The problems of relating the results of experiments in the laboratory to events in nature are twofold: to equate the response to a single variable (hydrocarbons) with the natural variability in the biological material in a multivariate environment, and to consider whether the response established experimentally has any relevance to the animal's chances of survival and reproduction (i.e. its fitness) in the natural population. Recent studies of the effects of petroleum hydrocarbons on marine invertebrates are reviewed, with an emphasis on the physiological and cytochemical responses by bivalve molluscs. The dose-response relations that emerge suggest the intensity of the 'signal' that must be detected in nature if the chronic, sublethal effects of petroleum pollution are to be measured. The natural variability in these physiological and cytochemical processes are then reviewed and the main causes of variability in natural populations, both endogenous and exogenous, discussed. These results indicate the extent of the `noise' above which the signal from possible pollution effects must be detected. The results from recent field studies on the common mussel, Mytilus edulis, are discussed. The results are as complex as expected, but it proves possible to reduce the variance in the measured responses so that pollution effects, including those due to hydrocarbons, can be detected. The ecological consequences of the observed effects of petroleum hydrocarbons are then discussed in terms of reproductive effort and reproductive value. Considerable variation between populations exists here also and this can be used to help in the interpretation of the extent of the impact of the environment on the ecology of the population. The result is to place the findings of the laboratory experiments in an ecological context of natural variability and of the physiological costs of adaptation.