Strong Inference

Abstract

Scientists these days tend to keep up a polite fiction that all science is equal. Except for the work of the misguided opponent whose arguments we happen to be refuting at the time, we speak as though every scientist's field and methods of study are as good as every other scientist's, and perhaps a little better. This keeps us all cordial when it comes to recommending each other for government grants.But I think anyone who looks at the matter closely will agree that some fields of science are moving forward very much faster than others, perhaps by an order of magnitude, if numbers could be put on such estimates. The discoveries leap from the headlinesand they are real advances in complex and difficult subjects, like molecular biology and high-energy physics. As Alvin Weinberg says (1), "Hardly a month goes by without a stunning success in molecular biology being reported in the Proceedings of the National Academy of Sciences." Why should there be such rapid advances in some fields and not in others? I think the usual explanations that we tend to think of-such as the tractability of the subject, or the quality or education of the men drawn into it, or the size of research contracts-are important but inadequate. I have begun to believe that the primary factor "nature" or the experimental outcome chooses-to go to the right branch or the left; at the next fork, to go left or right; and so on. There are similar branch points in a "conditional computer program," where the next move depends on the result of the last calculation. And there is a "conditional inductive tree" or "logical tree" of this kind written out in detail in many first-year chemistry books, in the table of steps for qualitative analysis of an unknown sample, where the student is led through a real problem of consecutive inference: Add reagent A; if you get a red precipitate, it is subgroup alpha and you filter and add reagent B; if not, you add the other reagent, B'; and so on.On any new problem, of course, inductive inference is not as simple and certain as deduction, because it involves reaching out into the unknown. Steps 1 and 2 require intellectual inventions, which must be cleverly chosen so that hypothesis, experiment, outcome, and exclusion will be related in. a rigorous syllogism; and the question of how to generate such inventions is one which has been extensively discussed elsewhere (2, 3). What the formal schema reminds us to do is to try to make these inventions, to take the next step, to proceed to the next fork, without dawdling or getting tied up in irrelevancies.It is clear why this makes for rapid and powerful progress. For exploring the unknown, there is no faster method; this is the minimum sequence of steps. Any conclusion that is not an exclusion is insecure and must be rechecked. Any delay in recycling to the next set of hypotheses is only a delay. Strong inference, and the logical tree it generates, are to inductive reasoning what the syllogism is to deductive reasoning, in that it offers a regular method for reaching firm indu...

Can we save large carnivores without losing large carnivore science?

“…A good understanding of species ecology can serve to exclude spurious or unreasonable interpretations of correlative data, and such studies can also exclude certain hypotheses or provide indirect support for other hypotheses for which experiments could be designed to provide a definitive test (e.g. Platt, 1964;Kershaw, 1969;Underwood, 1997;Fairweather and Quinn, 2006). While these types of lower-inference studies may not overcome all the aforementioned issues we describe, they do have the advantage of being far cheaper and faster to conduct under a wide range of different ecological conditions, which can address problems associated with the transferability of knowledge between contexts.…”

“…Syntheses and literature reviews of large carnivores' ecological roles should identify these issues, but they usually do not, instead routinely failing to assess the internal validity of the original studies reviewed, as described by Bilotta et al (2014). When the individual empirical studies that form the content of these reviews are judged against Platt's (1964) criteria for strong inference, Hone's (2007) deconstruction of experimental design capabilities, or Sutherland et al' (2013) 20 tips for interpreting scientific claims, it is clear that even literature reviews (e.g. Ritchie and Johnson, 2009;Estes et al, 2011;Ripple et al, 2014b) seldom offer reliable guidance on the state of the literature addressing the MRH, TCH and BMTCH.…”

“…However, manipulative experiments clearly trump anecdotal, observational and/or correlative information for their informative value, and should therefore be valued more highly in the decision making process (Platt, 1964;Fleming et al, 2013). While the weight of evidence for the general role of large carnivores in triggering trophic cascades is indeterminate at this time (but we look forward to this potentially changing one day), we caution researchers and science communicators to carefully consider the implications of simultaneously advocating for both large carnivore conservation and the primacy of top-down trophic cascades.…”

On strong inferences and irreproducibility in reproductive medicine

“…Such was the theme of Pratt's premonition earmarked for future generations, a cautionary tale that holds as much relevance to the discipline of reproductive medicine today as it does for many other research disciplines, medical or otherwise [1]. Consider the opening paragraph of his message, again being mindful of a frontier yet to have been discovered:…”

The Interpretation of Regression Coefficients in a School Effects Model1

“…result" of this type of concrete consideration may be a more "reasonably specific formulation of what we do not know in a region of inquiry, of why we want to know it, and, in the favorable case, of how we might proceed to find out" (Merton, 1968). Our commitment is to the "strong inference" approach enunciated by Platt (1964). -5-influences will in general lead to "specification error," Le., bias in calculated parameters due to incorrect specification of the structural model.…”