a well-defined vantage point. The vista specifies your own unique location (Gibson, 1979). If you take a photo that day, the photo tells where you were standing. It says "you were here!" The contours of the hills not only reveal where they are, silhouetted against the sky, they also indicate the special spot from which the photo was taken. If you are on a mountain track, and you move to one side, to take more pictures, the shapes of the brows of the hills will change slightly to specify your trajectory. Parts of distant hills evident in one picture may be hidden in a shot from a neighbouring viewpoint along the track. If the vista opens out to the ocean in one direction, you may see as far as the horizon. 

     Visually, the dimension of distance has as one anchor the distal object. The other anchor is the observer's vantage point: no vantage point, no distance. Acting as a far target for observation, the horizon often anchors one end of the dimension of distance. It is a visual limit for a terrestrial plane. The other end is the observer's vantage point. It is the origin for measures of the distance of the target from the eye. The origin  is the centre of a sphere of directions. From the origin we can move our gaze in six ways . We can change our heading via yaw, pitch and  roll, and we can move our origin up, sideways and forward. That is, not only do we  have to look into space from our own limited standpoint, we also have to gaze in a particular direction at any given moment. We look up (changing pitch) to the heavens above the horizon, or down to the ground. We can  look left and right (yaw) to where our path may take us, perhaps along a cliff edge of a plateau. We can revolve to stand on our head (roll). Also, our origin can be moved left and right along the path as if on a moving belt, be raised or lowered as if on an elevator, or tempt fate by allowing itself to go to the front and towards the edge of the bluffs or to the back and safely away from the precipice. There are three ways to change our direction of gaze from a fixed origin, and three ways to move the origin: Six degrees of freedom for our singular vantage point.

     The observer's vantage point is evident in vision and it is made known precisely and exactly in pictures based on optic projection. It has six degrees of freedom we take liberties with daily. Is there anything like the eye's vantage point in touch? Or does touch depend entirely on direct contact, that is on stimuli in proximity to the body? Does touch rely so much on proximal arrays that it resists any use of a vantage point? Is there any way in which touch enquires about distant objects, far removed from the observer, not abutting the body? In what ways might touch act as a distal sense as well as a proximal sense? Does touch have degrees of freedom? Does it use contours of objects, like the brows of hills, to specify its location? How might a  change in tactual location  be reflected in a tactual vantage point? Where is the tactual sky and the horizon?

     Visual impressions of a distinct, precise vantage point are well matched by the photos we take of the attractive vista. The photo is a record of the light rays coming to a single point, through a lens. If we made pictures for touch, using displays with raised elements (Edman, 1992) would touch provide us with information about a particular vantage point? Some might think the pictures would  do so only  for the sighted, who can imagine what the tactile display could look like, and interpret the picture bearing in mind a visual vantage point (Revesz, 1950). Surely there would be a lot to learn  about vantage points if a blind person unfamiliar with pictures were able to use a picture drawn from a single vantage point.

      I will discuss the questions here in an argument about dimensions of distance in touch. Fundamentally, I will propose that vantage points abound in touch. And I will contend they can be used usefully in tactile pictures, and be interpretable through touch in similar ways by the sighted and the blind.

   Let us begin with a thought experiment. Imagine touching or looking at a line of raised dots (Arnheim, 1974;  Kennedy, 1993, 1997; Holmes et al, 1998). The dots are elements that induce a perceived line. The perceived line crosses the empty space between the dots. Much of our perception of space is like this. We see or touch a few objects on a surface and gain an impression of the relations between them. We also get an impression of the relations between the inducers, the induced line and our own vantage point. The vagaries of these relations are the topic to be discussed here.  

The historical legacy: Molyneux's question and Murphy's Law

     Many scholars have asked about the relation between vision and touch. Vision's vantage point was often plain, and often mistakenly assumed to be entirely obvious, in these discussions. Alas, the idea that touch might have use for vantage points was often conspicuously absent. The result was, I think, a very lop-sided debate.

     William Molyneux , an Irish barrister of the late seventeenth century, had a wife who was blind, and was moved to throw a celebrated question into the pool of philosophical debate.  Query: would a blind person familiar with cubes and spheres  be able to recognize them if given sight by some enterprising operation? Many eminent scholars who did not know the answer rushed to reply. The character of  their arguments set the tone of enquiry for many years, with vision given abilities explicitly, and touch belittled by errors of omission. In a variation of Murphy's Law that what can go wrong will, parallels that might be misconstrued were.

    John Locke ( 1690, see also Boring, 1942, 1950) in an "Essay Concerning Human Understanding" commented  that vision gives us more than light and colours. It also gives us the "far different ideas of space, figures and motion". Locke discussed projection to a vantage point. He noted that a sphere is projected in vision as a circle and a cube as a square or hexagon. He described

these projections in terms of vision. He did not consider projection in touch.

