Abstract

"Metalinguists" depict the truth conditions of counterfactuals in terms of logical relationships between certain kinds of sentences. Possible-worlds theorists more candidly acknowledge unactualized possibilities, which metalinguists end up having to postulate anyway. However, similarity is not the right criterion for picking out the relevant antecedent-worlds. Rather, the relevant antecedent-worlds are those in which the appropriate ceteris paribus conditions obtain. ;True ceteris paribus conditions are importantly distinct from two other ways of qualifying natural laws: ideality and normality conditions. Ordinary counterfactuals involve less "back-tracking" than some philosophers have claimed. Control and randomization give experiments certain advantages, but problems can afflict even well-designed experiments. Inference from experiments to larger populations should be seen as a one-step, rather than a two-step, process of induction. ;Disputes arose over ceteris paribus conditions for experiments on character displacement and null models for island biogeography. The first dispute reminds us that some ceteris paribus conditions are incompatible with others. The second controversy illustrates the consequences of holding certain "other" factors constant, when they should be allowed to vary. Methodological critiques are not as likely to unseat a ruling paradigm like competition theory as are alternative explanations. The methodological paradigms of Popperian falsificationism and standard statistical hypothesis testing are incompatible with each other. ;Another dispute arose over ceteris paribus conditions for mathematical models of the relationship between the diversity and stability of ecological communities. This demonstrates that mathematical models involve some choices about ceteris paribus conditions that actual experiments do not. Special problems arise because of the possibility that stability is adaptive at the community level. Different modes of visual representation may have led scientists to adopt different ceteris paribus assumptions in this case. ;One rule of thumb for deciding between competing definitions of scientific terms is: Regard with suspicion any definition that turns a seemingly important empirical matter into an a priori exercise. Some, but not all, definitions of ecological "stability" are suspect according to this rule. The faulty definitions may have resulted from an over-emphasis on population dynamics. Machine metaphors of nature may have induced a related problem of experimental design