20. Tactile Interaction
The following chapter describes a variety of ways in which Tactile Interaction may be used to enhance the human computer interface, i.e. the design of interactive products. Opening with a general discussion on a broad range of potential applications for Tactile Interaction, the chapter quickly moves onto to consider the key physical, perceptual and technological issues that are likely to influence the ways in which we can tap into this potentially rich source of interaction. A broad range of research topics are visited along the way with suggestions being offered for the fundamental design principles that should be considered within any interface that seeks to harness some level of tactile interaction.
Let us consider why we might be interested in exploring tactile interaction within the design process at all? After all, is it not generally the visual channel that features most prominently within any given interface? In terms of display and feedback, this is often true; as users we perhaps expect to first see our available controls and options before we begin to interact with them, and similarly we probably expect to see the results of any actions we take. Interface design has long embraced the notion of offering additional methods of reinforcement where possible, with particular emphasis on the use of auditory feedback. So, any given action is likely to be reinforced through a combination of visual and auditory feedback, indeed, in an ever-expanding market for music and audio software, the feedback may be primarily auditory anyway; the music or sound itself. This undoubtedly powerful combination enables rich forms of both display and feedback, but the interaction that takes us from one to the other will generally be supported through touch.
The manner in which we select and manipulate these various physical and virtual objects will be through a combination of movement and touch or haptic interaction. This is perhaps where things begin to become interesting for though there is a considerable body of research and well documented 'good' practice on interface-design in terms of visual and auditory feedback, there is still very little on the specifics of 'good' design for haptic interaction. So, a first answer to our opening question could be that tactile interaction is perhaps being undervalued in terms of the potential source of feedback that it might offer. The primary focus has perhaps been too firmly fixed on the 'doing' rather than the 'receiving'. So, in a world where visual displays are already quite cluttered, off-loading some of that information to the auditory channel could be of real benefit, but it might also be that some of this feedback is more immediately meaningful to our sense of touch. By way of example, the action of selecting and positioning a virtual on-screen fader might simply feel more meaningful than it can ever sound.
This argument alone is quite compelling, but let us now consider the numerous situations and environments in which a visually dominant display is either impractical or impossible. Perhaps the most immediately apparent example would be the design of non-visual interfaces for users who are blind. There are many issues to consider here, and some of these will surface later in this chapter. However, the key point for now is that the graphical user interface (GUI) will need to be translated and communicated effectively using non-visual means. There are some aids to draw upon here, screen-readers can use synthetic speech to read on-screen text or to describe a structural layout — but might it not be faster and more informing to display structural components in such a way that the user can 'feel' the controls? So, there are extraordinary needs for some computer users where a significant move away from a visually dominant interface is likely to be of benefit. However, there are all manner of occasions where the same is likely to be true for users who do not have additional individual needs. When controlling a vehicle, for example, or operating specialist machinery, it might simply be 'safer' to be able to maintain near-constant visual contact with our surroundings. Indeed, the location of controls for in-car entertainment systems is now often found on the steering column — within easy reach and with no requirement to look away from the road. In some situations, the immediate environment may not offer enough light to easily see what is happening. At the extreme end of these scenarios, perhaps an environment is currently under emergency lighting, and at the less safety-critical end, perhaps the living room is a little too dark to easily control the home-cinema by remote control.
So, there is a broad range of reasons why, as designers, we might wish to explore non-visual modes of communication. Our interest within the context of this chapter, though, is with tactile interaction, and later we will consider the various technologies (existing and emerging) that can be drawn upon to make this happen. However, as with all other forms of interaction and feedback there are physical and perceptual limits and boundaries that will influence how effective any new design might be.
20.2 The Psychology of Touch
To begin to appreciate how tactual interaction might be successfully integrated into the human-computer interface, it is essential to understand how the human body retrieves and processes information about its immediate surroundings. This happens at two levels: physical and perceptual. At the physical level, our peripheral nervous systems gather information using a number of different nerve types each of which is sensitive to a particular type of stimulus. All the information gathered by the peripheral nervous system is conveyed through the central nervous system to its ultimate focal point: the brain. It is here that the information is interpreted and then acted upon, and it is this process of interpretation that constitutes the perceptual level. For the purposes of this chapter, we shall avoid any further discussion on the physiological make-up of the human nervous system, but for anyone considering exploring tactile interaction within a design context, this really should feature as part of your recommended further reading.
Loomis and Lederman (1986) provide a useful overview of the three aspects of interpreting information through touch that could be referred to collectively as tactual perception. They state that there are two fundamental and distinct senses that together provide us with a sense of touch: the cutaneous sense and kinesthesis. The cutaneous sense provides an awareness of the stimulation of the receptors within the skin, whereas the kinesthetic sense provides an awareness of the relative positioning of the body (head, torso, limbs etc.). Perception that involves one or more of these can be regarded as tactual perception and there are, therefore, three forms of such perception.
Tactile perception is solely dependent upon variations in cutaneous stimulation by such actions as tracing a pattern upon an individual's skin. Tactile perception alone means that the individual in question must be static; otherwise the kinesthetic sense will be incorporated.