     Reid (in 1764,  see Boring, 1942, 1950) astutely described how vision sees changeable aspects of an object in its projections. But in touch, he thought, we gain the impression that all these objects are identical. Vision, Reid wrote, initially takes a sphere as a circular form, variously coloured if it is partly in light and partly in shadow. But the genius of perceptual learning is that aided by touch we can discover that different distances are relevant, not just various colourings, and "this perception" gives the circular form convexity, adding a third dimension (VI, section 23).

     Synge (in 1693, see Eriksson, 1998) debated the basis for Molyneux's question, asking what a person born blind might have as an idea of a sphere or a cube.  A tactile idea of a sphere, Synge proposed, was of an object that felt the same all over. In contrast, a cube has distinct parts. Some are sharp vertices, some are flat, and some are long straight corners between flat areas.

     Berkeley (in 1709, see Eriksson, 1998) noted when we look at a point, the point will not tell us whether it comes from a short distance or a long distance. Its distance is indeterminate.         

     Locke, Reid , Synge and Berkeley do not offer systematic conjectures about touch having vantage points, dealing with projections and anchoring distance information. Locke's spheres and cubes project shapes in vision, and not in touch. But the direction of parts of objects vary just as much in touch as they do in vision. Reid failed to mention that cubes have different aspects to touch. Synge's spheres always feel the same. He does not mention there are many ways one might vary the vantage points from which contact is made. When Synge makes the point about distinct parts of a cube, one wants in vain to have an orderly treatment of the fact that some of the parts of the cube could be  "near" and others "far". Berkeley's visual point on our retina could be compared usefully to a tactile point on our skin. The visual point does not tell us how far it's straight-line transmission has come. Thar requires a specific informative context for the point (Gibson, 1979). The tactile point does not tell us how long a rod is behind it. Wielding a rod does tell us about its length ( Turvey, 1995).

     Diderot (in 1749, see Morgan 1977), French encyclopedist, is a radical in this company of  British  Empiricists. Ironically, he offered more empirical observations than the Empiricists. His observations make notions of a tactile vantage point relevant.  He discussed the abilities of two blind men -- a man from Puiseau, and a mathematician from Cambridge. Both of these men dealt with shape and distance. They appreciated that we reach out for objects, from wherever we are standing. The man from Puiseau spoke of reaching out with his stick. Sometimes an obstacle might block his access to the object he was trying to touch with his stick. The outcome is a valid awareness of spatial properties, the body as origin, several objects around it, some near and some far, an outstretched arm, extended by a stick, and occlusion of one object by another. 

    Is there direction in touch? A static stick does not have direction, but, as Diderot mentioned, a motion is in a particular direction. What about shape? Our hands pass through a succession of places, in following a string, Diderot wrote.  If the string is taut, it provides a succession of points or places that can be combined, using memory, into a straight line. If it is slack, the combination that will result will be a curve. We can recall the shapes and refer to the properties we discover through touch , across  a succession of points, he conjectured.

      Diderot's discussion lead him into fierce contradictions. We combine tactile points, using memory, and can later "refer" to  the products, Diderot believed. But he went on to argue that touch will not allow us to "imagine" figures. He argued that to imagine figures  we have to separate the lines or borders of shapes from their background, and this requires the lines or borders of figures to be defined by "different colours" than the background. This is clearly silly. We can imagine raised lines, not just coloured lines. In both vision and touch, we perceive continuous lines induced by rows of dots. The spaces between the dots have no specific colour or height. The perceived lines too have no colour or height.  We can combine the dots, use memory, and later refer to the result, but not imagine what we have done? Tut-tut! Imagining and referring seem suspiciously like the same operation by two names.

     There are chicken-and-egg  problems in Diderot. What enables us to know that a set of points is in a straight line? The line itself cannot tell us, because we could be suffering an illusion. Evidently there is an empirical question here: What is taken to be straight and crooked in touch? We cannot  just assert straight things perforce are perceived as straight. 

    Let us take away from Diderot one valuable idea: touch involves reaching from a direction, and so we have at least one kind of vantage point in touch.

    Running counter to his own restrictions on imagination,  Diderot noted that a blind man could consider a sphere, and then envisage a smaller or larger object, with the same shape. Change of scale leaves shape invariant. In this fashion, the blind man could imagine the terrestrial globe. Atoms and molecules can be imagined in the same fashion. But further, surely the blind man can imagine where atoms or celestial spheres are in direction from us: in front of us or to one side, near or far, small distances or huge ones. That is, direction in touch implies a wide range of distances cognitively, it is likely. 

   The idea that touch supports implied relations between ourselves and objects deserves attention. Just as touch's reaction to truly straight things needs to be explored, so too our consideration of possible vantage points, some real and some imaginary, cannot be taken for granted. Just as there are explicit numbers, numbers that are implied, numbers that are real and numbers that are imaginary, so too touch may serve an observer entertaining many kinds of vantage points, some real and some imaginary. Since we reach out in certain directions deliberately, we obviously have goals before we reach: The imagined or perceived directions of targets. Since we can intend to move our vantage point to pick up objects that are just out of reach at the moment, we can imagine moving our vantage point.    

