In Defence of Hypothesis Testing in Ecology

In two recent scientific meetings I have attended (which must remain nameless to protect the innocent), I have found myself wondering about the state of hypothesis testing in ecological science. I have always assumed that science consists of testing hypotheses, yet I would estimate roughly that 75% of the talks I have been able to attend showed no sign of any hypothesis. I need to qualify that. Some of these studies are completely descriptive – what species of ferns occur in national park X? Much effort now is devoted to sequencing genomes, the ultimate in descriptive biology. This kind of research work can be classified as alpha-biology, basic description which is necessary before any problems can be formulated. In my particular specialty of population cycles in mammals, much descriptive work had to be carried out to recognize the phenomenon of “cycles”. But then the question arises – at what point should we stop simple descriptions of mammal populations rising and falling? Do we need to study the dynamics of every rodent species that exists? Or in genetics, is our objective to sequence the genome of every species on earth? My point is that after we have enough basic description, we should move into hypothesis testing, or asking why some phenomenon occurs, the mechanisms behind the simple observations. The important point here is that we should not have a single hypothesis or explanation for any set of observations but rather several alternative hypotheses. As a simple example, if we find our favourite plant species is declining in abundance, we should not simply try to connect this decline with climatic warming without having a series of alternative explanations with the emphasis that our observations or experiments should be capable of distinguishing among the alternative hypotheses.

The alternative argument is that we do not know enough about ecological systems to set up a series of credible alternative hypotheses. It is quite possible to go on describing events endlessly in science in the hope that some wisdom will emerge. I do not think this is a profitable use of time or money in science. In ecology in particular I would argue that there is not a single question one can ask that cannot be answered by at least 2 or 3 different mechanistic hypotheses. Our job is to articulate these alternatives and to do whatever studies or experiments are needed to distinguish among them. Of course it is always possible that the correct answer is not among the 2 or 3 hypotheses we suggest at the start of an investigation, and this is often why one study leads to a further one. Consequently we cannot accept statements like “I have no idea why this observation has occurred”. Such a statement means you have not thought deeply enough about what you are studying. Ecological surprises certainly occur while we study any particular community or ecosystem, but we know enough now to suggest several possible mechanisms by which any ecological surprise might be generated.

So I think it incumbent on every ecologist to ask (1) what is the problem or question my research is addressing? And (2) what probable mechanisms can be invoked as the cause of this problem or the answer to this question. Vagueness may be a virtue in politics but it is not a virtue in science. And I look forward to future conferences in which every paper specifies a precise hypothesis and alternative hypotheses. Chamberlin (1897) stated the case for multiple hypotheses, Karl Popper (1963) asked very specifically what your hypothesis forbids from happening, and John Platt (1964) pulled it together in a critical paper. There was important work done before the Iphone was invented. Good reading.

Chamberlin, T. C. 1897. The method of multiple working hypotheses. Journal of Geology 5:837-848 (reprinted in Science 148: 754-759 in 1965).

Platt, J. R. 1964. Strong inference. Science 146:347-353.

Popper, K. R. 1963. Conjectures and Refutations: The Growth of Scientific Knowledge. Routledge and Kegan Paul, London.

Leave a Reply