Kinesthetic perception is concerned with variations in kinesthetic stimulation. However, tactual perception without contribution from the cutaneous sense can only really be achieved under contrived circumstances such as using anaesthetic to suppress the cutaneous contribution.
Haptic perception is the form of tactual perception that involves both tactile and kinesthetic perception, and it is this that we use on an everyday basis to explore and understand our surroundings using touch.
The level of control that an individual has at any given time over the collection of information by either the cutaneous or kinesthetic sense leads to the following five tactual modes, where for the first three modes there is no control. Indeed, it can be seen from these definitions that of the five modes only the last, active haptic perception, is likely to be of real significance within the design of interfaces that employ tactile interaction.
- Tactile perception - Cutaneous information alone.
- Passive kinesthetic perception - Afferent kinesthesis.
- Passive haptic perception - Cutaneous information and afferent kinesthesis.
- Active kinesthetic perception - Afferent kinesthesis and efference copy.
- Active haptic perception - Cutaneous information, afferent kinesthesis and efference copy.
20.3 The Practicalities of Tactile Interaction
Ultimately, the design of any human-computer interface that incorporates tactile interaction is bound to share many of the same considerations over construction that need to be made when designing, for example, tactile diagrams, and there is a considerable body of research within this area that we can turn to. Before we consider some of these perceptual properties in depth, let us first quickly consider the relationship between touch and sight.
20.3.1 Vision versus Touch
Our general reliance on vision suggests that if a conflict were to arise between our sense of touch and our sense of sight, it would be the visual aspect that becomes dominant. Such conflict of sensory information formed the basis for research by Rock and Victor (1964) who used an optical cylinder to give a subject group the visual impression that a solid square object was really rectangular. To the subjects in the experiment, the tactual impression was unchanged. Subjects who were presented with such conflicts generally made similar judgements on size and shape as those members of a control group who were only presented with visual information.
However, McDonnell and Duffett (1972) suggest that Rock and Victor's experiment may have had methodological failings that could have produced a bias in favour of vision. In their reworking of the original experiment, subjects were asked to examine blocks of wood on the top surface of a table. To examine the lower part of the blocks, the subjects had to feel beneath the table. The impression that the subjects were given was that the blocks went through the table, whereas there were really two blocks being used to give that impression. In effect, each block that the subjects examined was a pair of blocks where the width of the wood and the protrusion on either side of the table was the same but the lengths could be different. Five pairs of blocks were used with the discrepancy ratios of 1:1, 1.29:1, 1.67:1, 2.2:1 and 3:1. After examining each block, the subjects would have to choose the best match from a group of comparison blocks that had no length discrepancy. The mean scores for the group showed that the subjects had chosen a comparison block that was a compromise between their visual and tactual impression. However, closer examination showed that the subjects had made their choice conform to either the visual or tactual impression and that there was considerable diversity of response pattern such that visual dominance could not be regarded as significant.
Heller (1992) has also shown that Rock and Victor's original findings should not have been applied so generally as to state that vision will always be the more dominant sense when there is conflict between touch and vision. In Heller's experiment, subjects were required to explore the embossed letters p, q, b, d, w and m whilst looking at them in a mirror. A subject would, for example, be exploring the letter w whilst looking at the letter m. When asked to identify the letters, there were a wide range of responses where the majority relied on touch, some on a compromise between the two and only one relied on vision.
It seems, therefore, that visual dominance versus tactual dominance should not be thought of as a dichotomy as there is evidence of compromise between the two senses when they are in conflict. In addition, this level of compromise is likely to be highly individual and will also be affected by a bias towards the suitability of one or both senses to the nature of the task.
20.3.2 Visual to Tactile Mapping
Klatzky and Lederman (1987) argues that many tactile diagrams will probably have problems due to the limiting nature of spatial resolution describing, for example, line drawings that are seen well might not be felt that well because the scale simply becomes too small for such limited bandwidth. Klatzky and Lederman also suggest that a fundamentally flawed model of haptic processing is often employed in the design of tactile graphics displays. Referred to as the image mediation model, the "hand functions like a roving eye that is badly in need of glasses" (Klatzky and Lederman 1987). The assumption made is that a spatial image is produced which is equivalent to one produced using vision. However, this image is affected by factors such as the low resolution of the haptic sensors along with demand placed upon memory due to the nature of exploring an 'image' over time. This image is then passed to the visual system's image interpreters resulting in a mental-image which perhaps seems like the original diagram was examined visually by someone with poor eyesight.
In contrast to this model, Klatzky and Lederman suggest that the haptic system has its own perceptual system and interpretive processors. Haptics and vision are simply different ways of perceiving, although at a much higher cognitive level there may be some convergence of the haptic system into the visual. Perhaps the most significant suggestion from Klatzky and Lederman is that the haptic system is not an efficient mediator of images. By way of example, it is described that if an individual is asked to think of looking at a cat, he/she is likely to visualise the shape of its body and perhaps the colour and pattern of its fur. However, if the same person is then asked to think of touching a cat, a very different image will result; the softness of its fur and perhaps the warmth of its body.