    What about the relations  between two directions? It is this that gives us perspective, let us note. Descartes (in 1638, see Boring, 1942) came close to posing systematic parallels between vision and touch in their use of perspective, and his conjectures may have influenced Diderot writing about a blind man reaching out with a stick. Descartes explained that in visual fixation our two eyes converge on a target (Cabe et al, in preparation). The directions the two eyes are gazing are like two rods crossing at the target (Boring, 1942). If we can look from two directions in a fashion that is like using two sticks, and gain perspective on an object's location and distance, surely we can reach with one stick in two directions and gain similar knowledge. Descartes described a blind man holding two sticks that intersect a short distance in front of him. Descartes suggested the man could estimate the distance to the crossing-point. The estimate would be made using a kind of natural geometry, Descartes believed, based on the angles of the hands, wrists and arms, and similar effects would arise in visual convergence. In fact, vision is poor at using convergence angles, Gogel (1961) concludes. However, Cabe et al (in preparation) report  touch is quite good at estimating  the distance to the intersection of two hand-held sticks. The pairs of sticks Cabe et al tested were positioned one to the left of the median plane, one to the right. Distances of intersections in front were estimated  more accurately than ones to the side, which were underestimated more as the intersection departed from straight ahead ( possibly because the angle of intersection diminished considerably, and when it was tiny it was overestimated). 

Tactile pictures: Eriksson's history

    Eriksson (1998) argues that the discussion of the relation between vision and touch from the Seventeenth century onwards likely influenced many educators of the blind. She has written about the manufacture of tactile displays for the blind from 1784 to 1940. Many of these displays were pictures in raised form, using solid lines, or dotted lines, or bas-reliefs. In the writings of the pedagogues Eriksson surveyed ideas about touch as a spatial sense are evident. But once again the claims about touch offer imperfect parallels with vision.

     Like Diderot and Locke, educators in France, Germany and Britain stressed that motion was needed for tactile perception of shape, size and distance. Consider a few discussed by Eriksson. In France, in the early Nineteenth century, Guilli e noted the blind can only have successive ideas of the objects they touch. But then he added that they can perform a secondary task, to bring these impressions together, and perhaps a third task in order to compare impressions.  In Britain, in the middle of the Nineteenth century,  Fowler discussed passing our fingers over a table slowly or quickly. The rate of motion and the time taken indicate the size of the table. We only need a few contacts to get the impression of a continuous table surface. In Germany, at the turn of the century, Heller  wrote that the mobility of the hand is a key condition for the development of the blind person's sense of direction. He wrote about blind pupils examining tactile displays. using active exploration. He also described a subject reporting imagining an index finger moving from one point on a tactile display to another. But Heller then argued that it would be in vain to try to form a total impression of a large object close to us via touch. To form just such an impression we have to remove the object to a larger distance, mentally, and somehow reduce it. It is impossible to get a simultaneous impression from a large object -- only a small object will permit this in touch, Heller hypothesized.

      Eriksson found tactile displays made by Martin Kunz (1847--1923) in institutes for the blind throughout Europe and North America. Kunz not only made many displays, he theorized about the abilities they called on. In one extraordinary conjecture, he claimed blind people often do not have a sense of distance. He argued the sense of distance does not develop if one becomes blind, notes Eriksson (1998, page 77).  In a curious turn of events, Kunz proposed the blind could have a well-developed sense of location, indicating the places of many objects, but this was by no means the same as a sense of distance.    

     The teachers who worked with the blind, and the manufacturers and designers who prepared tactile pictures, were likely deeply influenced by the debates among philosophers and education theorists. Gall (in 1837, cited in Eriksson, page 92) , a Scottish clergyman who prepared tactile pictures, wrote  "The Blind can feel the shape of any image they can handle; but not having any idea of perspective, it is only an outline which can be perceived". Later, in the century Martin Kunz , discussing tactile pictures, wrote that for the finger there is no perspective, Eriksson reports. 

     Perspective and directions from the observer's vantage point are often at issue, but in disjointed ways, in the instructions accompanying pictures for the blind. Consider the caption for a picture of hot-air balloons in a picture book for the blind, published by the National Institute for the Blind, London, in the 1920s. Eriksson (page 127--128) quotes the caption: "Imagine the rectangular border represents an open window.... Inside the border represents open space... Stretch out your arm through the window... and move it about in every direction.... If you could stretch out your arm till it was five or six hundred yards long, you would be able to touch the nearest balloon...... a second balloon is shown....It is really the same size as the first, but being much further away it has the appearance of being  smaller and fainter...." ( I have abbreviated the caption.)

