Understanding Driving

Applying Cognitive Psychology to a Complex Everyday Task

John A. Groeger

• Publisher: Routledge
• Year: 2013
• Print ISBN:
  9780415187527
• eBook ISBN:
  9781134690978
• Topic: Psychology
• Subtopic: History & Theory in Psychology
• Index: Psychology

This book closely examines what is involved in driving. It identifies the aspects of perception, attention, learning, memory, decision making and action control which are drawn upon in order to enable us to drive, and the brain systems involved. It attempts to show how studying tasks such as driving can help to understand how these fundamental aspects of cognition combine to facilitate performance in complex everyday tasks. In doing so it shows how a very broad range of laboratory based findings can be applied, and that through our attempts to apply this knowledge to complex everyday tasks, we gain, in return, a greater understanding of fundamental aspects of human cognition.

CHAPTER ONE

Assessing distance, speed, and time

Perhaps the most immediately apparent aspects of the driving task are that drivers must assess the distance between themselves and other objects, the speed at which both are travelling, and how these might combine to allow the driver to act appropriately in the time available. As will emerge as the book proceeds, I think there is rather more involved in driving than merely combining different sources of visual information with particular motor patterns, but these elements of the task are themselves non-trivial. In this chapter I briefly outline some of the factors that influence distance and speed perception, and the way in which we judge when we may reach another object, or when it may reach us.

Although engine, wind, and road noise make some contribution, our impression of speed and certainly of distance depends on what we can see. I wish to begin by considering how fundamental the information is that we use as a basis for assessing speed and distance.

LIGHT AND THE EYES

How is it that we see something clearly in some situations, but not in others, such as on a foggy day or at dusk? Why is it that some things are easier to see than others? How does our visual system use the light available to help us build the complex visual world we generally take for granted?

Between ultra violet and infra red, both invisible to the human eye, lies the band of electromagnetic radiation that gives rise to our sense of seeing the world. For most of us this sense of seeing the world consists of the impression that there are “objects” which have a particular shape (i.e. surface, edge, and texture), and colour, the latter comprising the wavelengths of the visible spectrum ranging from violet (390 nanometres) to red (700 nanometres). Light-energy is lost (e.g. absorbed), scattered (e.g. through defraction), bent (e.g. by refraction), and reflected by different media (air, water, etc.) and surfaces (some surfaces reflect some wavelengths more strongly than others; some textures reflect light at different angles). This information, the way our eyes operate, and the knowledge we have of the world, allow us to see what we see.

The light reflected from surfaces across an angle of approximately 150° is directed by the lens of the eye onto the retina, an area of approximately 120–130 million light-sensitive cells at the back of the eye. The extent to which the lens changes in shape—becoming thinner, thus dispersing the light reaching the eye over a greater area of the retina, or fatter, directing the light towards a smaller area of the retina—partly governs the clarity with which we see near and distant objects. The ability of the ciliary muscles to contract or relax, and hence the power of the lens, decreases with age.

The other major factor involved is the nature and concentration of light-sensitive cells in the area of the retina onto which light is directed, whether by lens accommodation or eye movements. Light-sensitive cells comprise cones, which do not respond at low light levels, and rods, which do. Some cones are sensitive to blue, others to green, and others to red. However, because cones respond in bright light (i.e. photopic vision) do not respond in low light (i.e. scotopic vision) conditions, we see colour in reasonably bright conditions, but not in dim or dark conditions. Rods outnumber cones by about 20 to 1 on the surface of the retina as a whole, but the concentration of rods to cones changes over the retina surface, from the cone-rich fovea to more peripheral rod-rich areas. The former provide high-acuity vision in bright light, while rod-rich areas, which comprise the vast majority of the retina, provide considerable sensitivity, but not acuity, in dim light.

As light falls on different areas of the retina, causing some mixture of rods and/or cones to fire, nerve cells beyond the retina in visual pathways to the brain also respond. Which nerve cells will respond depends on the areas of the retina stimulated (i.e. receptive fields), and the periodicity of the pattern of light reaching the retina. The periodicity of this light pattern, usually expressed as the number of cycles per unit of visual angle, is called “spatial frequency”.

