SASI Home| Publications | Data | Maps | SASI Staff
Dorling, D. (1991) The Visualization of Spatial Structure, PhD Thesis, Department of Geography, University of Newcastle upon Tyne

Introduction: Human Cartography

Images are only images. But if they are numerous, repeated, identical, they cannot all be wrong. They show us that in a varied universe, forms and performances can be similar: there are towns, routes, states, patterns ... which in spite of everything resemble each other.
[Braudel F., 1979, p.133]

This dissertation presents the thesis that the study of society can be enlightened through the visualization of social structure. Spatial social patterns provide the most arresting pictures of the underlying order and workings of the system, but other facets of the process can also be transformed into images to illuminate their organization. Visualization in social science throws light into a dark world of specialisms and obscurity, showing at an instant how all is connected and everywhere is different. Most importantly we can begin to see how the structure is changing, what was, and what could be.

The antecedents of this work lie most firmly in human geography and cartography while being strongly influenced by writings in, and the combinations of, many other disciplines (Arnheim R. 1976, Muehrcke P.C. 1978, 1981, Bertin J. 1983b, Szegö J. 1984, 1987, Anderson J.M. 1988, Cuff D.J. 1989). There are contributions from studies in computer and statistical graphics, graphic design and art, mathematical abstraction (Print I) and political science. Current thinking in the study of history and sociology guides much of the writing. Above all, this dissertation is concerned with designing new ways of seeing the social world we live in. Before doing that, it is necessary to explain why the accepted geographical techniques are being discarded by the visual methodology proposed here. In particular, the conventional use of maps of physical geography, to show the spatial structure of society, is rejected.

Maps were designed to explore new territories and fight over old ones. They show where oceans lie and rivers run. Their projections are calculated to aid navigation by compass or depict the quantity of land under crops (Print II). They are a flat representation of part of the surface of the globe; they show things which often cannot be seen. How can we see social structure, as the map opens up land to the eye1? How can we begin to see the patterns of society, which, from being part of it we know are there, but have never seen?

Maps were not designed to show the spatial distributions of people, although the single spatial distribution of people upon the surface of the globe, at one instance in time, can be shown on them. They cannot illustrate the simplest human geography of population. People are but points on the map, clustered into collections of points called homes, into groups of points known as villages or cities. Communities of people are not like fields of crops. The paths through space which they follow are not long wide rivers of water, and yet, to see anything on maps of people they must be shown as such.

Conventional maps cannot show how many people live in small areas, instead they show how little land supports so many people. They cannot show who the people are, what they do, where they go. They show no temporal distribution, they do not need to — how quickly do rivers move or mountains shrink on a human timescale? They will not show the distributions of people changing — international migration, moving house, or just going to work. They cannot portray the distribution of the wealthy or the poor; on the map, at almost any scale, they live in much the same square inch of paper. Nor will they show where and when people had certain jobs, certain power, voted, were out of work, lived and died. What, after all, is a map2 (Hsu M.L. 1979, Brannon G. 1989, Phillips R.J. 1989)?

Pictures can make ideas plausible, paper beautiful, millions of numbers meaningful. They have intrigued many, as maps and charts of rivers and mountains, to the point of being the pretext for their studying geography. Here traditional maps are the inspiration, but not the foundation, for the generation of new graphics to form pictures of people, with their rivers of roads down which they flow, and mountains of cities up which they climb. The theory of how the patterns, movement and evolution of the lives of millions can be transformed to be represented visually, is presented here3.

We want to make sense of the reality of thousands of people simultaneously threading their way through life (Print III). What are they doing and why are they doing it? How can we see into every home, know what everyone does? We can't, but we can guess and we have some clues. We can guess from what, introspectively, we know from being part of society. We amass clues when people are counted. There has been an obsession for counting people since recording became possible. Every ten years, in many countries, hundreds of thousands of people count people (the census). Increasingly our actions are being recorded; we are each noted now several times a day, from the heat we register on satellite images, to almost every transaction and phone call we make or unit of electricity we consume. The question this thesis addresses is how can the part of this huge disparate collection of clues that is available to us be built up to form an at least partial picture of the patterns we imagine exist. The answer is, as it has often before been, in the form of pictures.