      Directions from the observer's vantage point are adumbrated usefully in this caption. But inconsistent use of spatial terms obscures the lesson. The caption tells us that a balloon lies in a certain direction. But then it alters its set of key terms. It does not spell out what aspects of direction are relevant. It changes its terms to appearance, size and faintness. As Lopes (1997, p. 438) writes, having "identified the picture surface with the visual field", the use of the significant term "direction" is over-ridden. The caption could have said the direction of the top of the nearby balloon is close to the direction of the top of the window frame. It could have added that the direction of the basket at the bottom of the balloon is close to the bottom of the window frame. The differences in direction result in the nearby balloon almost filling the window frame. The directions to the top and bottom of the frame, it could have said, are slightly wider apart than the directions to the top of the balloon and the basket. Then, the caption could have added, the more distant balloon has a small difference in direction between its top and its basket. It could have pointed out that as the balloons recede the differences in the directions diminish. 

    It is distinctly odd that most theories of touch argue spatial touch requires motion, in particular directions, in straight and curved paths, and stress that otherwise we have little except pressure in the finger and the resistance of surfaces, but then draw in their horns when talking about pictures. Often, theory of touch in concert with pictures describes "fingers", and  fails to entertain motion, direction, mobility, and any of the other degrees of freedom that give spatial touch its flexible modus vivendi and its information. Theorists opine that the blind can touch surfaces and imagine directions. They do not go on to say that blind people touching a picture can take some picture elements as telling us about directions from a vantage point. The result is interminably one-sided discussions.

      Lopes (1997) argues that the proper conclusion "to draw here is that perspectival perception is not unique to vision. It is part of any conception of space that enables us to move around our environment, and will be present in experiences in any sense modality that represents space. If perspective is spatial and not distinctively visual, then the argument that vision differs from touch because a component of its content is perspectival, characterized as shapes and sizes on a visual field, is unsound" (page 437).  He adds that we have made a mistake in interpreting vision as like a picture, and a picture as solely visual. We redoubled the error in interpreting images in the head as like pictures, and therefore as like vision. He continues "The error is compounded when, having postulated pictures in the head, we then explain pictures on the canvas by means of their alleged similarity to those postulated  mental pictures." (p.439). See also Costall (1990), and Cutting and Massironi (1998).

  Recent evidence

       The history of thought on touch, the blind and space is full of unfortunate assumptions. There is no a priori reason to insist touch is a non-spatial medium. Motion around an environment could well give a blind person an astute awareness of the relative distances between points, and their directions with respect to each other and the observer.

      In practice, how well do blind subjects know the relative distances between parts of a room? Here I will describe some of the key studies of the past decade. For an analysis of studies that paved the way to these reports see Millar (1994),  Kennedy (1993) and Kennedy, Gabias and Heller (1992).

     Haber, Haber, Levin and Hollyfield  (1993) tested 7  blind, highly mobile adults ( two of whom were congenitally blind) estimating the absolute distances between 10 objects in a familiar  room. The subject sat at one location in the room, surrounded by the target objects,  to make the estimates.  The estimates were all closely correlated to actual distance, regression analyses showed (all correlations within the range .84 to .99). There were no differences between early-blind and late-blind subjects. The two-dimensional map of distances and locations one could draw using the subjects' estimates closely matched the actual room, as an 88% scale model.  There was a single zero point for the estimates-- the observer's vantage point. 

       Haber et al compared the distance estimates of the blind to estimates made by sighted subjects, in the same room. The sighted subjects were familiar with the room. They found no significant differences between the estimates of the blind and the sighted, except that the sighted underestimated the distances slightly less (5%) than the blind (12%).  Haber et al asked how minor variations in the zero point for the blind compared to similar minor variations in the sighted. They found no significant differences.

     Instructively, Haber asked the sighted and the blind subjects about the 10 objects in a second condition. In this second condition, the subjects sat in a second room, and were asked about the remote room.  The results were the same for both the sighted and the blind: the original, now remote room space was reported as if it had shrunk by 30%!

     Haber et al were not sure of the reasons for the underestimations of distances, but whatever factors were involved they may be similar in the blind and the sighted. The major findings are the accuracy of the estimates and their close correlations with actual variations in distance.

    Haber et al compared the estimates of distances between pairs of objects that could be joined without intervening barriers, and pairs of objects that had intervening barriers. There were no significant differences between these distinct pairs, for either the blind or the sighted. Evidently, occlusion that requires a detour in travelling between objects did not impair distance judgments in the blind. 

      Haber et al.'s results are similar to those from Loomis et al (1993), comparing sighted and blind  adults on spatial navigation tasks. Landau, Spelke and Gleitman (1984, see also Landau and Gleitman, 1985;  Landau, 1991) tested blind children on their ability to find their way from one object to another in a room. She reported that if a preschool child learns about four objects A, B, C and D, by walking from A to B to C to D, thereafter the child will be able to walk from A to C, ignoring B. The specific route from A to C does not have to be taught. However, Lockman et al (1981) and Reiser et al (1980) find the more experience one has with travelling without sight the more accurate the response to spatial-layout tasks. 