Imagine, for a moment, that you are a hawk hovering above a zebra crossing. If you are hovering just a few metres above the crossing, the characteristic repetition of black and white stripes goes through just a few cycles for every degree of visual angle of the scene. If you hover, say, twice as high, the same visual angle will comprise more repetitions of the black–white pattern. That is, within the visible range, the higher up the hawk is, the higher the spatial frequency of the zebra crossing. Nerve cells beyond the retina are tuned to respond to different spatial frequencies, thus allowing our brains to build up the complex images, comprising many different spatial frequencies, that guide our everyday behaviour.

In most lighting conditions, telling the difference between the black and white portions of the zebra crossing is relatively easy, because the contrast between adjacent parts of the image is very high; formally, contrast is the difference between the maximum and minimum luminances in a pattern of light expressed as a proportion of the mean intensity or luminance. Telling the difference, under all conditions of illumination, between white and light grey stripes, or black and dark grey stripes, would be a lot more difficult. Where more light is available, assuming the light source does not cause “white-out”, discrimination of colour differences, object edges, and textures becomes easier and more accurate.

At midday during a UK summer, the intensity of the light from the unobscured sun is in the region of 10,000 lux. At sunrise or sunset the intensity of this light falls by more than a factor of 10, while by the light of a full moon the light intensity is around 1 lux—still well above the absolute human visual threshold. It is not that we humans cannot see in relative darkness, it is rather a case of what we cannot see. Colour information is lost in dim light (i.e. scotopic vision), when cones do not operate. Unless the luminance difference between the colours is very large, as in the case of a white object against a black background or vice versa, contrast discrimination is also impaired as light levels drop. In daylight foggy conditions, the light is sufficient for gross differences in colour to be distinguished, but image-edges are blurred, and the luminance of objects is made more homogeneous, rendering contrast sensitivity much less useful than it would otherwise be. As a result of this reduction in contrast sensitivity, and the blurring of edges and texture information, spatial frequency information is also lost in fog. As we will see later, this renders drivers’ ability to assess distance and motion quite inaccurate. Because as we age we tend to lose contrast sensitivity, and our eyes adapt less rapidly to changes in light levels (see Kline & Fuchs, 1993), the elderly driver is especially prone to problems in low light and blurred visual conditions. This, as we shall see, has implications for the visual standards drivers should be required to meet, and also for the driving environment we design.

DEPTH AND DISTANCE

As drivers we need to be able to judge reliably how far away objects are, for a variety of reasons, e.g. in order to stop before we reach them, to turn before or after reaching them, or to overtake them. This might involve using some precise measurement in metres, or some more vague approximate subjective scale that allows us to know reliably which things are further away than others, or even some still finer classification of distances into those that permit a certain action (e.g. overtaking before a bend is reached) and those that preclude that action. We may not be able to verbalise these estimates, but nevertheless we reveal consistent use of such estimates in the way we behave. In the second part of this section I review how we estimate distance, but first I want to describe the distance-related information that is available in our visual world.

Different sources of information are used in making judgements of distance. These partly reflect properties of the object and its environment, such as colour, texture, size, height in the visual field, linear perspective, etc. The information used also partly reflects properties of the perceiver's visual system, such as lens accommodation, blur, eye vergence, and stereopsis. These sources of information are more available and more useful in some circumstances than in others.

Under optimal viewing conditions, a focused image is formed within and beyond certain distances from the eye. This range of distances, referred to as the depth of field, stretches from approximately five metres to infinity, for an eye focused on infinity. For objects closer to the observer than five metres, a clear image is maintained by adjustments of the lens. The requirement to “accommodate”, and the extent of accommodation required, provide information on object eccentricity closer than this to the observer. Beyond this range, a single stable image is maintained by turning the eyes towards or away from each other (“vergence”). The amount of vergence and resultant strain provide some cues to distance, but obviously only relate to a single object or distance at a time. Lack of clarity of the image, that is image blur, perhaps arising from inadequate accommodation and vergence, can provide yet another source of depth information (Pentland, 1987). This can be misleading when it arises from environmental conditions such as mist or fog, leading the driver to suppose objects are further away than they actually are. Usually, however, the combination of information from two eyes provides a rather more useful source of information about distance, arising from the difference between the images available from each eye.

