Review - Theories of Scientific Method
An Introduction
by Robert Nola and Howard Sankey
McGill-Queen's University Press, 2007
Review by Keith Harris, Ph.D.
Sep 16th 2008 (Volume 12, Issue 38)

Most non-scientists have a common, if unspoken, understanding of what science is, how it works, what its aims are, and what it is used for.  Science is, more or less, the process by which we discover new ideas and knowledge, especially those that lead to new (or better) technologies and everyday conveniences than we already have.  Forward progress in medicine, genetics, space exploration and so on seem to happen as a matter of course, and the general public rarely needs to think much beyond consideration of the immediate outputs or consequences of the scientific enterprise.

The thought of scientific methodology usually brings up images of brainy-looking people in lab coats tinkering with beakers or electronics, or scruffily-dressed professors scribbling on chalk boards.  Scientific theories are seen as unproven assumptions or even just opinions: some sensible and well-founded in fact, others strange or even bizarre.  Discussions of the theory (actually theories) of scientific methodology itself are not likely to gain much interest around the dinner table.

Nola and Sankey, however, make it clear in this hefty book why scientists, at least, should put some thought into the issue.  First, the way scientists approach their work is grounded in a specific methodology, or at least assumes a methodology; and the assumptions inherent in that methodology affects how they develop their research agendas and what types of testing and validation are deemed acceptable.  At a deeper level, the scientific enterprise itself moves forward according to fundamental assumptions about what's knowable and how that is to be known.

Very appropriately, In Part 1 the authors take up the meaning of methodology as a concept, and whether there even is a set of methodologies that can be seen as common to the scientific enterprise as a whole.  They divide up their initial discussion into four categories:  (1) methods of practice, (2) methods of discovery and justification, (3) methods of heuristics, and (4) methods in relation to logic.

Of special interest in this section is the authors' consideration of Karl Popper's apparent denial that there is actually any definable method for the scientific enterprise.  According to Popper's well known and highly regarded work, The Logic of Scientific Discovery (1959), there actually is no logic of scientific discovery, thus no definite or circumscribed set of methods by which to engage in science.  Nola and Sankey neatly refute, or at least delimit, this position, and subsequently put to rest other challenges to the existence of methodologies as well.

The importance of understanding methodology is clear.  By what rules should scientists select areas worthy of research or study?  How does one establish standards of validation for analysis of findings?  By what standards should competing theories be compared?  Can logic itself provide enough support upon which to build a scientific methodology?

The next sections of the book are much thicker going.  In Part 2, the authors discuss in detail the inductive method, abductive reasoning and inference to the best explanation (IBE), and the hypothetico-deductivist approach.   These systems, especially induction and deduction, are probably the most familiar to non-academics, although in many areas abductive reasoning is probably the most reasonable approach.

Part 3 explains Bayesian methodologies, which are those based in probabilistic inference.  In this section, Chapter 8 explains the development and refinement of approaches based on Bayes initial discoveries in the 18^th century.  In the next chapter, the authors take up various concerns about the logic (so to speak) of Bayes' Theorem, including the "problem of priors."  The book provides so many interesting entries that it would be impossible to list them, but since this one is well known it will serve as an example.

The problem is thus:  In order for Bayes' Theorem to be applied with precision, knowledge of "the priors" is critical.  That is, applying all the known or knowable factors to the estimation of a probability is essential to the method.  As an example, if we know that at any given time, 12% of drivers in a certain city are drunk while driving, and we know that a breathalyzer test used to detect the presence of alcohol is 96% accurate, we can use Bayes Theorem to estimate the probability that any single case apprehended as a DUI is correct if based on the test.  But if we also know that among those who drive while drunk, younger drivers have a higher incidence than older drivers, we should include that factor in our estimation of the single case.  But we may not know, at that moment, what the differential due to age actually is.  This illustrates the problem of priors.  Some theorists argue that this issue will resolve with time as more data are collected, but how does that help us with this particular case?  The methodological challenge remains of how much freedom researchers should exercise when assigning degrees of belief at any particular point in time.

Part 4 of the book carefully dissects the relevant ideas of Karl Popper, Imre Lakatos, Thomas Kuhn and Paul Feyerabend, whose approaches are collectively (and perhaps humorously) concatenated into the designation "Popkuhnlakabend" by philosopher David Stove.  It is in this section that Popper's ideas and concerns are fully explained and considered.  And finally, Part 5 looks at the naturalistic or pragmatic theories of W.V. Quine, Larry Laudan and Nicholas Rescher.  To the novice, this part of the book will read something like a collection of encyclopedia entries, although of course there is a common thread.

Indeed, for the non-philosopher, this book presents a very daunting challenge.  It is replete with carefully detailed analyses and lengthy discussions of nuances such as only philosophers can perform (or truly understand). 

It is important to emphasize that the book's writing is clear and succinct; it is the concepts are challenging.  To the extent that readers can follow the detailed discussion, the book provides a very thorough job of substantiating the practical consequences of the central issues (e.g., whether there is one or more definable scientific methodologies, and if so what might those be).  Even for general science readers the book will provide much food for thought about an aspect of science rarely thought of outside academia.

© 2008 Keith Harris

Keith Harris, Ph.D., is Chief of Research for the Department of Behavioral Health in San Bernardino County, California. His current interests include the empirical basis for mental health research, behavioral genetics, and the shaping of human nature by evolutionary forces.