Meetings 5-6: Strengths and Limitations of Hypothetico-deductive Science
Each of the following papers discuss scientific methodologies that might be different from classical
hypothetico-deductive science (HDS) as articulated by, for example, Popper, Giere, and
Underwood. Consider which arguments are refinements of HDS vs. being antithetical to HDS. For those
that are refinements, are they essential or not? For those that are antithetical, do they undermine HDS
or are they wrong? Can you identify other weaknesses of HDS besides those represented
here? Imagine that you were given an hour with incoming graduate students to summarize a process
for conducting effective science in your discipline. How would you do it?
Platt, J.R. 1964. Strong Inference. Science 146:347:353. In preparation for discussion, identify an
example of research in your discipline that has employed multiple working hypotheses (see some
examples from Matt’s research at end of this document).
Krebs, Charley. 2015. A Survey of Strong Inference in Ecology Papers: Platt’s Test and Medawar’s Fraud
Model. Posted 15 March 2015 on Ecological Rants, Ecological opinions of Charley Krebs and Judy Myers.
Is hypothetico-deductive science the exception rather than the rule in ecological research?
Loehle, C. 1987. Hypothesis testing in ecology: psychological aspects and the importance of theory
maturation. Quart. Rev. Biol. 62:397:409. Be prepared to evaluate the thesis that confirmation bias
and theory tenacity influence the scientific process. Identify at least one example from your discipline
of:
(1) failure to publish negative results –
(2) misuse of mathematical models –
(3) confusion resulting from ambiguous terms (4) biases against new ideas -
Lloyd, E. A. 1994. "Confirmation of Evolutionary Models.” Chapter 8 in The structure and confirmation
of evolutionary models. Princeton University Press, NJ. Identify examples of theories from your
discipline that have gained support as a result of:
(1) fit between model and data –
(2) independent support –
(3) variety of evidence.-
Anderson, D.R., K.P. Burnham, and W. L. Thompson. 2000. Null hypothesis testing: problems,
prevalence, and an alternative. Journal of Wildlife Management 64:912:923.
Trafimow, David & Michael Marks (2015) Basic and Applied Social Psychology bans use in its publications
of the "null hypothesis significance testing procedure". Editorial, Basic and Applied Social
Psychology, 37:1, 1-2. Related editorial in Nature by Regina Nuzzo: "P values, the gold standard of
statistical validity, are not as reliable as many scientists assume."
Evaluate the thesis that “tests of statistical null hypotheses have relatively little utility in science and are
not a fundamental aspect of the scientific method.” Why is the P value under attack now after reigning
for decades? Is this a trajectory or just a blip?
Forber, Patrick. 2011. Reconceiving eliminative inference. Philosophy of Science 78(2): 185-208.
Is Forber in agreement with Popper that we tend to learn the most when we eliminate possibilities?
Recreational reading
Burnham, K. P., D. R. Anderson, and K. P. Huyvaert. 2011. AIC model selection and multimodel inference
in behavioral ecology: some background, observations, and comparisons. Behavioral Ecology and
Sociobiology 65:23-35.
Elliott, L. P. and Brook, B. W. 2007. Revisiting Chamberlin: multiple working hypotheses for the 21st
century. Bioscience 57: 608-614. Interprets Platt 1964 with respect to information-theoretic
approaches and Bayesian statistics.
Chamberlin, T.C. 1890. The method of multiple working hypotheses. Science (7 February 1890).
Michael R. Dietrich and Robert A. Skipper, Jr. 2013. R. A. Fisher and the Foundations of Statistical
Biology. Pages 147-160 in Outsider Scientists: Routes to Innovation in Biology. Oren Harman and
Michael R. Dietrich, Editors. University of Chicago Press.
Stephens, P. A., S. W. Buskirk, and C. M. del Rio. 2007. Inference in ecology and evolution. Trends in
Ecology & Evolution 22:192-197.
Gibbons, J. M., N. M. J. Crout, and J. R. Healey. 2007. What role should null-hypothesis significance tests
have in statistical education and hypothesis falsification? Trends in Ecology & Evolution 22:445446.
Rio, C. M. d., S. W. Buskirk, and P. A. Stephens. 2007. Response to Gibbons et al.: Null-hypothesis
significance tests in education and inference. Trends in Ecology & Evolution 22:446.
Link to more references on frequentist statistics, information theory, and Bayesian statistics.
Three examples of multiple working hypotheses
Matt Ayres
All have the structure of being a question with a number of different possible answers.
Research question #1: Why are stands of longleaf pine less susceptible to southern pine beetle than
stands of loblolly pine?
H1. Null hypothesis. Within forests, D. frontalis infestations occur within stands of longleaf pine and
loblolly pine in proportion to their abundance. The conventional wisdom that there are fewer
infestations with longleaf is a false impression created by the low abundance of longleaf pine in
the contemporary landscape.
H2. Compared to loblolly pines, the resin defenses of longleaf afford better protection against bark
beetles.
H3. Greater inter-tree spacing in longleaf stands compared to loblolly limits the ability of frontalis to
aggregate during attacks and dampens population growth as a result.
H4. The searching behavior of dispersing adults of D. frontalis is biased against stands of longleaf pine
compared to stands of loblolly pine.
H5. When struck by lightning, longleaf pines are less suitable than loblolly pines for colonization by D.
frontalis and its fungal associates.
Research question #2: What produces positive density-dependence (Allee effect) in the population
dynamics of D. frontalis? i.e., why do large infestations grow more than small infestations?
H1. The apparent phenomenon is actually just exponential growth and not truly an Allee effect.
H2. When local populations are larger, tree-specific attack rates are greater, resin defenses of trees are
more quickly depleted, and per capita reproductive success is greater.
H3. Small infestations are more likely to collapse due to an interruption in attacks (from demographic
stochasticity and variance in emergence times).
H4. In relatively small infestations, there is increased mortality of attacking adults because it is difficult
for beetles to locate trees with a favorable (low to moderate) density of adults already present.
Research question #3: Why are there large fluctuations in the abundance of some forest insects?
H1: The populations cycle due to delayed density-dependent feedback from specialist predators.
H2: The populations cycle due to delayed density-dependent feedback from inducible plant defenses
H3: The populations fluctuate due to density-independent effects from climatic variation
H4: The populations fluctuate due to density-independent variation in host quality, natural enemies,
mutualists, or competitors.