The conventional statistical treatments of numerical information about people averages them, agglomerates them and destroys the detail that is of interest4. They take a million numbers and return half a dozen. These techniques were conceived when little better could be done. Now it is possible to show you a million bits of information at a glance that would be challenging to describe in a thousand words (Print IV). Our minds are the most powerful tools we have to address these problems. The difficulty comes in trying to address these problems to our minds.

Orthodox cartographic methodology has been translated onto the computer screen (Bickmore D.P. 1975, Hagen C.B. 1982, Taylor D.R.F. 1985, Jupe D. 1987, Goodchild M.F. 1988b, Visvalingham M. 1989, Muller J.C. 1989). The name has changed to geographic information systems, but the fundamental basis of physical geography has remained5. Thematic maps drastically distort the reality they purport to contain, at worse reversing the patterns that exist. People who study people, who are interested in societies, politics, history, economics and increasingly even human geography, do not use these maps. A topographic map base allows, at most, the depiction of human land use. People have created maps based on human geography in the past, but only with the advent of sophisticated computer graphics has it become possible to do this on an easily replicable basis.

This thesis is presented as part of the academic revolution known as visualization, and parallels are drawn with the wider world of computer graphics. Therefore the basics of what can physically be seen have to be introduced. How these images are created has to be explained — from the theoretical to the practical problems. Most importantly I address the problem of how time and space can be transformed to represent clearly the patterns within them, on paper, or in animation.

That transformation is essential for representation is a most difficult argument to accept, for it completely alters the images produced and hopefully the emphasis of the viewer to places, and, more importantly, the relationship between places and times — the metric. The argument for transforming to population space, distributions which exist only in that space, has been made repeatedly over many decades in human geography. It is simply reiterated in stronger terms here, the new images being traditionally referred to as cartograms. Put simply, people no longer exist on paper as points, but as areas6, so can now be legitimately drawn as fields, their paths as rivers (Print V) perhaps running through a landscape of accessibility covered with the vegetation of some aspect of social structure.

This thesis draws on those patterns of people that are familiar to this writer and the envisaged reader. Britain in the 1970s and 80s is all that I have known in any detail, and only a very small part of that. The clues given by the official sources consist mostly of the absolute numbers, age and sex of people across the country. Then, every ten years the combinations of their answers to a few questions at the census are provided, where they were born, what job they do or did, where they did it, where they moved (Print VI) and so on. But there are other forms of information that can be drawn on, and, as one claim of visualization is the ability to handle large quantities of loosely related data coherently, other sources and surveys are called upon. How people voted in the local and national elections of the decades, national surveys of workplaces which were conducted in several years, the health service records of migration, building societies' lists of house sales, and information on the infrastructure of roads, railways and settlements for example, are all digitally available7. What is sought here is the means of putting these numbers together, as a collection of images forming one picture of one place during a short, twenty year or so period of time.

The most simple of spatial distributions to envisage are those captured at single instances of time, and so it is these with which the examples of the visualization of spatial distributions in this dissertation begin. Much of the static spatial social structure is already known intuitively to social scientists, if not in such great detail and with all places shown in immediate relation to each other. The degree of complexity and interdependence shown by the images in this work may also be new to many. The dissertation then moves on to show changes in the population over time in a single picture (Print VII). Much of what this shows about Britain will be unexpected, as it is only through the methods employed here that such things can be seen. The way people move about, day to day, and year to year, is visualized as streams flowing through space. It must always be remembered that I am not concerned with two hundred, or a few thousand people, but the activities of as many as fifty million. The computer is used to handle these vast numbers, not to produce more numbers, but pictures — black and white, coloured, and, when required, moving.

