The Dictionary of Human Geography

complete sets of joint and marginal posterior
distributions may be constructed by thissimu-
lationmethod. Starting from the (diffuse)
prior distributions, the conditional distribu-
tions of theu-parameters are sampled using
randomsampling(hence the ‘Monte Carlo’
part of the name) and these sampleduvalues
are then brought together with the data to
estimate the likelihood. This information is
then used to update the conditional distri-
butions, and the process is repeated many
hundreds or thousands of times, gradually
building up samples representing the posterior
distribution. The sampling at any stagetis
based on updating from the conditionals at
timet1, and hence it is a (first-order)mar-
kov chain process. This remarkable method
can be developed for very complicated models
with many parameters and difficult structures,
and is being used in many disciplines. Models
in bothspatial econometrics andmulti-
level modellingmay now be estimated by
these Bayesian methods.
Bayesian methods have been applied in
several areas of geographical andspatial an-
alysis. The specific version of ‘empirical
Bayes’ estimation is widely used in spatial
interpolation and in disease mapping and spa-
tialepidemiology. Other applications include
population and economic forecasting, crime
‘hotspot’ modelling, and hierarchical Bayes
estimation to lend insight into the problem
ofecological inference. Brunsdon (2001)
provides a case-study using MCMC in a Baye-
sian model predicting school performance fig-
ures for pupil-tests. Inspatial econometrics,
Bayesian methods, both numerical integration
and MCMC, have been used to estimate
models with spatialendogeneityand more
complicated forms (Hepple, 1995; LeSage,
1997). The geographer Peter Congdon has
written two major statistical texts on Bayesian
statistical modelling (Congdon, 2001). Opin-
ions still differ about the role of subjective
prior information, but modern Bayesian
methods are one of the fastest developing
areas of quantitative analysis. lwh

Suggested reading
Withers (2002).

behavioural geography A sub-discipline
emphasizing the psychological underpinnings
of individual spatial behaviour; in particular,
the cognitive and decision-making processes
that intervene between a complex environment
and human action. In its earliest expression this
work was more humanistic, exemplified in the

historical musings of J.K. Wright in the 1940s

(Keighren, 2005), and the influential essays of

Lowenthal (1961) and Brookfield (1969) on

environmentalexperience andperception.

While this tradition led into humanistic

geography, behavioural geography was typi-

cally more formal and analytic, drawn into

thepositivistparadigm oflocational analy-

sis. Its characteristic question was: Given the

assumption of rational behaviour, why did an

actual location or pattern of spatial behaviour

depart from an optimal form? (Seelocation

theory.) The answer was seen to be a product

ofdecision-making, and notably the human

tendency to have only incomplete information,

to make imperfect choices, and to be satisfied

with sub-optimal options. Applications in-

cluded Wolpert’s (1964) study of Swedish

farmers and Pred’s (1967) analysis of indus-

trial location. In each instance, behaviour was

seen to besatisficingrather than optimizing

as predicted, for decision-makers were not

only incapable but even unwilling to com-

promise other values in order to maximize

their utility functions. Similar work examined

the journey to shop, and showed again how,

both in terms of retail location and shopping

behaviour, cognitive variables intervened to

complicate geographically rational behaviour

(seeretailing). A particular emphasis was

upon preference structures in spatial behav-

iour, modelling such topics as place utility

and residential search. The most celebrated

work was conducted by Peter Gould and his

students who examined themental maps,or

preference surfaces, within different countries

held, usually by students, and which might

permit the prediction of subsequentmigra-

tion(Gould and White, 1993 [1974]).

One of the most interesting and applied

aspects of behavioural geography was work

examining human perception ofenvironmen-

tal hazards. Typically, this research ad-

dressed itself to a seemingly anachronistic

location decision. Why did people or industry

locate in unpredictable sites such as flood-

plains or areas of earthquake or avalanche

hazard? How was such irrational behaviour to

be explained? The pioneering work by Robert

Kates and Gilbert White on floodplain hazards

inspired many subsequent studies, which

included increasing methodological sophisti-

cation. For example, Saarinen’s (1966)

innovative study of the perception of drought

hazard by farmers on the Great Plains postu-

lated the existence of a distinctive personality

disposition, which he explored using the

thematic apperception test, a personality

Gregory / The Dictionary of Human Geography 9781405132879_4_B Final Proof page 44 31.3.2009 11:01am

BEHAVIOURAL GEOGRAPHY