During studies by Klatzky and Lederman into exploration of three-dimensional objects, a series of methods for exploration were recorded. Of particular interest was that substance dimensions (e.g. hardness and texture) can be extracted quickly and reliably, whereas structural information is extracted slowly and is error prone. If it is assumed that the haptic system will favour encoding mechanisms which produce maximum return for minimum effort, then it would be expected that an economical system would favour substance based exploration and encoding. One particular study involved subjects sorting objects that varied along four dimensions: hardness, surface-roughness, size and shape. There were three variations for each dimension, and all possible combinations were covered within the whole group. The subjects were simply asked to place objects that were similar into a common bin. By examining the objects within each bin, the dimension that had proved most salient could be established. For example, had all the rough objects been in one bin, the medium rough in another and the smooth in another, then surface texture would have been the most salient dimension.
The results varied depending on the exploratory conditions which were allowed such that when denied sight, the subjects would prefer the substance dimensions of either hardness or texture. This pattern was unchanged when similarity was defined as how objects 'felt', but there was a large shift towards reliance on structural dimensions when similarity was defined in terms of objects being similar 'visually'. Finally, when the subjects were allowed to see the objects, the structural dimensions were the strongest again. Although these studies were carried out using three-dimensional objects, the findings might be just as valid within tactile diagram design although only contour and texture would apply. In particular, if a display is to be partly visual but also reinforced with some tactile elements, Klatzky and Lederman's findings might help establish how best to distribute the information between the visual and haptic channels.
20.3.3 Line Symbols
Perhaps the simplest tactile object is that which represents the drawing primitive of a line. Straight and curved lines could be considered as the building blocks of most graphical representations, and graphs, maps, and other diagrams can rely heavily on the use of lines within their construction. There are two issues that quickly become apparent when using raised lines as tactile substitutes for visual ones:
- How easily can they be traced?
- How easily can different line widths be distinguished from one another?
On the traceability of lines, a comparison study has been made by Bentzen and Peck (1979) which addressed the issue of which styles of lines are easiest to follow by tracing with a finger. Four line styles were used: single continuous (smooth), double continuous, single dotted (rough) and double dotted. These were chosen because they seem to be accepted as the four most commonly used, but without any supporting evidence as to which is superior. Other than identifying which lines are generally the easiest to trace, Bentzen was also interested in how tracing can be affected within two particular scenarios:
Two displays were created using embossed plastic sheets. A simple display used all four line types, without intersection, which all included a right angle, an obtuse angle and an acute angle and a semicircle of 1.5in. These were connected using three straight-line sections of 3in and three of 1.5in. The complex display had the four line styles using the same tracing components, but each line intersected the other three at some point.
The conclusions reached were that the performance of rough against smooth lines is not significantly different and is therefore not a real design issue in the use of tactile diagrams. Single lines, rough or smooth, are preferable to double lines (0.25in apart) in tactile displays that do not have intersecting lines. No real conclusion was found with double over single lines, other than double did perform better than single at intersections. This, however, was believed to be particular to the design of the displays used and was therefore not significant at this stage. One final observation was that a single narrow line intersected by a double is an undesirable feature.
An experiment by Lederman and Campbell (1983) explored the use of raised lines with tangible graph displays for blind people. Four different methods of presenting the graphs were used.
Three line styles were used: smooth, large-dotted and small-dotted. From a traceability perspective, the results were encouraging, but with graphs where all three line-types were present in close proximity, a significantly longer time was taken to perform tasks and these were more likely to be inaccurate. Some notes by the investigators mention that when initially presented with a graph, a subject would tend to explore its general format such as major axes, symbols and labels. Sandpaper squares at the corners were used to great effect in that by checking the squares and major axes, a broad sweep of the hand allowed the subject to become swiftly familiar with the graph's dimensions and proportions. Individual differences were observed as some people would use both hands to explore, whilst others would prefer to keep one hand on the origin as a point of reference. It also appeared that both the no-grid and grid-on-graph formats were preferred to the other options; both seemed to be equally easy to use.
One other exploration of the factors that affect the discriminability of tactile lines concerned the perception of one line as being wider than another (Berla and Murr 1975). In essence, five standard line thicknesses were each compared with a set of lines half of which were narrower and the rest thicker. The five standard line thicknesses used were 0.1, 0.15, 0.2, 0.25 and 0.3cm wide. The 0.1cm line had six progressively narrower lines and six wider, in steps of 0.01cm, whereas all the other lines used steps of 0.013cm. Each line was approximately 0.64cm in height. When presented with one standard line and one variable, the subject had to indicate which was widest. The results showed that the percentage of extra width that was required for an individual to perceive a line as wider than the standard decreased as the standard increased. For example, the thinnest standard line (0.1cm) needed to be 20% to 48% wider, whereas the thickest line (0.3cm) needed to be 11% to 27% thicker. However, this trend did not continue without finish. When the standard lines included thicknesses of 0.64cm and 1.27cm, performance began to diminish. This was attributed to the line width beginning to extend beyond the width of an average fingertip with the result being that both sides (or edges) to the line could not easily be perceived. As the results stand, though, there is a very useful set of highly distinguishable line widths that can safely be employed where difference in line width will play a significant part in a tactile display.