Finally I can begin to produce images to depict the little that is known about large numbers of people, which are totally different from anything we would recognise in current practice. A notional surface is proposed where the distance between points is equal to how long it would take to travel between them, upon which we can then drape other distributions. It may soon be possible to create true volumes of pattern and colour to depict the entire evolution of a single phenomenon, for example unemployment at every place, every month. The alternative is to cut through this distribution, collapsing all of space to one point, to draw graphs over time, or all of time to a point, to show a simple spatial pattern. Inevitably we ask, can we now combine these disparate images and compare the evolution of one thing with the flows of another and the distribution of yet others, without collapsing reality into dimensions which cannot contain its complexity?

Presented here is a methodology for studying relatively data rich spatio-temporal distributions and their interrelations. This goes beyond the accepted format of book chapters (containing a few tables, perhaps a graph, or a coarse thematic map) on each of a small number of topics, with an overview chapter implying that everything is related but that it's all very complicated. If it is complicated it is interesting — so let's look at it, rather than repeatedly explaining away the simplest points, tabulating and sorting the basic rates, or drawing yet more examples of inappropriate poor quality choropleth maps (by computer of course).

Images of recent British history are being created here which allow new questions to be asked, show different distributions to be explained, the distributions that many social scientists know are there, but which traditional cartography fails to depict, and hence to explore8. There are also glaring patterns to be seen in the well trodden census tables and government figures which have been ignored, before we even begin to look for the more subtle or complex and detailed relationships.

Visualization can be claimed to solve many of the fundamental problems identified in studying spatial social distributions (Prints VIII & IX). The fact that the way you subdivide the space and time you choose to study can drastically alter the overall impression of your results, suggests that there are a variety of different views to be gained, and we should choose those which we wish to believe, in the light of all possibilities. Here it is argued that previous numerical solutions to this problem often encouraged even worse symptoms to emerge. The philosophy adopted here, is to ask how you amalgamate individuals rather than subdivide society. A logical unit of analysis does exist for the study of spacetime in human geography — it is a human life. As yet we have very little information on single people, but, at least from the census, the data is given at a resolution whereby, for national pictures what is produced would hardly appear any different from pictures drawn with the benefit of such information. For spatial data with a slightly less fine temporal resolution, what we have can appear as the full picture would, but somewhat blurred. As long as methods which depend critically on the spatial and temporal units (or units which would distort any method) are not adopted, such problems may well be circumvented9.

Social science does need maps; but the maps that are currently drawn in its name, apart from often being bad examples of physical geography's cartography, are bad social science. They make concentrations appear where they are not, and dissolve existing patterns. They rarely portray anything but the most simple of spatial distributions, certainly not spacetime evolution, or the interrelation of a dozen different influences (Print X). Here some of the particular solutions to mapping that social science requires are given. It is hoped that while a new methodology is being explained an alternative picture of Britain will develop through the subjects covered.

Prints

I Two images from the infinity of the Mandelbrot set (Colour).
II Land use close-up of Northern Britain (Colour).
III Journey to work flows of over ten people between wards from the 10% sample.
IV The changing distribution of housing by price, attributes and sales, 1983-1987.
V Migration flows between all regions in 1976 — flows sorted by contiguity order.
VI Yearly migration flows between English and Welsh wards 1980/1981.
VII The changing distribution of age and gender in Britain 1971-1981 (Colour).
VIII Voting composition on the electoral cartograms of Northern Britain (Colour).
IX Voting composition on the electoral cartograms of Southern Britain (Colour).
X The distribution of employment by industry, status and gender (Colour).

