The Future of Geography (RLE Social & Cultural Geography)

Ron Johnston

• Publisher: Routledge
• Year: 2014
• ISBN:
  9781317907121
• Edition: 1
• Topic: Physical Sciences
• Subtopic: Geography
• Index: Geography

The chapters in this book address fundamental questions of the nature and purpose of geography, scrutinising its contents, philosophy and methodology.

Aimed at undergraduates its purpose is to broaden the debate about what geography had become during the 1980s and what shape it might take in the future.

---

III

GEOGRAPHY FOR SOCIETY

---

11

QUANTIFICATION AND RELEVANCE¹

____________________________R.J. Bennett

Geography, like many of the social sciences, went through a form of quantitative revolution in the late 1950s and 1960s. In retrospect this can be seen as part of the post-war surge in interest in technological innovation, technical approaches to management, and planning. In the 1970s, rather later than some other disciplines, geography experienced an attempt at a ‘radical’ revolution. The intention in this case was to displace supposed value-free approaches to scientific method by ones which explicitly acknowledged a value system based upon, in most cases, a theory of labour value. Much of the early vigour of this radical group of geographers was concentrated on attacking quantitative geography and what was termed ‘spatial science’. The 1980s, however, have seen the emergence of a more mature approach by both quantitative workers and their critics. The thrust of applied research has played a critical part in the formation of this new base. It is thus highly pertinent to take up in this chapter, if only briefly, the interrelated questions of quantification, relevance and utility. In doing this I wish to develop three main points, and from these to derive a fourth:

1 The necessity for a ‘paradigm shift’ in geography as a whole away from the view that the core of the discipline is framed by spatial geometry.

2 The necessity of reappraising the nature and appropriateness of the methods of statistical inference within quantitative geography, especially those derived from the statistical theories of Neyman and Pearson.

3 The need of geography to respond to the increasing stimuli provided by ‘applied’, social, political and policy issues which are tending to shift the emphasis of the subject towards frame disciplines and away from a central geographical core.

4 The need to put in place an appropriate response to the critique of quantitative geography as positivism. This is a somewhat separate issue from the previous three, but derives from each of them.

I hope that not too much injustice will be done to each theme by a discussion that, in the limited space available, is necessarily brief.

PARADIGM SHIFT AWAY FROM SPATIAL GEOMETRY

William Bunge, in his 1962 Theoretical Geography, introduced the concept of structuring geographical data in terms of their elementary geometrical properties of point, line, flow, zone and surface. Much of Bunge's thinking was inspired by Fred K. Schaefer who, in a seminal paper in 1953 and in writings reported in Bunge (1962, 1966), propounded the thesis that spatial patterns were morphological laws. The proposal was that generalities of location link unique relative positions into geometrical and topological regularities which apply across space in a repetitive and predictable form. This was part of a vehement attack on what he saw as a stultifying tradition emphasizing the uniqueness of location and undermining the capacity to propound generalizations. Schaefer's influence, however, together with that of many other writers of this period, stimulated an approach to spatial data primarily emphasizing the inductive search for pattern which was matched against prior theory derived from geometric and topological properties of space. Both Schaefer and Bunge exerted a considerable influence on Peter Haggett's 1965 Locational Analysis in Human Geography, which came to dominate much subsequent thinking, especially in the UK. The objectives of this lineage of work are the detection and analysis of spatial pattern, and it is this lineage which is most characteristically termed ‘spatial science’. Spatial science is clearly defined by Haggett et al. (1977, 1) as (1) describing aggregate arrangements in space, (2) methods for detecting these arrangements and (3) using this information in applications.

This approach to spatial problems using the structures of point, line, flow, zone and surface has immensely stimulated geographical research, but it has tended to divert attention from analysis of spatial process, and to emphasize inductivism.

As a result of dissatisfaction with this emphasis on spatial pattern and inductivism, many researchers have sought alternative frameworks to attack geographical problems. At the extremes these alternatives have been based on an outright rejection of spatial science. For example, in human geography the spatial view has been characterized as a form of fetishism which obscures the more fundamental social questions. This view is vehemently pressed by some Marxist sociologists. For example Castells (1976) claims that there is no specifically spatial theory that is not part of a more general social theory; in other words human geography has no valid methodological base within the social sciences. For physical geography the rejection of spatial science comes from a different direction, but is no less extreme. There, the methodology of other subjects has increasingly overridden any rather weak spatial view the geographer might adopt. In particular, the criteria of intellectual status have often dictated publication of research findings in journals outside geography, and this has introduced non-geographical criteria into research in physical geography. Whilst useful in many ways, this influence has eroded the geographical core of the discipline.

Apart from methodological critiques, this separate development of human and physical geography erodes the unity of the subject. Where physical and human issues were frequently linked as part of population-environment concerns, now many human geographers grasp the methods of social science which, in the most extreme Marxist and critical theorist cases, reject any substantial role for the influence of the physical environment; and the physical geographers largely seek theoretical development divorced from social and human needs.