       Wagner et al (1996) explored the connection between stepping motions and spatial judgments. Using a treadmill, they varied the rate at which a step changed the observer's spatial location. Subjects swiftly adopted the new coordination. The effect was perceptual, not a conscious correction, because there was an aftereffect once the normal coordination was restored.

       Morrongiello et al (1995) videorecorded blind and blindfolded-sighted children undertaking the Landau ABCD four-locations task. They coded the paths taken, accuracy of initial turns, closest positions and final positions relative to the target locations. They also devised a composite score to assess the efficiency of the path taken. The mean age of the blind children was about 7 years, with a range from 4 years 5 months to 9 years two months. They were all congenitally totally blind (that is, none had ever had sensitivity to light). The blind children performed like the sighted children on all the measures except accuracy of the final position, at which they were slightly worse.

       The children were also asked to draw some visible or tactile maps, e.g. showing the route from B to D. The proportion of sighted children  who drew correct maps of their routes  was about 20% for visible maps or tactile maps at age 4-5. At age 6-7, the proportions were still about 20% for visible maps, but had risen to 33% for tactile maps. At age 8-9, the proportions were 50% for visible maps and 58% for tactile maps. The majority of the children received the same score for both maps. Evidently, the map-making ability is modest in preschoolers, grows slowly, and tactile-map making is at least the equal of visible-map making. The tasks tap one ability, in the sighted at least, it seems likely.

     Morrongiello et al. note that "examining the nature of spatial representation is a challenging task because of ambiguity in the link between representations and behaviour." (p. 228). This was one reason their study used three tasks. On their easiest task--walking between the ABCD destinations--only a few of their youngest subjects accurately reproduced distance and angle information when pursuing novel routes. No three-year-old available to the investigators could be persuaded to undertake the task.  In contrast, the actual routes on which the subjects were trained, and reversals of the routes, 

were executed summarily.

Measures of spatial ability

      Were the youngest children discussed in Morrongiello et al (1995) ever reluctant to run novel routes for fear of novel obstacles? Could they have pointed in the right direction (from B to D)? Standing at B could they have told whether a sounding brass or a tinkling symbol was located at D?  Some response measures reveal more about basic spatial abilities than others (Millar, 1985, 1994).

       Haber, Haber, Penningroth, Novak and Radgowski (1993)  tested body postures, including the use of body extensions such as pointers, and external pointers, as ways of indicating the direction of sound sources. The subjects were twenty blind adults.  The sources occupied a semicircle from the extreme left, through straight ahead to the extreme right. The body parts pointed at the target included the index finger of an outstretched arm, the observer's nose and the chest, to which a pointer was affixed at right angles. The observers also tried to point hand-held rods at the target. All these measures were superior to methods that used  external pointers on their own bases, like rotating a dial. The least accurate methods involved drawing or offering a verbal description using clockface labels such as noon for extreme left, 3pm for straight ahead and 6pm for extreme right. Generally, the body part  or hand-held rod methods were less variable as well as more accurate.

          Subjects who travelled less were particularly poor at the more difficult tasks with high variance such as the clock-face or drawing tasks. On the other tasks, the groups performed alike. Travel skills and experience may affect performance on relatively indirect measures of space perception, and have little bearing on basic body orientation to the local environment, it seems (Millar, 1985).

Triangular routes and convergence

    Worchel (1950) argued blind people guided along a right-angled triangle route from origin A to the right angle at B and then to a terminus C often cannot then walk back from C to A along the hypoteneuse. This extreme claim  is surely false (Klatzky et al, 1990; Millar, 1994). Errors will be made, but the basic principle that the triangle exists in a two dimensional space, with directions  and spatial extents, is understood by most blind people intuitively, provided they have no loss other than sight (Kennedy and Campbell, 1985). That is, the blind person knows that extents subtend different angles at our vantage point as they recede: Directions to the ends of the extent converge.

   The principle of convergence applies in both the horizontal plane and the vertical plane. We can point to the bottom and top of nearby and distant trees or to the gaps between columns that are near and far. Both entail convergence. If we point upwards to a bird flying away from above our head we will find the direction changes swiftly at first and then more slowly. Likewise, if we point to a  mouse running away from between our feet we find the directions change swiftly at first and then slowly. Generally, blind people understand this (Kennedy, 1993). The arms  pointing up to the bird and down to the mouse converge. They converge more and more slowly as the imagined distances of the bird and mouse increase. Both arms stop at the horizontal. When they stop they are pointing at the horizon.

       The principle of convergence applies to all dimensions of space. It applies to small scales--the tabletop or manipulative space-- and to larger scales such as the domestic or room-sized space and the ambulatory space of say a few hours walk (Millar, 1994).