1 The advantage of maps is simple — they provide context:
Maps frown upon the isolation of single items. They preserve the continuity of the real world. They show things in their surroundings and therefore call for more active discernment on the part of the user, who is offered more than he came for; but the user is also being taught how to look at things intelligently. One aspect of looking at things intelligently is to look at them in context. [Arnheim R. 1976 p.5]

2 [a] The search for a definition of "maps" never ceases:
The current definition of cartography is inadequate largely because it does not define clearly the focus of the subject, namely maps. The description of maps is circular — "maps may be regarded as including all types of maps, charts, sections ...". This implies two types of maps, namely a subclass of specific forms, called maps, and a superclass of generic forms also called maps. The subclass of maps is defined as a "representation, normally to scale and on a flat medium, of a selection of material or abstract features on, or in relation to, the surface of the earth or of a celestial body" (ICA [International Cartographic Association], 1973, p.7). This second definition makes it clear that the subclass differs from its generic class in some ways. But, the two definitions taken together do not identify the common properties shared by all maps, which set them apart from artefacts which are not maps. [Visvalingam 1989 p.26]

[b] The most important aspect of definition concerns visualization:
For the ICA, oblivious to the contradiction inherent in its own definition, the end 'product' or cartographic process (the map) is to be 'visual, digital, or tactile'. Yet how can numbers, the constituents of what has been called, appropriately enough, the 'invisible map' be described as a map before they have been processed into an image (the visual map)? In following the politics of expediency rather than linguistic logic, and anxious to ward off (in the words of one President of the ICA) the threat of 'rapid submergence' by the new GIS-based technology, the ICA has managed to shoot itself in the foot. It has given the non-map parity with the map! [Harley J.B. 1990 p.16]

[c] An old definition is surprisingly apt:
One of the definitions of the word "map" that appears in the Oxford Dictionary dates from a source of 1586, where it was used to describe "a circumstantial account of a state of things ("circumstantial" is defined as "full of circumstances, details or minutiae"): not a bad objective 400 years later! [Bickmore D.P. 1975 p.328]

3 The term and philosophy of visualization did not appear overnight:
The medium of graphics has long been used to create two-dimensional representation of spatial phenomena for the primary purpose of visualization and, for many, this has also been the essence of "cartography". [Muehrcke P. 1972 p.27]

4 It is the great wealth of pattern and variation that is of interest in many of the pictures drawn in this dissertation:
The dismissal of geographical diversity as merely 'noise' or 'residuals' is a betrayal of what geography is. [Taylor P.J. 1991 p.24]

5 [a] The cartographic basis of physical geography dates from a time when land was all important and people had few rights:
Traditional cartography is seen as an optimal response to a highly constrained technology based largely on pen and paper. Although many of the conventions of manual cartography appear to be intelligent choices, they have nevertheless been made in an extremely restricted environment which imposes a limited view of reality. Early digital technology did little to broaden the constraints, and led cartography, map analysis and spatial analysis in different directions. More recent hardware and the results of intensive research have produced a digital cartography which can successfully emulate its analogue parent. However, its true potential lies in less conventional methods of analysis and display and in the degree to which it can escape its traditional constraints. [Goodchild M.F. 1988 p.311]

[b] Putting the argument less gracefully:
In physical geography, only that which has an effect on mankind is studied. Now that men are much less dependent on the countryside than on cities, why have geographers not followed mankind? Why have geographers left their minds back on the farm? [Bunge W. 1975 p.177]

[c] New computer systems often fail practically as well as theoretically:
There seems to be an inverse law where, as the sophistication of GIS software grows, the attention to basic principles of graphic design lessens. The emphasis is on getting something on the screen quickly rather than getting something on the screen that is meaningful. [Medyckyj-Scott D. 1991 p.21]