Nolan and Morris (1971) suggested that only around eight tactile linear symbols could be in use at any given time before similarities would occur and that there was likely to only be a maximum of ten. A study by James and Gill (1975) showed there to be ten such distinguishable linear symbols, but they were unable to progress beyond Nolan and Morris's upper limit.
20.3.4 Point Symbols
Point symbols are probably best described as those symbols which are designed to be explored with only minimal movement of the fingertip. An important aspect of the use of such symbols is how well they can be perceived in contrast to the background. This is commonly referred to as the figure-ground problem. One aspect of this is whether raised symbols are easier to recognise than incised. An assumption that there is no particular difference between either has been tested (Nolan and Morris 1971) and in contrast to popular belief it seems that there is a significant difference. Braille readers were asked to trace a tactile symbol and then find that same symbol from a set of five possibilities. This was carried out using a set of raised figures and an equivalent set of incised figures. Nolan's result showed that, in fact, raised figures are significantly superior to incised figures. Although there was an increase in error rate with the incised symbols, the significant result was that there was a required increase in reading time of around 38%. This implies that most tactile symbols will be read faster and more accurately if raised figures are used.
Research by Lambert and Lederman (1989) into the relative legibility and meaningfulness of tactile map symbols has shown that there are three categories of tactile symbol to be used. Some symbols have inherent meanings such as a telephone shape, some can imply a meaning such as a pointed symbol for stop and some can have quite arbitrary meanings such as using a square to represent a washroom. The size of the symbols used was between 0.635cm and 1.27cm per side. These were based on previous observations and studies with blind people. The notable feature that Lederman discusses is that, although some people will instantly prefer or recognise certain symbols, the question of whether a symbol possesses inherent meaning is not paramount. With prolonged use, any symbol will begin to imply a specific meaning and to that effect there will be a learning overhead expected in the successful implementation of any tactile diagram.
20.3.5 Areal Symbols
The term areal symbol is used to describe those areas of a tactile diagram which use either a texture or a tactile pattern to communicate information. The term tactile-pattern perhaps implies an area covered with incised or raised symbols which can easily be perceived as being identifiable and distinguishable from similar patterns. However, Lederman (1982) makes the point that a fundamental aspect of tactile pattern perception is the perception of texture. She uses examples of the 'smoothness' of a baby's skin, 'roughness' of sandpaper, 'softness' of cashmere, 'rubberiness' of elastic and the 'slipperiness' of ice. Texture alone provides much of the tactual feedback that we need to decide whether one areal tactile pattern is highly distinguishable from other areal patterns.
Lederman discusses some of the known issues with the perception of texture outlining how she has studied the perception of roughness using aluminium plates with linear grooves cut into the surface. Her results have shown that the main factor affecting the perception of roughness is the width of the groove cuts in relation to their spacing. However, she has also shown that applied force is the next most significant factor, followed by hand speed. A greater applied force produces results that show a higher ratio of perceived roughness, and this can be shown to be true with a gradually decreased hand speed. These findings are significant in that they could play an important part in the design of appropriate tactile patterns to employ as areal symbols.
Loomis (1981) describes a series of limitations that affect tactile pattern perception: spatial resolution, interactions between stimuli more widely spaced than the resolution limit (interference), temporal resolution, perceptual integration and limited attention.
Spatial resolution - This is normally associated with the two-point limen test where the two points of a draftsman's compass are placed onto an individual's skin in close proximity to each other. The distance of interest is the threshold at which the two points are perceived as one. However, Loomis breaks spatial resolution into three further factors. Firstly, there is the mechanical property of the skin. When a point is placed on the skin, the gradient of skin deformation will be considerably less than that of the stimulus. Secondly, there is the property of mechanical wave spreading. When a point is stimulated, travelling waves are produced which, when picked up by other mechanoreceptors, result in a 'blurring' effect. Lastly, there is the neural organisation that means that spatial resolution is dependent upon:
- The density of mechanoreceptive units in a particular area of the skin.
- The size and sensitivity of these units.
- The number of neurons in the cortical projection areas that represent that same field.
Interactions between widely spaced stimuli - The two-point limen test is a clear example of this, but the focus is to assess the distance at which two stimuli are perceived as one. Having two such stimuli presented to body locations that are wide apart, such as separate hands, can lead to a phantom sensation at a point between the two. Another example of phantom sensations occurs when rapid stimuli are presented to one point and then immediately to another. The resulting feeling is of an evenly spaced series of sensations travelling from one location to the other. Cutaneous masking can also occur in that one stimulus can be completely masked by the presence of another much stronger stimulus.
