Individual variation in aquatic toxicology: Not only unwanted noise.

The mean value of any parameter and its changes are usually discussed, when ecotoxicological studies are carried out. However, also the variation of any parameter and its changes can be important components of the responses to environmental contamination. Although the homogeneity of variances is commonly tested, testing is done for the use of correct statistical methods, not because of exploring the possibility that variability and its changes could be important components of environmental responses. We evaluated recent aquatic toxicological literature and found that in the majority of articles indicating that homogeneity of variances was tested and giving the result of testing, the assumption of homogeneity was not fulfilled. Further, it was observed that in some studies experimental treatment clearly affected the variability. In this commentary we discuss the reasons for variability: measurement errors, experimental design, genetic heterogeneity and phenotypic plasticity, and conclude that even after accounting for experimental design and genetic makeup significant variability remains. This plasticity may change in environmental responses as suggested by a hypothetical example, and as confirmed by experimental data. As a consequence, the changes of variability can be significant, even when the means do not differ. Because of this, variability and its changes should always be analysed and reported. This will be easy, since the datasets are exactly the same for comparing the variances and means, and as normally variances are tested for homogeneity. It is likely that much new information about the responses of organisms to environmental contamination will be obtained. However, the present journal practises tend to discourage one from concentrating on anything but the mean. In contrast, we think it is imperative that variability is always included as an endpoint in data analysis in the future.