The ABC method for evaluating pollution-induced stress was tested usin
g data from the Chesapeake Bay, Virginia, collected between 1985 and 1
989. Three predictions were tested: (1) benthic communities from estua
rine transitional regions with salinities near the range of 5 to 8 par
ts per thousand (horohalinicium) should be classified highly stressed
due to major shifts in ionic composition producing physiological stres
s; (2) benthic communities from regions subjected to summer low dissol
ved oxygen conditions (anoxia or hypoxia) should be classified as high
ly stressed after such events; and (3) benthic communities from sedime
nts contaminated with heavy metals and polynuclear aromatic hydrocarbo
ns should be classified as highly stressed. Only partial support for e
ach of these predictions was found and several problems with the ABC m
ethod were obvious. A small number of large-sized species, particularl
y in mesohaline and polyhaline regions of the estuary, greatly affecte
d the analysis. Similar designations of stress could be produced by si
mply sampling only for these rare, large species. Regions of the estua
ry considered a priori as highly stressed were sometimes designated as
unstressed due to (1) minor shifts in dominance patterns in benthic c
ommunities with low absolute numbers of individuals and biomass, e.g.
in regions affected by anoxia/hypoxia, and (2) collection of rare, but
large species, such as the tubiculous polychaete, Diopatra cuprea, in
contaminated sediments. Regions of the estuary considered a priori as
unstressed were sometimes designated as highly stressed due to dense
recruitment events. Contrary to assumptions of the ABC method, increas
ed sample size (replication) may result in the collection of rare, lar
ge-sized individuals in highly stressed communities. Partial dominance
curves were applied to the data and (1) removed the effect of biomass
dominants in contaminated sediments changing the classification of co
mmunities from unstressed to stressed, (2) did not change the stressed
classification due to dense recruitment events, and (3) changed the c
lassification of mesohaline and polyhaline communities from unstressed
to stressed, even in the absence of low dissolved oxygen events or co
ntaminated sediments. No single method or analysis is likely to produc
e stress classifications without unacceptable misclassifications. We p
ropose that ecological stress, from any source, is best measured using
multiple methods or analyses with different assumptions. The consiste
ncy of classification between different approaches would provide the r
obustness necessary to judge the reliability of a stress classificatio
n.