      The space to which convergence applies consists of directions (angles) and distances from a vantage point. Do we confuse these two? Klatzky (1999) tested angle and distance errors in touch, finding they were often unrelated.

       Klatzky (1999) looked for the origins of systematic errors made when blindfolded subjects attempt to walk two legs of a triangle and then to indicate where the origin is from the final destination.  She asked whether errors are introduced when subjects attempt to imagine their walk and final destination has been displaced by a rotation about the terminus, or by displacement to one side.  She found more errors for distances than directions (angles).  Interestingly, she suggests some errors are due to subjects emphasing body-centred coordinates: information about space is referred to the observer's vantage point, and any errors arise from misestimating the body's location and where it is facing.    

Other errors, she argues, come from use of objects or landmarks as the primary basis for spatial understanding: object-centred coordinates.  She contends body-centred errors are more evident before subjects entertain imagined rotations or displacements. But object-centered errors are more common after imagined rotations.           

     Despite the presence of some errors, the subjects tested by Haber et al, and by Klatzky, performed quite consistently in spatial tasks, whether they were in the target location, or imagining locations, directions or displacements.  

 Pictures and the field of directions from a vantage point

        When we use a picture, the very medium of picturing has distinctive implications (Pierantoni, 1986). We can look at a picture pinned to a wall, and take it to show a person in profile standing upright. If we take the picture off the wall and lay it on our desk it will be horizontal. Does this make the man appear to be lying down? Well, no. We take the picture to be a medium whose vertical or horizontal orientation is not relevant to the orientation of the profile (Kennedy, 1993). The profile remains that of a person standing erect.

     Consider the picture lying on the table. The profile can be turned so that its nose points toward the observer, or away from the observer.  Then blind and sighted subjects can be asked when the profile seems to be facing down to the ground and when it faces up to the sky. In a demonstration I recently conducted,  LT, a totally blind man, whose blindness had an early onset due to retinitis pigmentosa, reported when the nose (in a horizontal picture, made of raised lines) was pointing towards him the profile seemed to be facing the ground. When the nose was pointing away from him, the profile was facing up to the sky.

       The  profile demonstration suggests people use a field of directions around a vantage point centred on their heads to interpret tactile pictures. What is higher in the field of directions is taken as representing objects that are higher in the vertical plane in the world. As another example of this use of a field of directions, consider the text on this page. Notice that the text can be read while resting on the table, vertical as if fixed to a wall, or held overhead  as though fixed to a ceiling. Sometimes the text letters have their tops farther from the observer than their bottoms, as is the case when they are on the table. Sometimes the tops of the letters are closer to the observer than their bottoms, as is the case when they are on the ceiling. But in both cases the tops of the letters are higher in the field of directions than the bottoms. And in both cases the letters appear upright (Mirabella and Kennedy, in press).

        If someone draws a U on our hand as it lies palm-up on a table in front of us, and we have our eyes closed, we can read it as a U. If we turn our forearm to bring the hand in front of our tummy, palm up, thumb pointing away from us, still resting on the table, the same shape on our skin will usually  be read as a C. This suggests the vantage point at our head controls the apparent shape and identity of the form traced on our skin. What is higher in the field of directions from that vantage point is the top of the letter.

        Many investigators have used "cutaneous perception" tasks in which a letter is drawn on the skin, with a blunt stylus, on sites scattered around the observer's body. Subjects seem to entertain a variety of vantage points in three dimensions. Natsoulas (1966) drew b, d, p and q on the subject's forehead.  Subjects generally used one of two vantage points, the results suggested. One had an internal location. The subject behaved as if "looking"  from a position within their head. The second had an external location, like a "disembodied eye" (Concoran, 1977). The subject behaved as if looking from the experimenter's position standing in front of the subject. Often subjects could readily change from one apparent vantage point to another. Similar flexibility of vantage points in assessing forms drawn on the skin is evident in studies on blind subjects (Shimojo, Sasaki, Parsons and Torii, 1989).  Further, the confusions subjects report are between p and q, or b and d. Subjects maintain the vertical orientation of forms on the forehead.

         The studies on profiles and letter forms suggest what is high in the field of directions around a vantage point is "the top" of the object, whether the form is on a table, a wall, a palm or a forehead. But they also suggest vantage points that are disembodied can move horizontally to some extent. It is also relatively easy to imagine being above an array of objects, with "a bird's eye view", as tactile maps often require us to do. But rotation of a vantage point and its field of directions from erect to tilted or inverted (a roll, as opposed to yaw or pitch) is likely more problematic for the observer. Also, rotation of tactile maps in the plane (yaw) by 180 degrees , like rotation of visual maps on a table, often confuses observers about left and right.