6 [a] The geographical features most of us recognise are not physical — we do not live near mountains:
Base data is so traditional that it invites a critical review. Consider the use of rivers on base maps. With the invention of bridges to cross them and railroads to compete with transportation on them, it could be argued that rivers have become unimportant enough to be eliminated from the map. They might be replaced by major railroad lines. In general, the traditional base map data is especially unsatisfactory to human geographers. Terrain features might be profitably dropped in favor of a surface of population density. The "continents" of population clusters on the Eastern shores of the United States, Western Europe, China and elsewhere are many times more important to the economic geographer than the distinction between land and water traditionally shown and memorized. Major cities are more important "islands" for many purposes than the atolls of the Pacific. It is probably true that of all the degrees of latitude and longitude shown on the map, only the equator and the poles are on the mental map and, therefore, the other degrees might be dropped as superfluous. Much of what has become traditional base map material might have been selected for no better reason than the ease with which the material could be gathered by early explorers. It is much easier to plot the continental outlines, rivers and mountain peaks than to obtain a census of population or an accurate map of arable land. [Bunge W. 1966 pp.45-46]

[b] A more human-based geography is being called for:
The inspiration may come not only from the field of geography / cartography but also from different fields of artistic endeavour, and lead to the design of maps of human activities which are much more vital than the thematic maps of today. Paradoxically, developments in computer technology may lead to the creation of maps which, when it comes to spontaneity and liveliness, have more in common with the popularly-admired and beloved, hand-drawn maps of the middle-ages and the renaissance than with strict, formalized cartography of the modern day.
However, certain conditions must be fulfilled for this to happen. Suitable cartographic data must be made available, and computers must be adapted to user needs in such a way that the technology does not impose itself between the user and his future map. An additional precondition is a revived interest in working with spatial / geographical problems, and a renewal of the skills involved in solving such problems by graphic means as well as in presenting these solutions in a creative way.
If these requirements are fulfilled — and the geographer / cartographer must assume a great deal of responsibility for this — a new era will be initiated for human cartography. [Szegö J. 1987 p.231]

7 [a] There are many sources of digital information about people in Britain:
The only nation-wide count of the population in Britain occurs at census year; the last two censuses were held in 1971 and 1981. With no 1976 or 1986 mid-term censuses, this information is currently produced only once every ten years. However, during the inter-censal years there are a number of other sources which can provide information on the changing socioeconomic, demographic and manpower characteristics of the population at the local scale ... [McKee C. 1989 p.1]

[b] Pre-eminent in all these data sets is the decennial census:
Some information on other characteristics of the population such as house-hold structure, employment status, ethnic composition and housing situation can be gleaned from the annual General Household and Labour Force surveys, but the problem of small sample size virtually rules out their use at scales below the Standard Region. As a result, the Population Census is not just the best, but in practice the only, source of reliable data on a reasonably wide range of demographic and socio-economic characteristics at sub-regional scale. Moreover, it has the advantage of providing data down to the level of the individual enumeration district covering roughly 500 inhabitants, which, even if too small for certain purposes, can be treated as a building block for areas specially defined by the user (Rhind, 1983). [Champion A. G. 1989 p.113]

[c] National and local election results also provide information:
In many ways elections are a positivist's dream. Millions of people go through the process of voting in numerous countries every year and these decisions are put together and published by areal units ready for analysis by social scientists. [Taylor P.J. 1978 p.153]

[d] We use whatever information is accessible:
In this book, votes receive rather more emphasis than other activities only because they have become the currency of political sociology rather than because they are more "special" or necessarily more legitimate than other activities. [Agnew J.A. 1987 p.6]

8 [a] The fundamental cartographic questions are:
"What to map?" "How to map?" "What to do with the maps?"
These three questions sum up the main problems connected with the mapping of population phenomena and statistical data generally. Each question gives birth to a brood of lesser questions, the lesser questions to a third generation, and so on. The outlines of this genealogy will be traced in the present paper.
I. WHAT TO MAP?
The offspring of "What to map? are (1) "What has been mapped?" (2) "What can be mapped?" and (3) "What should be mapped?" ... [Wright J.K. (ed.) 1938 p.1]