Suppose a driver is travelling along a three-lane highway, for the sake of argument in the centre lane. He is following a red car, which is also in the centre lane. Assuming that both of his eyes are focused on the red car, he will have the impression of a stable image of the red car ahead. It is actually a combination of two foveated images. Were he to look at the red vehicle only with his right eye, the image of the car ahead would be displaced relative to the combined image, and to the image he would have if he shut his right eye and looked with his left. Now suppose there is a blue car in the outside lane. Assuming he continues to fixate on the red car, the driver will nevertheless know whether the blue car is closer or further away. Suppose the right eye image of the blue car, which is actually displaced to the right of the point of fixation, is closer to the fovea of the right eye than is the left eye image to the fovea of the left eye. This only arises where the blue car is further away than the fixated red car. If the blue car was closer than the red car, this pattern would be reversed. But the disparity actually yields more information than just relative depth. As disparity depends on the difference in depth between the two vehicles, and change in disparity is closely related to the difference in depth divided by squared viewing distance, the size and sign of relative distance are readily available. Thus, disparity might, in the case described here, provide some information as to whether the driver can overtake the vehicle in front.

Information from one's two eyes remains separate, as it travels along distinct pathways in the lateral geniculate nucleus, until it reaches the observer's primary visual cortex. This is the earliest stage of the visual system, which is responsive to binocular information, and there is evidence that cells here allow disparities to be cortically measured (e.g. Poggio & Poggio, 1984). However, important as disparity is, because similar disparity can arise either from a near object that is close but deep, or from a distant object that has minimal depth, disparity alone would be insufficient for us to negotiate our way around the world. Let us assume, for example, a motorist is following another vehicle, intending to overtake it. In order to ascertain whether the road ahead of that vehicle is clear, the motorist pulls towards, perhaps over, the centre line. A relatively close long articulated vehicle would produce as much disparity as a distant family saloon. Fortunately, there are a range of visual and other cues that generally override this ambiguity, but it is clear that we require some form of scaling to recover real depth or distance.

Beyond the distances where vergence and stereopsis provide veridical distance information, colour and texture may be still be available to the viewer and may provide information useful to the viewer. As there is a correlation between size, which can itself related to colour and texture, and disparity, both can inform inferences about distance. For example, where only coarse texture information is available, sensitivity is best for larger disparities, whereas at much higher spatial frequencies sensitivity is optimal for small disparities (see Smallman & MacLeod, 1994). This and other evidence supports the view that spatial filters, and resulting receptive fields, underlie the correlation between size and disparity.

Far beyond the point at which colour and texture information is unavailable, and perhaps up to several kilometres in optimal viewing conditions, presumably at much lower spatial frequencies, factors such as linear perspective, height in the visual field, relative size, shading, and occlusion may all provide cues to relative, if not actual distance. The first of these, linear perspective, arises where one has the impression that a long stretch of straight road is actually narrowing so that the edges meet far ahead. Objects further away are assumed to be higher in the visual field, and the corollary of this is that slope of regard influences distance estimation, i.e. given the same flat scene, passengers in low-bodied sports cars will tend to believe distant objects are further away than will those travelling in the upper deck of a double-decker bus. Shading can provide a sense that the object has some depth and thus gives rise to disparity, and depending on which parts of an object are seen to be shaded can indicate both the slope of regard and relative distance. Occlusion of one object by another also provides information about their relative distance from the observer, and where known objects are being viewed, the apparent gradient of the image viewed, together with the assumption that light comes from above, can also provide information about relative size, and consequently distance.

As many of these sources of depth- and distance-related information are relatively independent of each other, each can provide different information about how far away an object is, and thus they can potentially conflict. Good evidence is emerging that such cues are combined and some come to dominate more in some situations than others (see Bruce, Green, & Georgeson, 1996 and Cutting, 1986, for more detailed discussion). Having sketched briefly what types of distance information can be available to observers, let us now explore how such information might influence drivers’ impressions of how far they are away from other objects they can see.

Distance estimation

Although it does not necessarily follow, it might be expected that with all of the aforementioned information available, people should estimate distance accurately. A variety of studies show that this is not the case, with people being both lawful and consistent in making judgements, but generally underestimating distance, i.e. assuming objects are closer than they actually are. Teghtsoonian and Teghtsoonian (1970) report that for distances ranging from 5 to 480 feet, estimates are a power function of the actual distance, with an exponent of 0.85. The suggestion that estimates are a power function of the actual distance means that, where the exponent is less than 1, far distances will be underestimated much more than near distances. That is, an object 5 feet away is believed to be just 3.93 feet away, one 480 feet away is thought to be some 190 feet distant. Similar findings are reported by Da Silva and Dos Santos (1982) and Da Silva and Da Silva (1983), although a recent meta-analysis of distance estimation studies shows that the exponent of the power function varies considerably, depending on the estimation task used, the context in which judgements are made, and those making the judgements (Weist & Bell, 1985). Exponents of a similar nature have been reported for driving under simulated conditions (Groeger, 2000), but it is important to recognise that estimates made are also subject to a number of other biases.