It is now accepted by most geographers that the spatial geometric base of spatial science has little major utility, except in a few special situations. However, what must remain very relevant to spatial science are spatial relations, spatial processes, the significance of place, and place differences. To take as an example the definition of a spatial process, this is now usually interpreted not in terms of geometric or topological situations, but instead as one realization of an underlying generating process. Part of the modern position is stated well by Haining (1981). He distinguishes two levels of conception. At the first level is the mathematical theory of spatial process, which is expressed in terms of a structure of variables, relationships between variables and the environment of variables (or parameters). The value of a variable then defines the system, and a process is defined by the rules governing the changes over time in the system as a chain of events. Spatial pattern is the evidence of the second level, the map pattern of a single realization of the underlying spatial process. This is the surface of relations constituting the data which can be used for empirical analysis. Analysis is thus founded on analysing the generating process, with the map pattern constituting the data used to assess realizations, not vice versa. A wider view of spatial process developed by Sack (1974, 1980) is discussed more fully in my conclusion.

REAPPRAISAL OF STATISTICAL INFERENCE

Associated with the approaches of spatial science, but not essential to it, was an overburdening emphasis on statistical inference in much geographical research of the 1960s and 1970s. In practice many quantitative researchers never used methods of inference, but a number of widely quoted papers and all the text-books on quantitative geography emphasized this approach, so that inference became the dominantly voiced methodology. Certainly it dominated to the extent that Harvey, in his Explanation in Geography (1969), represented it as the paradigm within which geography had come to work, and Gould (1970) attacked it by asking whether statistix inferens was the name of a geographical ‘wild goose’. Wrapped up in the difficulties of spatial science are many of the problems which result from an unthinking use of statistical inference; and the emphasis upon inference again allowed a more ready conjunction of scientific method and positivist methodology, as discussed later. It is now clear that emphasizing inference per se produced numerous difficulties.

Within all scientific method there is a complex interplay between induction and deduction as the two main approaches by which theories are developed about phenomena. This applies to most physical and social science. In general the two approaches must be seen as inseparable so that, although we may choose to emphasize a more inductive style of argument in one case and a more deductive style in another, it is inconceivable that the development of theory is seen as solely an inductive or deductive exercise. Instead the best of geographical analysis draws deeply upon both the analyst's inductive experience of empirical reality and upon his prior deductive theory. It is to be hoped that he will be sufficiently emancipated from each to respond both to reality and to the philosophical positions of his theory. In its simplest terms the division arising from an emphasis on either the inductive or the deductive approach to statistics has given rise to two major schools of statistical theory. Both lead to similar estimators, but there is an important line of division in the premisses of their approaches.

The first of these approaches is the ‘classical’ theory of inferential statistics and hypothesis-testing as developed by Neyman and Pearson in the 1920s. This often appears to be primarily inductive. It has a ‘frequency’ interpretation in which probabilities can be asociated only with events arising from repeatable experiments. Hence probability is interpreted as relative frequency over a large number of repetitions of an experiment. The theory as developed by Neyman and Pearson has the properties of a decision theory. A prior cut-off is specified by statistical significance level, which defines a critical region. Then, under a specified null hypothesis and test statistic, a calculated result for the test statistic yields a result which permits a decision on whether a sample statistic accords or does not accord with the properties of a parent population. For such an approach, various, usually rather rigorous, assumptions are required. Particularly important are (1) that the sample is random, (2) that the decision experience is repeatable and (3) that the sample is not the entire population.

The second main approach in statistical theory is that utilizing Bayesian methods. In this approach the intention is not to reach a decision about whether to accept or reject a null hypothesis, but instead to reach a conclusion as to the degrees of belief which can be placed in a given statement. The outcome of this approach is a posterior probability density function (pdf) which expresses the degree of belief. This is based on the prior pdf, the unconditional pdf and the conditional pdf or likelihood function. The prior pdf is the best estimate we have at the start of the experiment. It incorporates a priori information and theory and places no restrictions on the form of information that can be employed. It thus overcomes many of the restrictive, statistical assumptions of the Neyman and Pearson theory. The unconditional pdf is the observed distribution of the data for all possible probabilities of given parameters that might affect the situation under study. The conditional pdf, also termed the likelihood function, is, perhaps, the most important element, and it expresses the probability of given experimental outcome in terms of the unknown parameters. Hence it gives us the entire evidence of our experiment in an empirical situation.

The advantage of the Bayesian theory is that it is a unified approach. It concentrates attention on the practical process of research design and refinement of hypothesis, which it regards as a sequential process moving not to a final decision (as in the Neyman and Pearson theory) but to a level of support for a particular set of ideas. It is thus symmetrical to the view of geography as an inductive-deductive process of refining theory. The method has wide application since the approach does not rely on rigid statistical assumptions, on large sample properties, or on precise repeatability of experiments. Hence, it has a greater appropriateness to most geographical problems than the Neyman and Pearson theory. Yet, despite its advantages, Bayesian theory has been developed to a much lesser extent in geography and planning than the theory of statistical inference. (A recent exception is the approach of Wilson and Bennnett 1985.)

The Fisher approach to statistical analysis also leads to a level of support for ...

Table of contents

• Cover Page
• Half Title page
• Title Page
• Copyright Page
• Original Title Page
• Original Copyright Page
• Contents
• List of illustrations
• Notes on contributors
• Acknowledgements
• Preface
• I The Content of Geography
• II Quantum Neurodynamics
• III Geography for Society
• Index