[b] Yet fewer and fewer people are asking these crucial questions:
Eavesdropping in the conference bar, the cartographer's chatter is of the virtuoso Macintosh rather than the question of why and what we map. Are the mechanics of the new technology so preoccupying that cartographers have lost interest in the meaning of what they represent? And in its social consequences? And in the evidence that maps themselves can be said to embody a social structure? If material efficiency is allowed to dominate the design and construction of maps, we can see why the ethical issues tend to pass unnoticed. The technology of Geographic Information Systems (GIS) becomes the message, not just the new form or medium of our knowledge. [Harley J.B. 1990 p.7]

[c] Questions can often be more important than answers:
It is surprising to learn that such a seemingly perverse world view is embraced by modern physicists. In the words of John Wheeler, one of the grand old men of physics, "No elementary phenomenon is a phenomenon until it is an observed phenomenon." By this, Wheeler means that the rise of quantum mechanics has demolished the view that the universe sits "out there" while we sit back and observe it. The kinds of questions one asks — and the order one asks them in — has a profound influence on the answer one gets, and on the world view one builds up. [Rucker R. 1984 p.193]

[d] What are we doing this for?:
The analytic power to order data has potential equally for control or liberation. It is all a matter of questions asked and interpretations made. [Taylor P.J. 1991 p.30]

9 [a] Many eminent cartographers have called for a change of approach:

A second challenge requires a greater effort by cartographers to escape from the constraints of euclidean space and to exercise more imagination and originality in producing maps. Barbara Petchenik (Chapter 3) makes a plea that we "... move our consideration from the domain of rationality and analysis to an exploration of the domain of synthetic intuition". The map is a designed object and in our concern with the "scientific basis" of cartography in recent years we may have lost sight of the need for more imaginative design. Here cartographers may have to learn from graphic arts. An increasing number of thematic maps are being produced by graphic artists, not by cartographers.
Part of the reason for this is that cartographers are a fairly conservative group and are still largely prisoners of euclidean space. Kishimoto (1980) recently drew attention to this fact. We are increasingly coming to accept the essential difference between the thematic map and the topographic map but have not yet accepted that locational accuracy is not always a basic requirement of the thematic map. We can more effectively and imaginatively map other "spaces" and give more emphasis to map content than to geographic location.
Here again, cartographers should take note of the work of psychologists like Arnheim (1975) and Norman and Rumelhart (1975) who argue that what a cartographer would regard as a "distortion" of the "real" euclidean space may in fact lead to an increase in map clarity.[Taylor D.R.F. 1983 p.288]

[b] Computer cartography could aid, rather than set back, this new approach:
Cartography in the information age will centre about a multifaceted model of geographic reality, the spatial data base. The challenge facing cartographers will be to devise the theories, methods, and techniques needed to collect, load, manage, and transform the data items into useable information. The new cartographic process will form a continuum of information flow that can be described in terms of the various generic functions of a spatial data processor. Technological advances will provide the potential for collection of vast quantities of basic spatial data. The distillation of the data into descriptions of geographic reality that we can understand will require a conception of the abstract modelling process used by a human to comprehend spatial entities. Processes to manipulate the data must bridge the gap between a user's perception and a computer's representation of spatial reality. Automated cartography will expand from its robot draftsman roots to a spatial information system using artificial intelligence techniques to allow the cartographer not only to produce cartographic products but also to convey the user-designed view of geographic reality. [Guptill S.C. & Starr L.E. 1984 p.14]

[c] What ever we do, we must always keep the basic reason we draw maps in mind:
If the student already carries in his mind's eye the image of a base map showing the boundaries of the administrative units by which statistics are tabulated, he may derive from a table of statistics a hazy idea of the form of a distribution. If no such picture is present in his mind he can gain no such concept whatever without the aid of a map. How many of us could picture the distribution of population in our own state by studying the census tables alone? Hence statistical maps are tools for the discovery of new truth. [Wright J.K. (ed.) 1938 p.16]