Reviews in Fish Biology and Fisheries 9: 117–123, 1999.
The EcologicalDetective: Confronting Models with
Data (Monographs in Population Biology, 28). By
Ray Hilborn and Marc Mangel. Princeton Univer-
sity Press, Princeton, NJ and Chichester, UK, 1997.
ISBN 0-691-03496-6 (hard cover), 0-691-03497-4
(p/b, £16.95). Acid-free paper, pp. xvii + 316, 21
tables, 5 boxes, 54 ﬁgures.
The Ecological Detective aims to bridge the gap
between ecological models (and modellers) and data.
Each reader will form a different opinion of this
book, largely depending upon their existing modelling
background. For example, the preface summarizes
comments from readers of earlier drafts, and these
span the range from “too hard” to “too simple”.
The reader is expected to be computer-literate and
is encouraged to implement the computational algo-
rithms associated with the examples.
Chapter 1 introduces the notions of models,
hypothesis testing, likelihood, and the Bayesian
approach. Chapter 2 begins with a historical overview
of four different philosophies to modelling, from
which the authors argue that the process of modelling
is a contest between competing hypotheses and data,
rather than betweena single hypotheses and data. They
also express a preference for the Bayesian approach to
modelling. (The reader should not be concerned that
this is a Bayesian book – methodology is developed
following the classical approach, with the Bayesian
approach viewed as a subsequent modiﬁcation.) The
1897 paper of Chamberlain, on the method of multiple
working hypotheses, is provided in an appendix.
The third chapter is the largest in the book.
It is primarily devoted to statistical notions such
as randomness, probability distributions and density
functions, random variables and Bayes theorem, and
covers some of the most widely used statistical distri-
butions in moderate details. Useful tools (e.g. the delta
method, bootstrapping) are brieﬂy covered and the
connection with ecological modelling is maintained
via examples. Full comprehension of this chapter
requires a working knowledge of derivatives, Taylor
series expansions, and integration.
After Chapter 3, the chapters alternate between
in-depth examples and methodology. The four exam-
ples are ‘Incidental catch in ﬁsheries: seabirds in the
New Zealand squid trawl ﬁshery’ (Chapter 4), ‘The
evolutionary ecology of insect oviposition behaviour’
(Chapter 6), ‘Conservation biology of wildebeest in
the Serengeti’ (Chapter 8), and ‘Management of hake
ﬁsheries in Namibia’ (Chapter 10). The alternating
methodology chapters are ‘The confrontation: sum
of squares’ (Chapter 5), ‘The confrontation: likeli-
hood and maximum likelihood’ (Chapter 7), ‘The
confrontation: Bayesian goodness of ﬁt’ (Chapter 9),
‘The confrontation: understanding how the best ﬁt is
found’ (Chapter 11).
In the concluding paragraph of the introductory
chapter, the reader is warned, “We are practicing ecol-
ogists. We are not statisticians, numerical analysts,
or philosophers, and the appropriate chapters will no
doubt offend the appropriate experts.” This is too
strong a statement, but it does voice the fact that the
line that separates acceptable modelling from unac-
ceptable modelling is one that is very subjectively
placed. The comments below arise in part from my
placing the line somewhatdifferently from the authors.
I was concerned by Chapter 5, which presents
the combination of sums of squares and bootstrap-
ping as an omnibus method of modelling, and the
ensuing application of this in Chapter 6. Perhaps
this is a manifestation of the last sentence of the
introductory chapter, “And if the techniques do not
exist, then we must invent them.” However, Chapter
5 is little concerned with whether a technique already
exists. This omnibus approach certainly has its role in
modelling, but it should not be used blindly. Care must
be taken that the reader is not encouraged to invent
a technique when there is already a better approach
available. For example, in the ﬁsheries literature,