Arcs and elevators

    Another kind of rotation moves the object in a vertical arc around the vantage point (pitch), from a table top to the ceiling. Objects with conventional fronts such as raised letters and pictures can move in an arc like this and remain upright and easily legible.  Their orientation is invariant across this arc. They are always "directed to" the vantage point, and their top is "upwards" in our field of directions. Orientation and direction are related concepts, by definition. A tactile vantage point is a location from which orientation or direction can be established.  

      Some objects maintain their orientation in an "elevator" transformation but not an arc. Cups would spill their contents if moved in an arc, but  elevated vertically from a desktop to a shelf they keep their usefulness (that is, their "affordances", Gibson, 1979). They are "directed upwards", as is an elevator. The elevator transformation is often independent of our vantage point. The arc has a vantage point as its origin. The elevator transformation, however, can take an object past our vantage point. The cup varies in its accessability, for it can be too low or too high for our reach, from a given vantage point. Also, when we imagine a bird's eye view on a room we have often raised our vantage point as if on an elevator, while maintaining what is to our front and back, left and right, and likely this is straightforward for many blind and sighted observers. (A birds's eye view may be what Haber et  al.'s observers undertook when asked to report distances in one room while sitting in another room.) When we move left or right (slide), or to the front or back (to-and-fro) we maintain what is higher or lower than us. These motions preserve direction--what is straight ahead on the horizon--tactually and visually. Again, these motions of the vantage point in the plane are familiar to the blind and the sighted, surely.

       The elevator transformation does not invert an object. But a raise reveals the tops of solid objects (and lowering ourselves reveals the bottom). The arc transformation does not invert a raised picture when it moves from a table to overhead.  But if the display passes overhead and starts downwards again ( behind us, for example) it inverts while following the arc. Both inverted solid objects and pictures likely are relatively hard to identify in touch. A u-shape becomes an n-shape. Cookie cutters that are inverted are often hard to identify in touch. Busts that are inverted are too, informal class demonstrations suggest.

        In sum, orientation at a tactile vantage point is often dependent on a field of directions from that point. The vantage point may be disembodied to some extent, in a variety of locations in the three dimensional space around the observer, so the locus is moved forward or back, displaced to one side or elevated. As Klatzky pointed out, it can rotate in the plane, remaining upright, and still be quite usable. Invariant upright orientations of targets facilitate tactile perception of form, it is likely.  

Borders, media and foreground in vision and touch   

    A disembodied vantage point from which we take directions to a few objects can be envisaged relatively freely. It can be where we intend to go, before we actually go there. But much of our tactile exploration aims to use richer or fuller environments, with real edges of actual surfaces and textures of surfaces being examined to determine where we are in relation to objects. Similarly much of vision involves scanning borders of various kinds to get to know the vista around us and our place in it.  

      How do vision and touch use surfaces, textures and borders, and the media of perception,  to recognize where we are? Rubin (1915) defined a kind of foreground and  background perceived at a contour or line as figure and ground. He might have added that our vantage point is always indicated by what is foreground and background.

     A line or contour can be defined by colour or luminance borders in vision. What operates in a similar vein in touch? Touch reveals borders of surfaces as changes in resistance, notably, and thermal and wetness or friction properties secondarily. We can certainly detect one surface foregrounded on a background in touch. One example is we can feel that a sheet of newspaper is lying on a rubber mat such as a mouse pad. Further, we can feel one section of our newspaper is lying on and partially covering another, and that on another, and that on yet another etc. We feel a set of foregrounds and backgrounds, specifying our vantage point in front of the most foregrounded sheet.

      A flat surface offers resistance, ending at a border, giving way to another kind of resistance. We can also feel two similar surfaces meeting at a corner, with change in slant (where the top of our desk meets a wall, say). The corner encloses our vantage point, if the corner is concave. We are outside the corner, we feel, if it is convex, like the corner where the top of the desk meets the side). A roofline of a model house is a tangible border (Heller et al, 1995). A rounded object seems to offer a definite border between front and back to vision, and it also offers a border to touch if we take it to be in front of a particular vantage point, say one from which we are reaching. To that vantage point, the object presents a front, a back and a clear division between the two: an occluding boundary of a rounded surface.

       Raised line drawings of objects showing  corners and boundaries of rounded surfaces by lines are recognized in similar fashion by blind and blindfolded sighted 8--13 year olds (D'Angiulli et al, 1998). The blind children were congenitally totally blind. The performances of the sighted and the blind children were highly correlated. The scores of the blind and sighted children exploring the displays actively, with no external guidance, were correlated .81. 

      Perhaps some aspects of textures operate similarly in vision and touch. A sheet of paper and a rubber mouse pad have tangible textures, just as they have visual textures. Variation in tactile texture is readily used as an indicator of  the slopes of surfaces, though it may be more immediately understandable in vision (Holmes et al. 1998). Holmes et al.  presented blindfolded subjects with texture patterns with distinct gradients, say dense at the top and gradually expanding to sparse at the bottom. In a series of the patterns, the rate at which the texture expanded was varied. Subjects attempted to match the texture patterns to panels that sloped from vertical to nearly horizontal. Subjects scaled the magnitudes of the physical slants of the panels to suitable texture gradients, provided the extremes of both were evident, and some opportunities for learning were offered.  Early blind subjects performed like sighted subjects.   