Temporal resolution - The minimum period with which an individual can still clearly perceive two brief pulses as being separate. Loomis states that a number of experiments into this have arrived at estimates ranging from 2ms to 40ms.
Perceptual integration - The way in which all this information is used at a cortical stage even if it were to arrive there with no significant loss of detail. The suggestion is that, unlike with visual perception, the information from a stimulus pattern may fail to be recognised.
Limited attention - Even if the information was perceived without loss and integrated cortically, the individual might still not perceive the pattern as a result of insufficient attentional capacity. With the visual channel, it is widely acknowledged that people possess the ability to focus their attention when the quantity of information begins to exceed their processing capacity. The tactual channel does not seem as efficient for this form of focused attention.
When it comes to the practical application of tactile patterns as part of a display, Lederman and Kinch (1979) have provided a review of existing work in the field. A general conclusion that can be reached is that although there are around forty tactile patterns that can be found that are easy to recognise only up to eight can be used together. It is very difficult to find any more than eight patterns that can be used as a group without ambiguities creeping in. As an example, a typical poor choice of patterns would be the use of diagonal lines in opposing directions to signify separate functions. So, within the example set of twelve areal symbols shown earlier in Figure 5, there are already ambiguities beginning to creep in such that rapid distinction between certain pairs of symbols will be inhibited. A significant effect that Lederman discusses is the use of height to provide a filtering method within a tactile display. Symbols could be presented at one of three heights to provide an indication of significance, and through using a sweeping hand movement, this could be used as way of filtering out unwanted information.
20.3.6 Strategies for Exploration
Berla (1972) defines three problem areas in the use of tactile diagrams by blind people: legibility, organisation and strategies for exploration. Of these three problems, the last is of particular interest as it is concerned with how individuals adopt different methods for exploring diagrams. The results of an experiment in this area by Berla showed that some individuals will use a one-handed scanning strategy and others would use two. Those who used two hands proved to maintain a superior sense of orientation and position within the diagram than those who used only one hand. He suggests that this is the result of being able to use one hand as a reference point. The strategies that Berla identified can be described as:
Berla concludes that these strategies all have relative merits and disadvantages which, therefore, make it difficult to suggest the 'ideal' strategy to adopt. He also describes how the most appropriate approach might be to teach all of these strategies to people who will be working with such diagrams. They will then be able to apply the most suitable strategy to any given exploration task. Berla and Butterfield (1977) has also suggested that individuals will probably need to be trained how to use a particular type of tactile diagram before any significant level of success can be attained. In his studies, he showed that a student's performance at distinguishing between, and understanding, tactile symbols increased if the student was first trained in line tracing and distinctive features analysis. There can be, perhaps, an expectancy for an individual to simply understand such tactile diagrams without any prior experience. This is an area that Berla underlines as being important when intending to design and implement a tactile display.
20.3.7 Braille Symbols
The Braille system is a method widely used by people who are visually impaired to read and write. Braille was devised in 1825 by Louis Braille, a blind Frenchman. Each Braille character, or cell, is made up of six dot positions, arranged in a rectangle containing two columns of three dots each.
Although braille symbols are not an integral aspect of tactile diagram design, each cell is effectively a tactile symbol, and factors that affect the legibility and perception of such cells could therefore be of interest. One particular piece of research into braille and pattern perception studied whether braille cells are perceived as groups of individual dots or as outline shapes. A series of experiments by Millar (1985) used partial outlines to represent the letters that would normally be shown in braille. These were not tactile outlines of the ordinary printed letters; instead they were joined up versions of the braille cells. Millar's results showed that braille letters can be read and recognised significantly faster with cells of dots rather than joined lines. This implies that small tactile symbols might also benefit from being made of patterns of dots rather than outlines.
20.3.8 Design Principles for Tactile Interaction
A preliminary set of design principles for including tactile information within the human-computer interface has been proposed (Challis and Edwards 2000, Challis and Edwards 2000). Using music notation as an example of a complex graphical information type, an experimental system was developed that could enable blind music-learners to adopt a non-visual multi-modal approach to reading music notation. Static overlays were used in conjunction with a resistive touch-pad to create interactive pages where a user could 'feel' the structure of the page layout and then select appropriate levels of information retrieval using audio and synthetic speech. Three founding principles were adopted at the outset of the study covering consistency of mapping, the use of height and the use of static data with additional principles being identified within the study to address aspects such as size of display, visual-to-tactile mapping, simplicity of symbol design and perhaps most significantly empty space; this simply introduces areas where the user has no information on where or how to explore. One simple but key observation put forward was that tactile diagrams might not look that good and that the design is not likely to benefit from over-reliance on direct visual-to-tactile mapping.
20.4 Tactile Interaction in the Human Computer Interface
When employing technologies for harnessing tactile interaction, there are three broad approaches that might be considered: static tactile displays, dynamic tactile displays and force-feedback technology. There are clear merits and constraints associated with each such that the nature of interaction task will dictate which category will be most effective. Indeed, Oakley et al. (2000) have suggested definitions by which such different technologies can be categorised, this being based on the sensory system that is most affected by the interaction.
