It is sometimes possible to substitute one species of animal for another. For example, lower vertebrates or invertebrates may be substituted for higher vertebrates, or so-called "laboratory species" (such as mice or rats) may be substituted for "companion animals" such as dogs or cats. Such substitutions are usually advocated on the grounds that species differ in their capacity to suffer pain or distress and it is assumed that invertebrates or lower vertebrates will suffer less than higher vertebrates, and laboratory species less than companion species.
If the research is to yield useful results, the animal species selected must be the one which will fulfill the requirements of the model and closely mimic the condition being studied. The relative capacity of different species to suffer pain or distress is intensely debated by animal welfare advocates and scientists alike. In addition, selection of a particular species for a research project is constrained by many considerations and there may be few or no other species which satisfy all the requirements of the model. In biomedical research, where the models are often of human diseases, invertebrates or lower vertebrates may share so few relevant characteristics with humans that substitution is impossible. In basic biological research, the problem being studied may be specific to a particular species or group of species and may not occur in other groups.
Plants and microorganisms have been suggested as substitutes for animals. For example, Salmonella is used in mechanistic studies in genetics, and the active steroid hormones found in yeasts are used in some endocrinological and immunological research. Use of plants and microorganisms as substitutes is limited by their evolutionary differences from humans and other higher vertebrates, and by their own unique characteristics.
Improvement in experimental design, or in the statistical analyses of results, may reduce the number of animals needed. Knowledge of statistics, or consultation with a statistician, can improve experimental design by including consideration of such factors as randomization, confounding variables, sample size and statistical power, and the problems associated with testing multiple hypotheses (Geller, 1983). Animals are expensive to use, so experimenters usually employ the minimum number of subjects, raising the possibility that improved design or statistics may increase, rather than decrease, the number of animals used. If too few animals are used, experimental results may be statistically invalid and therefore useless. Improved design and analysis could reduce animal use if more robust or clearer results reduced the total number of experiments, or eliminated unnecessary duplication of experiments (Still, 1982). Reproducibility of results is a key component of the scientific method. Therefore, all repetition of experiments cannot, and should not, be eliminated.
Alternative designs may reduce the numbers of animals used. For example, a two-stage screening process may replace a single pass design, cross-over designs may be able to replace parallel designs for studies of short-term effects, or group sequential controls can be used in which subjects are divided into equal sized groups and the experiment performed in stages with additional groups being added until cumulative differences reach significance (Elashoff & Beal, 1976; Geller, 1983).
In some cases, experimental design can be modified to reduce or eliminate pain and distress of the animal subject. For example, noninvasive systems for sampling physiological parameters are sometimes used to replace invasive methods. MRI and PET imaging techniques can not only allow investigators to study organ systems noninvasively, but also can allow individual animals to be studied longitudinally, thus reducing the number of animals required. In behavioral research there is some evidence that allowing a subject increased control over painful or unpleasant stimuli reduces distress for some species (e.g., Lea, 1979).