     Vision also uses changes in the visibility of textures to indicate what is foreground and background. Is there an equivalent at a tactile vantage point? Consider the possible cases. "Accretion" of texture is texture becoming evident in the optic array at our vantage point. "Deletion"  is the reverse.  Gibson (1979) described the following case. A mousepad being pulled our from under a foreground newspaper covering it is revealed by its texture visibly accreting at its common border with the cover. If the pad is progressively covered by the foreground newspaper being pushed over on top of it, its texture is optically deleted.

       Is there a tactile equivalent for Gibson's case? Certainly we can feel the pad's texture moving out from under a cover sheet, accreting in touch, indicating it was behind the cover and is now being exposed. And we also can feel a cover sheet moving over a pad and concealing its texture.

      Besides Gibson's case, vision also has texture accretion and deletion  that occurs because of illumination, and a rather different kind that occurs when the medium for light transmission has elements swirling in it. A searchlight playing over a prison wall in a 1930's black-and-white movie often reveals the wall's texture of bricks accreting at the leading edge of the pool of light. The texture at the trailing edge is deleted, falling into invisibility. The accretion and deletion specify a single surface, the wall, as foreground, with no other surface as background. Shadows racing over the wall have similar effects.

       A searchlight beam passing through empty air is invisible till it hits a reflector (the wall, in this prison movie). To add to our little drama, consider the night air is full of twirling snowflakes, gently falling. The snow may be invisible till it enters the searchlight beam (optically accretes). It passes into invisibility again when it falls through (deletes). The accretion specifies the texture elements are in the foreground. So too does the deletion. The elements in turn specify a volume of space, which we often informally call "the searchlight beam".

      Snow in daylight can provide another case of accretion and deletion. If our vantage point is in front of a dark surface, such as a dark hill set in a snow-covered field, silhouetted against a white sky, falling snow may only be visible when it has the hill as background. The snow can be invisible against the sky. It only comes into view when it is in front of the hill from our vantage point. It is invisible again when it is against the snow-covered field. From our vantage point, the accretion of snow at the brow of the hill specifies a volume of elements in front of the dark border, and the deletion  at the border at the base of the hill does the same.  

       Touch is hardly as richly endowed as vision with cases of accretion and deletion. Touch generally operates more like an eye scanning over elements on a fixed surface. We often look at an element, allow it to fall into our periphery, and then gaze away so the element is too far to the side to be seen. Similarly, when touch runs a hand over a surface, texture elements come onto our skin, pass across the skin, and then fall too far behind to be in contact.  However, we are not totally without some tactile media. In addition to direct contact, touch can use a rod as an intervening medium to feel the roughness of a surface (Lederman, 1982). Textural roughness in the surface we are stroking causes vibrations to arise in the rod , and to vary as we pass over different surfaces. Further, touch can use covering material as a medium to "palpate" surfaces beyond the cover. The media through while we feel add their own roughnesses, vibrations and thermal properties we have to discount to detect the distal target. When we drive a car or ride a bike, the car or bike will tell us about the rough road we are careering over. They also add their own vibrations. That is, we often have to realize what arises because of media between our vantage point and the target. Touch isolates vibrations in a medley and discerns which are from sources near to us and which are from afar.

     Like vision, touch offers accretion and deletion of texture from two surfaces in close proximity to each other. The event specifies foreground, background and our vantage point. Unlike vision, touch's use of media does not involve accretion and deletion due to a secondary source of energy, such as a searchlight, or texture in a medium accreting and deleting.

 Coda

    There are many ways in which vantage points arise in touch, in tactile tasks and in space perception served by touch, because touch deals with direction. Some show us limitations of the observer. Some are disembodied, like intended locations. Some suggest practical ways of using pictures, using outlines for edges, and directions of elements for directions of referents.  Some are especially easy to use when they allow invariants in the apparent orientation of moving objects. Some suggest perspective is present in touch. The parallels with vision are extensive, as consideration of direction shows, but not complete, as consideration of accretion and deletion of texture reveals.  

    I may have left the impression that vision's vantage point is plain at all times. This was handy to introduce my topic. But I should not let it go at that. Our visual impression that we have a single vantage point optically is deceptive. We look with two eyes, not one. But objects often seem to lie in one visual direction. The laws of visual direction, and the conditions under which we will seem to have one visual vantage point, are now a subject for animated debate (Ono and Mapp, 1995).  It seems fitting that we should realize we could try to discover the laws of vantage points in touch just as fresh views of visual vantage points are being explored.

  References