20.4.1 Static and Dynamic Tactile Displays
Where a permanent display is appropriate, an interactive information display, for example, or the controls on a dedicated device, then a static overlay superimposed upon an appropriate touch-responsive surface (e.g. resistive, capacitive, infrared, acoustic wave or force-sensing) will be practical whilst offering fine level of detail if required. In contrast, a dynamic display can offer increased flexibility as the interface is not 'tied' to a particular layout. Refreshable braille displays go some way towards offering this flexibility, but the size of the technology currently available does not lend itself to the creation of tactile symbols of a useful resolution.
For example, to achieve the sensation of tracing an unbroken line, a resolution of 20 dpi is required (Fricke and Bahring 1992, Fricke 1997). Even at this resolution, a display area of 20 inches by 15 inches will require 120,000 elements, and yet diagonal lines will still feel broken. Assuming that such technology was available to achieve a higher and more accurate resolution, the technology must be even smaller still. Each of these elements would need to be controlled separately; the system would have to be capable of addressing 120,000 elements individually, but also incredibly quickly. Mechanical technology for this purpose and of this size is simply not available yet, and even the most basic technology that is available is very expensive. One alternative solution to this compound problem has been explored within the Heidelberg Tactile Vision Substitution System (Maucher et al 2000) which greatly reduces the number of required taxels by using a virtual display area (think of taxels as tactile counterparts to visual pixels). To achieve this, 48 such taxels have been mounted on a carriage that is moved across the larger display area that would have required 2600 taxels; still a very low resolution in terms of overall definition.
20.4.2 Haptic Display Technology
Partly influenced by the rapid expansion of the computer games market and general interest in virtual environments, affordable force-feedback and haptic-technology has been available for some time now (force-feedback joysticks, 'rumble-pads', vibro-tactile mice etc.). Based on similar technology to that developed within the Phantom series of force-feedback devices from SensAble Technologies, sophisticated games devices such as the Novint Falcon are being used to offer a rich and immersive sense of physical presence within computer games. Offering three degrees of freedom, a motor controlled arm is used to convey sensations like recoil, impact or different levels of resistance to a player's hand held control device that might take the form of, for example, a gun or bat. The Phantom devices are similar, and like the Falcon can take control of the user's hand or limb, creating a rich variety of virtual textures and other effects. Remember though, haptics implies a combination of both tactile and kinesthetic feedback; however, the devices described here do not really provide much feedback at a tactile level. In this sense, haptic perception within the domain of virtual displays is unlikely to be active and therefore does not match the definition provided earlier. To that extent, the usefulness in such displays lies in their dynamic nature rather than in the richness of haptic interaction that can be achieved.
Building on their initial ventures with force-feedback and vibro-tactile feedback gaming devices, the company Immersion is now producing multi-actuator controllers aimed at bringing haptic interaction to the now ubiquitous touch-screen interface on smart-phones and 'tablets'. Designed to be placed behind or around the screen-area, up to 16 actuators can be programmed to produce various vibratory effects that can reinforce user actions. Pressing an onscreen button, for example, can be accompanied by an effect that will go some way towards informing the user that the action has been achieved. Again, this is useful feedback, and the dynamic nature of the display brings considerable flexibility to the interface, but the user, for example, will not be able to feel outline or detailed texture in the way that a static display can offer; the 'display' as such is still predominantly visual.
20.5 Future Directions
It is clear that current technologies cannot yet facilitate dynamic tactile displays that offer the same richness of detail and contrast that we experience with the objects around us on a daily basis. However, as much as this would be an obvious ultimate aim within a context of augmented reality or virtual environments, there are perhaps much more attainable targets that will still be of significant use. Access to information for visually impaired (VI) users has been a strong feature within our discussion, and it is clear that a truly dynamic and rich tactile display could bring considerable benefits to this area. For example, e-books are now commonplace, but how about e-books where the user can explore and interact with braille and tactile diagrams? Might there be other more mainstream uses for similar technology? At a personal level, my background is shared between technology and music, and though I have considerable enthusiasm for virtual synthesisers and music environments, I still much prefer the sensation of touch and 'feel' with real controls to those afforded by a touch-screen. In this context, which would be more useful to me, a rich colourful and graphical interface with no meaningful tactile interaction or a simple graphical display that allows me to touch and feel the controls? After all, both will be controlling and producing the same sonic output, and that will always be my main concern. Realistically, I imagine that I would probably benefit from a little bit of both. So, how close are we to having such an immersive level of dynamic tactile display? We considered earlier how the size of the mechanical parts creates restrictions in both control and resolution, but there have been significant breakthroughs in this area with the exploration of gel-based pixels that respond to heat. Measuring just 300 microns across, such a pixel will respond to heat from a fine light source and shrink to just half its original height of 0.5mm; the pixel also becomes opaque such that the change is quite visible. The experimental display contains over 4000 of these pixels with about 297 being available within each square centimetre such that fine resolution tactile graphics can be achieved. However, the refresh rate is still slow compared to a standard LCD display, for example, but so is that of a typical e-book. This is still an emerging technology, but one that if developed fully could easily lead to the creation of the tactile e-book that we just considered. If this same technology can be further enhanced to enable input on the part of the user, then this really could be the gateway to enabling rich haptic interaction within all kinds of novel tactile-interfaces.
