Académique Documents
Professionnel Documents
Culture Documents
BenchMarks
Two terms used as frameworks for scientific experimentation—the “hypothesis” and the
“model”—carry distinct philosophical assumptions, with important consequences for the
practicing scientist.
Scientists are commonly taught to frame (Anderson, 2008). Given the ability of tion for the hypothesis characterized
their experiments with a “hypothesis”— scientists to gather vast amounts of sci- the 1500s as a “century of confusion”
an idea or postulate that must be entific data—by sequencing genomes, (Hall, 1962), in which there was clear
phrased as a statement of fact, so that surveying changes in the expression excitement over new developments
it can be subjected to falsification. The of every gene, or analyzing proteomic but no programmatic march forward
hypothesis is constructed in advance of changes in response to a stimulus—is a due to a lack of an accepted method
the experiment; it is therefore unproven hypothesis the most appropriate way to to distinguish between various claims
in its original form. The very idea of frame such experiments (Glass, 2006)? of discovery. Galileo represented a
“proof” of a hypothesis is problematic Here, we discuss the philosophical rea- move to a hypothesis more grounded
on philosophical grounds because the soning that motivated original and cur- in realism, and an increased emphasis
hypothesis is established to be falsified, rent notions of the hypothesis and its on the experiment as the basis for con-
not verified. The second framework for implications for scientific experimental clusions, but he did not systematize
experimental design involves building a design. his approach into a methodology that
model as an explanation for a data set. could be clearly followed by others.
A model is distinct from a hypothesis The Novum Organum It was Francis Bacon who in 1620
in that it is constructed after data are Galileo helped to initiate the renais- wrote an approach to scientific meth-
derived. In contrast to the hypothesis, sance in science that occurred in odology—his Novum Organum or “new
the model must be held up for verifica- Europe in the 16th and 17th centuries, instrument”—new because Bacon took
tion—its success is determined by its by upending religious dogma as to the issue with Aristotle’s Organon (the term
ability to predict a particular outcome. centrality of Earth’s position in the uni- given by Aristotle’s followers to his sys-
Furthermore, an unsuccessful, or not verse, through mathematical reasoning tem of logic) in several respects (Bacon,
fully successful, model need not be and observation. He was described by 1620). Bacon noted that deductive rea-
scrapped in the way that the alterna- a Paduan contemporary as “the father soning by itself is not sufficient because
tive framework urges the rejection of a of experiments and all their exact- if the premise is set in advance of the
falsified hypothesis, but it may instead ness.” Galileo’s approach was in line experiment, for example by a hypoth-
serve as the starting point for a suit- with Greek antecedents, especially esis, the reasoning would be twisted
ably refined successor. The concept Aristotle and Archimedes, in his reli- to meet that premise. Bacon therefore
of a model’s “verification” requires an ance on “deductive” reasoning stem- argued that a purely experimentation-
acceptance of “inductive reasoning”—a ming from hypotheses. The hypoth- based methodology was necessary, and
form of logic that allows the scientist to esis as it was used in the 1500s was that to solve problems with pre-existing
both generalize a particular result and a premise—a starting point based on bias, “the only hope is true induction.”
say that the same result will occur in the unproven assumptions. From the initial With “inductive reasoning” a data set is
future—which itself has been criticized. premise, deductions would be made, taken and used to infer that under similar
Although many scientists use the term and their success or failure was deter- circumstances the result will be repeated,
“hypothesis” when they mean “model,” mined by subjective assessments as to and that the finding as applied to a spe-
we will maintain the distinction that whether they were satisfactory in their cific case may be generalized to other
the hypothesis is an unproven premise explanations of the premise. Although cases of like kind. In the case of grav-
whereas the model is data derived, to this method resulted in “satisfactory” ity, one observes first that a mass falls
discriminate between “top-down prem- conclusions as applied by Galileo, the toward earth at a particular rate, and later
ise/deduction” and “bottom-up data/ same cannot be said of some other that it does so in a predictable and “veri-
induction.” A recent article proposes 16th century physicists, who applied fiable” manner, meaning that after hav-
that the availability of large amounts fictitious and nonrealistic theories to ing established the rule for how quickly
of scientific data renders the need for physics as well as to astronomy (Blake objects descend, one can predict that
a pre-existing hypothesis obsolete et al., 1966). The lack of a founda- descending bodies will continue to follow
strating the reproducibility of the data, ect where comprehensive data sets Hall, N., and Hájek, A. (2001). Induction and Prob-
one can join in the inductive project are accumulated, such as genome ability. In Blackwell’s Companion to Philosophy
of Science, P. Machamer and M. Silberstein, eds.
with a clinical trial outcome. sequences, a hypothesis may not even (Blackwell), pp. 149–172.
The relevance of a philosophical pos- be feasible. What would such a hypoth-
ture for a physician or scientist is made esis be? Only a question is required: Hume, D. (1749). Text used: An Enquiry Concern-
ing Human Understanding, 1999 (Oxford: Oxford
clearer by considering the patient who “What is the sequence of genome X?”. University Press).
seeks treatment for his disease. The This would be followed by an answer,
Kuhn, T.S. (1977). The Essential Tension; Selected
patient who receives a new treatment the genome sequence, that can be Studies in Scientific Tradition and Change (Chi-
based on the results of an experimen- tested for reproducibility by further cago, IL: University of Chicago Press).
tal trial, and as a result survives (as sequencing, allowing for an increasingly
Newton, I. (1721). Text used: Opticks, or A Treatise of
revealed by the experience of those improved model for the genome. Thus, the Reflections, Refractions, Inflections, and Colours
patients who were not similarly treated), although a hypothesis might have been of Light, 1979 (New York: Dover Publications).
must continue to be offered the ther- thought to be necessary in the past, it no
Newton, I. (1729). Text used: The Principia, A New
apy and would not be well served by a longer seems to be so. It is better to see Translation by I. Bernard Cohen and Anne Whit-
philosophical program that claims this science as a quest for good questions man, Preceded by A Guide to Newton’s Principia
by I. Bernard Cohen, 1999 (Berkeley, CA: Univer-
past experience is irrelevant. Therefore, to try to answer, rather than a quest for
sity of California Press).
the clinical trial should not be framed bold hypotheses to try to refute.
with a hypothesis aimed at falsification Nozick, R. (2001). Invariances, the Structure of the
Acknowledgments Objective World (Cambridge, MA: Belknap Press).
but rather should be explicitly inductive
because the project of the physician is Poincaré, H. (1952). Science and Hypotheses
D.J.G. thanks M. Fishman, B. Richardson, the No- (New York: Dover Publications).
explicitly inductive. vartis community, S. Nelson, G. Smith, and L. Zim-
When one turns to basic biology, merman and also R. King and J. Brugge for help- Popper, K. (1959). Text used: The Logic of Scien-
the issue is straightforward. Here, the ing D.J.G. to establish the Experimental Design tific Discovery, 2002 (London: Routledge).
course at Harvard Medical School. N.H. thanks A.
scientist must ask whether verification Hájek for stimulating discussions about induction Russell, B. (1912). Text used: The Problems of Phi-
or falsification is being sought. If the and scientific methodology. losophy, 2007 (New York: Cosimo Classics).