Verification: The use of empirical
data, observation, test, or experiment to confirm the truth or rational
justification of a hypothesis.
Scientific beliefs must be evaluated and supported by empirical data. What does this require? Two concepts are fundamental in
discussing scientific method: truth and justification (warrant). A hypothesis is true if it corresponds
to the way the world is.
Justification has to do with the grounds we have for believing a given
statement to be true. A hypothesis
is rationally warranted if a body of evidence and inference has been provided
in support of it. Ideally, the
fact that a statement is rationally warranted ought to make it likely that the
statement is true. (This treatment
makes apparent the relevance of Baysian theory to scientific inference. The questions have to do with the
transmission of rational credibility from one body of beliefs to another.) Philosophers have introduced several
concepts and theories on the basis of which to analyze the logical and inferential
relations between empirical data and scientific hypotheses, including
observation, falsifiability, confirmation, and experimental method. One of the most important consequences
of this extended and complex debate is the conclusion that theories cannot be
“verified”, but they can be “confirmed,” “warranted,” or “falsified.”
The
general question of scientific inference can be formulated in these terms:
Given a body of evidence E and a hypothesis or theory T, how do we measure the
warrant of T given E? Are there
logical grounds for answering this question? (This does not define the full task for a theory of
scientific method, because scientific method should also give us some guidance
concerning the problem of discovering a relevant body of evidence.) It is rare for a scientific hypothesis
to be amenable to direct and certain confirmation along these lines: given E, H
is certainly true. That is, it is
rare that there is a finite body of observations that suffice to establish the
truth of a given scientific hypothesis.
This is so for two reasons: First, because scientific hypotheses
normally refer to entities, mechanisms, or processes that are not directly
observable; and second, because hypotheses and theories normally make universal
claims (laws) that go beyond any finite body of observations. Instead, verification normally takes
the form of indirect inductive or hypothetico-deductive support for the
hypothesis: given E, H is likely to be true.
Suppose
that H deductively implies O and that the body of evidence E contains “not
O”. In this case E falsifies
H. On these assumptions, the body
of evidence contains observations that demonstrate that H is false. (For example, suppose the hypothesis H
deductively entails that the metal will melt when heated to 500 degrees
Centigrade. We perform the
experiment; the metal does not melt; and we conclude that H is false.) Suppose now that H implies a series of
observations O1, O2, ..., On and that E
contains {OI}. (That
is, some of the observational consequences of the theory are found to be
true.) Does E confirm H, and to
what extent? This is the
fundamental question of inductive logic and scientific method. What constitutes a significant test and
confirmation to the theory?
Several logical points are important. No finite list of observations exhaustively confirms a
theory with universal generalizations; and different types of additional
evidence have very different incremental effects on the credibility of the
theory. E.g. additional observations of the same type have less epistemic
weight than an additional unexpected observation.
What
is an observation? This has been a
main source of controversy in the philosophy of science, in that it has long
been recognized that there is no sharp and permanent distinction between
observation and theory. Virtually
all scientific observations are theory-laden. But the essential idea is that an empirical observation is a
scientific belief with a relatively direct relationship between the evidence of
the senses and the truth conditions of the statement, based on reliable
techniques of data collection to which we can attach high rational
credibility. Here, for example, we
are to consider direct sensory observation, observation using instrumentation,
interviews, records of price data, etc.
Most are not “direct” or “certain”. But they are relatively unburdened by currently
controversial theories. This
description suggests an approach to empirical confirmation that can be
described as the “bootstrapping” method, where the credibility of some beliefs
is enhanced by assigning provisional credibility to others; we then return to
re-assess the provisional beliefs based on the wider set of theoretical
beliefs.
Several
general approaches to empirical evaluation of scientific hypotheses have been
offered in the past century. First
is the idea, most deeply explored by Carl Hempel, that we should draw out the
deductive consequences of a theory; evaluate the truth of some of those
consequences using observation and instrumentation; and assign a degree of
warrant to the theory based on the volume of its confirmed observational
consequences. This approach
constitutes the hypothetico-deductive theory of confirmation, and it represents
the logical basis for the experimental method. Second is the idea, advanced by Karl Popper, that the
scientific method works primarily through diligent and serious efforts to
refute scientific hypotheses. The
researcher needs to identify the most unlikely predictive consequences of the
hypothesis and make every effort to show that these predictions are not
upheld—thereby “falsifying” the theory in question. Only when a theory or hypothesis has
withstood serious and varied tests along these lines do we have a rational
basis for believing the theory.
The two approaches are logically similar, in that they attempt to assign
a degree of empirical warrant to a hypothesis based on the truth or falsity of
its deductive consequences. But
the underlying assumptions about how confirmation and testing proceed are quite
different. Confirmation theory
largely assumes that degree of warrant increases through the gradual
accumulation of true consequences; whereas falsifiability theory assumes that
warrant increases only as a result of our failure to refute the hypothesis in
question. Both approaches have
generated a large volume of critical discussion. Critical reflection upon classical confirmation theory notes
that it is difficult or impossible to provide a purely formal specification of
the set of deductive consequences of a theory that serve to enhance the warrant
of the theory. Further discussions
of the theory of falsifiability have noted that anomalies are too easy to find
in the development of scientific theories, and have attempted to offer a more historically
adequate account of scientific belief based on a theory of “scientific research
programmes” (Lakatos 1978).
Brown, Harold I. 1987. Observation
and Objectivity. New York: Oxford University Press.
Glymour, Clark N. 1980. Theory
and Evidence. Princeton, N.J.: Princeton University Press.
Hempel, Carl. 1965.
Confirmation, Induction, and Rational Belief. In Aspects of Scientific
Explanation, edited by C. Hempel.
Hempel, Carl. 1965. Aspects
of Scientific Explanation, and other essays in the philosophy of science.
New York: Free Press.
Lakatos, Imre. 1978. The
Methodology of Scientific Research Programmes: Philosophical Papers.
Vol. 1. Cambridge: Cambridge University Press.
Laudan, Larry. 1977. Progress
and Its Problems : Toward a Theory of Scientific Growth.
Berkeley: University of California Press.
Popper, Karl Raimund.
1965. Conjectures and Refutations; The Growth of Scientific Knowledge.
2d. ed. New York,: Basic Books.
Encyclopedia of Social Science Research
Methods,
edited
by Michael Lewis-Beck (University of Iowa), Alan Bryman (Loughborough
University), and Tim Futing Liao.
Sage Publications.