Theoretical models suggest that in natural communities complexity bege
ts instability. The predicted negative hyperbolic relationship between
connectance - the ratio of the number of realised interspecific inter
actions to the number of possible interactions - and the number of spe
cies of the community has been reported for many different systems. Ho
wever, we argue that these tests suffered from analytical and conceptu
al flaws, and that this and other predictions in community ecology can
be better approached using allometry. Four community datasets which a
re available in the literature (two for food-webs, one for plants and
their pollinators, and one for plants and their seed-dispersers) were
reanalysed to illustrate the use of the community allometry technique.
One of the four tests indicated that the number of realised interacti
ons is isometric in relation to the number of possible interactions, i
ndicating that connectedness does not decrease with increasing communi
ty size. The remaining datasets produced negative allometry, ostensibl
y meaning that larger communities tend to be more loosely connected th
an smaller communities; however, the dataset showing isometry is the o
nly one with a standardised sampling procedure, and this emphasises th
e importance of considering sampling bias. Community allometry was als
o applied to an additional food-web dataset to demonstrate how the met
hod can be extended to test other community structure hypotheses. Fina
lly, the complexity-stability theory itself is briefly discussed - spe
cifically, it is observed that the concept of connectance in the theor
y differs from that in empirical studies.