20.6 Where to Learn More
It may come as little surprise that much of the information referenced within this chapter has been drawn from research into non-visual approaches for enabling access to information for visually impaired users. Though there is significant interest in the application of tactile interaction within mainstream devices there is already a considerable body of knowledge that has been established because of research into accessibility through new technologies. There are a number of conferences that address non-visual interaction within a broader range of disability related topics and these include:
Conferences on Human Computer Interaction should also be included as potential sources of further information:
There are also a number of journals that will occasionally address the notion of Tactile Interaction though perhaps within a wider of interactivity. These include:
20.7 Commentary by Alistair D. N. Edwards
Tactile interaction occupies an odd position within the discipline of interaction – as is reflected in this chapter. On the one hand it is associated with futuristic, exotic technology that is still far from mainstream use, but on the other hand it is an element of most conventional human-computer interaction almost without our realizing it. That is to say, that the ubiquitous keyboard and mouse input relies heavily on haptic and proprioceptive senses in an unconscious way that we tend to take for granted. Thus, most people probably do not think that they engage in haptic interaction, but would acknowledge that it must be useful for those who lack other senses (notably sight) for whom tactile communication such as braille would seem invaluable.
It is to be hoped that this chapter will make readers much more aware of the truth of the situation and of the potential of tactile interaction. The keyboard and the mouse have both evolved over time, adapting to the physiological abilities of the user, but with minimal reference to the underlying biology. Of course, the keyboard is a classic example of non-optimal evolution: it is suggested (Noyes 1983) that the conventional qwerty layout was designed to slow typists down, to reduce consecutive-letter clashes on mechanical typewriters and now we are stuck with it. On the one hand, it is fun to speculate how the keyboard might have been designed if the inventors had had the benefit of reading a chapter such as this, but on the other, one has to note that even with the availability of this knowledge haptic interaction clearly a long way from reaching its full potential.
We have physical devices, including keyboards and mixing desks and lighting desks, which inherently give good haptic feedback and interaction, but in the digital, virtual world, we are constrained by the technology. Is the investment (of time, money, space etc.) required to use a device such as that depicted in Figure 21.13 worth the benefit derived?
The tactile picture is a device which seems in some sense 'obvious', particularly to those of us who have sight: if you can't see a picture, then you should be able to feel it. I have spent a lot of time and effort working with tactile diagrams, mainly within the University of York Centre for Tactile Images (sadly no longer in existence for lack of financial viability). We produced a wide range of diagrams for different purposes, working in collaboration with a number of important clients (Including The Deep, Hull; The Jorvik Centre; The National Trust; English Heritage and The National Railway Museum). Repeatedly we found ourselves repeating the assertion (mentioned by Challis) that a good tactile diagram may look bad. That is to say that the haptic senses are very different from vision.
The immediate differences are obvious – when one stops to think about them. Tactual sensitivity is low. Challis mentions the two-point limen but refrains from giving a figure for it. This is probably wise, since it is at least controversial as a measure of tactual sensitivity and is hard to gather agreement as to average values. The point is, though, that the value is of the order of millimetres – much lower than visual resolution. In Figure 21.5, Challis shows examples of patterns which could be produced on swell paper. In the visual representation they are all clearly distinct, but in experiments with swell paper samples (Magill 1999), we found that the average person can only distinguish three levels of pattern: smooth, medium and rough, regardless of their visual pattern. For instance, while the vertical and horizontal cross-hatching shown in Figure 21.5 are clearly different visually, they would probably be perceived as being the same on tactual inspection.
Furthermore, in exploring a tactile diagram, the person is likely to use at most two finger tips and more likely one. It is easy to characterize this as exploring through a tactile pinhole. Yet, in practice, it is likely that the situation is even worse than that. Vision is an inherently integrative sense. The angle sensed by the eye is very small, but because the eye moves constantly and the brain can integrate information we get a wide field of view which forms a picture that we perceive as coherent. The only people who need to be taught how to perceive visually are those who have lacked the sense of vision and have then had it given to them (usually through surgery, as in the case of Virgil in (Sacks 2009)), whereas even proficient braille readers need to be taught how to use tactile pictures. Evidence from such cases of lately-acquired sight suggests that it may be that people who mature without sight may never develop the ability to integrate fragments into a comprehensive picture (literally and metaphorically). In other words, a tactile picture may always be of limited value – even if it has been designed in such a way as to make the most of the haptic senses.
