(organisms sedentary on woody debris in the channel) of the warm water
Ogeechee River (Georgia, USA) was inconsistent with the theory (scaling expo-
nent of M and P/B, 0.50), though this discrepancy could be accounted for in
various ways. Thus, for Huryn and Benke’s data on stream benthos, metabolic
theory does fairly well.
Several other authors found encouraging, though patchy, fits of pattern with
metabolic theory in different contexts (Petchey, Long & Morin, this volume;
Humphries, this volume; Hall et al ., this volume; Jennings & Reynolds, this
volume), leaving us with the intellectually awkward task of judging whether
the theory survives because we can explain the discrepancies, or if the theory
falls in general. The problems of testing theory in this way are several and were
vigorously debated at the meeting, and there seem to be three major issues
involved. First, testing has essentially been apost-hocexercise in which data
gathered for completely different purposes are used to challenge the theory,
always a tricky process. Second, different measures of body size were used in the
present collection of chapters: mass or weight (dry, ash-free, wet, whole-body or
elemental), volume or equivalent spherical diameter, individual, average or
adult body size. Different measures will be appropriate for different research
questions and in many cases a clear rationale for selecting a particular measure
of body size is presented. For other studies, however, including many macro-
ecological investigations, the data used are simply those at hand, often adult
body sizes, which can be extracted from field guides or species survey lists. If
patterns that emerge from the analyses of such data are interpreted in the
context of theory that assumes other expressions of body size, such as individual
biomass, then the value of the research is compromised. One solution to this
potential problem is to make available a much greater variety of body-size
measurements for a broad range of systems within a public domain database,
so that analyses can be tailored more closely to theory.
Third, body-size data and relationships are presented in a variety of forms
within the present volume. Several chapters reported bi- or tri-variate relation-
ships, with body size as the explanatory variable, often seeking to allow compar-
isons with theory (mainly as judged by slopes of regression lines). However,
testing for deviation from theoretical relationships opens up a can of worms. For
instance, the choice of regression models (least squares, geometric mean, bisec-
tor regressions and others) is important because each estimates a different value
for the slope. Similarly, it is not good enough to assume that, because a slope
through the observed data is not significantly different from the theoretical
prediction, the data are consistent with theory. An alternative explanation is
that there is not enough power (essentially the number of data points) in the
analysis to falsify the model, and formal power analyses should be carried out to
check this, although this is rarely done. More suitable statistical procedures
to demonstrate negligible trends in data are now available (Dixon & Pechmann,328 A. G. HILDREWETAL.