structure and function may not be manifest in a single season or even a single year, it is important
for experimental/manipulative studies to monitor effects in the field with a frequency and duration
that are sufficient to capture these changes. This may require monitoring sites more than once per
year and a study duration in excess of 2 years (Underwood 1993).
Because of the spatial heterogeneity of contaminants in correlative studies and the inherent
variability associated with measuring responses in the field, it is important to have a replicated
design. A minimum of two, preferably three, replicate measurements should be made to ensure that
statistically meaningful differences are detected (Sanders 1984; Underwood 1993).
Since most field validation efforts are designed to evaluate contaminant-induced effects, it is also
important to account for factors that may influence the availability of the contaminants. Establishing
a correlation between measured contaminant levels and responses in either field- or laboratory-
exposed organisms can be facilitated by normalizing for factors that control bioavailability. Most
field validation studies performed to date have normalized concentrations of organic contaminants
to organic carbon levels present in the sediment. Similarly, selected divalent metals can be
normalized to the concentration of acid volatile sulfides. A separate but related issue, especially
for correlative studies, is accounting for differences in the magnitude and ratio of individual
contaminants present in the typical milieu of contaminants found in sediments across a range of
sites. Typically, quotient approaches are used in an attempt to account for these differences.
Quotient approaches assume additivity and have perhaps the greatest utility when they are used to
evaluate concentrations within a contaminant class (e.g., PAHs). The quotient approach accounts
for differences in the potency of the individual constituents within a given contaminant class that
has the same mode of action. For example, dioxins (i.e., tetrachloro-dibenzo-p-dioxins (TCDD) and
dibenzofurans) can be normalized to 2,3,7,8 TCDD toxic equivalents to account for changes in the
type and amount of the 14 individual dibenzodioxins and dibenzofuran congeners over a range of
sites. Similarly the ΣPAH model of Swartz et al. (1995) can be used to normalize for the types and
amount of PAHs over a range of sites. When multiple contaminant classes are present, the ERM-Q
approach of Long et al. (1995) can be used to provide qualitative information on the level of
contamination and associated toxicity over a range of sites.
The data of Swartz et al. (1994) and McGee and Fisher (1999) showed no significant differences in
variety of benthic community measurements along a contamination gradient that produced signifi-
cant toxicity in laboratory bioassays. These differences may be accounted for in the type of
organisms colonizing the sites as they have been either less sensitive or less exposed (i.e.,
tube-dwellers versus free-burrowing organisms) than the organisms used in the laboratory bioassay.
Another possibility is that resident organisms may have adapted to the contaminant loads over time.
In either case, field studies and correlative studies in particular must carefully consider the potential
masking of contaminant effects by factors such as these. As Swartz et al. (1985) point out in their
study of the Palos Verdes outfall, the absence of a specific taxa (e.g., phoxocephalid amphipods)
may be as important as more standard measures of benthic community analysis. A comparison of
resident and laboratory test organisms of the same species via toxicity testing (reference toxicant
or whole sediment) could shed light on whether an adaptive community has evolved in the field.
In correlative/observational studies, it is important to document and account for the distribution of
physicochemical features (e.g., current flow and direction, water depth) of the sampling sites along