A limitation of most tactile graphic technologies, such as swell paper, is that it is static and hence the idea of a tactile screen is attractive – and perennial. I have stopped paying much attention to announcements and papers by people who have had this idea again and believe they have found the one which will work. I would like to propose a moratorium on such publications until someone can present their device which works, is reliable and affordable. In the absence of such a ban, I will impose my own by not reading them. I have no doubt that one day someone will invent the appropriate technology, but – for the reasons set out in this chapter – it is a hard goal to achieve. Even this chapter mentions another of these attempts ('gel-based pixels that respond to heat'), but clearly it is quite speculative again, since there is not even a publication cited.
It is clear that the haptic sense becomes even more important in the absence of the dominating sense of vision. Braille is the richest form of tactile communication its use by blind people is very low. It is estimated that fewer than 2% of blind people read braille (Bruce, McKennell et al. 1991). The reasons for the low take-up are unclear and probably complex, but the essential explanation must be that blind people do not perceive that the benefits they will gain are worth the effort that it will take to learn braille. This particularly applies to those who have had sight and have lost it – which is the vast majority. For those of us who have sight, exclusion from (printed) literature seems a great loss, but for most of those who experience this the difficulty of learning to read tactually is not sufficient motivation.
Yet it is not sheer laziness on the part of sighted people that means that they do not make as much use of their non-visual senses as they might. In radical experiments by (Pascual-Leone, Theoret et al. 2004) sighted participants were blindfolded 24 hours per day for several days and given daily lessons in reading braille. They made good progress in learning braille and brain scans showed that areas of the brain cortex normally associated with visual processing were being reassigned to processing tactile information. A control group of non-blindfolded, sighted participants made minimal progress on learning braille and showed no brain adaptation. In other words it seems that braille is not only likely to be useful for blind people, they are the only people who are likely to be able to use it.
To summarize, haptic senses are used to a greater extent in conventional interaction than most people realize. They could be used more and to better effect if designers were more informed regarding them. Virtual haptics are being used increasingly and this trend will accelerate as the technology develops, but there is something to be said for the use of (well-designed), old-fashioned physical interactors (knobs, switches and the like). Ultimately, though, there is a limit to the use that can be made, given the physiological limitations of the haptic senses.
20.8 Commentary by Klaus Miesenberger
I like the chapter a lot — a brilliant source for students and practitioners in need of the state of the art. Two things in particular:
- We must not forget to look at the fundamentals, results, success stories and failures before jumping into a new adventure!
- Things which are difficult to grasp makes it hard for readers and students to cope with the subject. This chapter helps by summarizing and bringing the issue to the point.
The chapter presents the different components, the bits and pieces relevant for tactile information presentation, the potential of tactile/tangible/touch interfaces and the challenges we are facing on the way to better support efficiency, quality and joy of interaction with technical systems and content. The chapter does not promise what it can’t answer and does not try to answer what the next steps in tactile/tangible/touch interaction will be — as nobody could foresee that touch displays have suddenly come to proliferate: Technology, concepts, ideas and prototypes were available, described and discussed for decades, but only recently the concept and its implementations exploded. Why has tangible interaction been so successful in attracting a big part of the “non-techy” population and get so many people — including aging people and people with disabilities — engaged with ICT? Is it touch?, is it vision?, is it overcoming of hand/touch-eye-ear- coordination?, is it the mobile aspect?, is it non-text oriented aspect?, ...? In the end, many answers and factors have to be taken into account to grasp the big picture allowing us a next step in the journey of interface design e.g. by adding a new quality of perception into the tactile experience.
Ben Challis’ chapter collects what might form a next step in the interface revolution. This revolution could be summarized as “From the hand-axe to the pointer”: It started with switches and switchboards (with lots of tangible interaction), over console, WIMP, SILK towards tactile/touch/tangible. Interesting to see that first we moved away from immediate and intuitive interaction using tactile interaction towards a new virtual, intangible experience (i.e. the WIMP interface) with its advantages in terms of efficiency and independence, but also with its limitations in acceptance due to a lack of intuitiveness.
In our pursuit of creating user experiences that are intuitive, usable, enjoyable and efficient, research and development increasingly uses touch and sound as well as other senses like acceleration, gravity and the olfactory sense to achieve this. Tactile interaction combines several types of sensory experiences, which is perhaps the reason it seems intuitive — it give us a feeling of easy, efficient, effective, straight forward and joyful interaction. As designers, we must however first understand touch, tactile and tangible interfaces on their own before advancing into cross-sensing approaches.
Ben’s chapter provokes such questions, which makes it an ideal source for students, researchers and developers when diving into what seems “obvious” at first glance. We must never expect simple, easy to apply answers in interface and interaction design. We have to learn to put the right questions in the right context, to be prepared to question the obvious and to progress based on a sound and respectful analysis of the state of the art. This is what this chapter is.
And last but not least, I like how we learn from the extreme: People with reduced or no visual sense in many aspects show how tactile or alternative interfaces have to be designed. They are key experience and know-how owners.
Good interfaces “don’t make us think”. But good interface design depends on designers and developers willing and prepared to think about the big picture of interaction and interface design. Ben’s chapter is the right source to make us think before we implement!