Interpretations of the information available may also be biased by previous experience, the context in which decisions are made, and how the person assessing the distance communicates their judgement. This is most clearly seen with respect to the effect of familiarity on judgements of distance, where it is widely shown that known objects are thought to be further away than similar, less familiar objects which are as distant. Thus, for example, Predebon (1990) has shown in relatively free-field viewing that a woman, located at 90, 100, 140, or 150 metres, was always judged to be further away than a similar-sized board 120 metres from the observer. Similar results have been reported under laboratory conditions with smaller objects at shorter distances such as playing cards (e.g. Carlson & Tassone, 1971), and more recently in contrasts between a familiar object (red London bus) and a similar-sized untextured object at much larger simulated distances (range 10–450 m; see Groeger, 2000).

Earlier work tended to encourage comparisons between objects of normal size and off-size versions of the same objects. The emphasis on off-size representations has diverted attention away from the potentially significant issue of whether a normal-size object viewed under naturalistic conditions influences perceptions of extents. In a full-cue situation, the familiar size information provided by a normal-size object does not conflict with visual and oculomotor information, and for that reason might not be expected to influence perceived distance. It certainly appears to be the case that the effect does not occur for unfamiliar objects in the vicinity of familiar objects, and is weakened if the familiar and unfamiliar object are viewed in the presence of a common reference target (see Predebon, 1990).

A further difficulty with some previous research is that the methods used to allow observers to communicate their impressions of distance have ranged from raw verbal estimates to the interposing of other objects between the observer and the target, or other forms of bisection tasks. As not all methods yield similar results, and as one might not expect drivers to rely on verbal estimates but on production estimates to guide their behaviour, it is useful to have it confirmed that the effect of familiarity of the observer with the distant object, such that more familiar objects are thought to be further away than unfamiliar objects at the same eccentricity, occurs whether a verbal or production estimation method is used (Groeger, 2000). Similarly, in the same paper it has been shown with both verbal estimation and production methodologies that where repeated distance judgements are made, a small range in the distances encountered will tend to increase the observer's tendency to underestimate distance (Groeger, 2000; Teghtsoonian, 1973).

It has been suggested by some authors (e.g. Stewart, Cudworth, & Lishman, 1993) that one source of increased accident involvement of child pedestrians is the possibility that they are mistaken for a more distant adult. As will be clear from the foregoing discussion of how depth and distance information are combined, under normal viewing conditions such apparent size illusions will be prevented by the relative size between pedestrians and their surrounds. The research reported earlier implies that where visual information is reduced, e.g. in darkness, fog, rain, very bright light, drivers may well assume objects that are readily identified (e.g. road users, be they pedestrians, motorists, or cyclists), to be more distant than they actually are, and thus drive in such a way as to leave themselves with less time to react than they would intend under normal viewing conditions. Similarly, as discussed further shortly, availability of more environmental information enhances drivers’ estimates of when they will reach distant objects (Cavallo, Mestre, & Berthelon, 1997). Another difficulty in reduced cue situations is that the amount of disparity required for detection depends in part on the information available from the objects concerned. Dark objects against gloomy backgrounds, or objects that appear blurred (i.e. where only low spatial frequency information is available), will require far greater angular separation for relative depth to be detected. Where high spatial frequency information is available, very small tolerances in angular separation are sufficient for reliable disparity detection (see Smallman & MacLeod, 1994).

Thus, a wide variety of environmental, perceptual, and cognitive factors underlie our impressions of depth, and hence our sense of distance. Most of this information, it should be noted,...

Table of contents

• Front Cover
• Half Title
• Title Page
• Copyright
• Contents
• List of illustrations
• Acknowledgements
• Foreword: Drivers and the driving they do
• 1 Assessing distance, speed, and time
• 2 Motor responses and behavioural repertoires
• 3 Combining perceptual-motor skills
• 4 Attention, automaticity, and distraction
• 5 Learning, instruction, and training
• 6 Memory for driving
• 7 When driving is dangerous: Arousal, assessment, and hazard perception
• 8 Appraisal, efficacy, and action
• 9 Age, neurological damage, disease, and driving
• 10 Towards a cognitive account of driver behaviour
• References
• Author index
• Subject index