The advocates of the empiricist consensus assumed either the
H-D account of justification, or possibly the Popperian view of
falisificationism. Both of these accounts of why we are justified
in believing general statements ("laws" or "theories") hold that each
proposed general hypothesis has to meet the "test" of observational evidence,
or in other words what we "test" by the scientific method are hypotheses
one by one. The philosophical thesis known as "holism" presents
a serious challenge to this apparently innocent assumption.
"Holism," is the view that what is in fact "tested"
by observational evidence (if anything is) are not individual laws
or theories, but rather large constellations of belief which include not
only the one "theory" that is allegedly undergoing empirical "testing"
but also a whole array of "auxiliary hypotheses" which are more or less
tacitly taken for granted.
This point was first made by the non-consensus French philosopher/historian
of science, Pierre Duhem. Later the American pragmatist Willard
van Orman Quine made effectively the same point, but extended the set
of "auxiliary hypotheses" to include not only those statements which the
positivists had considered "synthetic" but also even those once regarded
as "analytic," i.e. true by definition. Quine's point was that faced
with "recalcitrant evidence" (observations which do not square with our
expectations), there is always the option of altering meanings,
or even logical laws, to save a cherished hypothesis from potential refutation.
Thus in famous phrases Quine argues that our whole system of scientific
claims, our "web of belief," faces the "tribunal of experience"
only as a "collective body" not as individual hypotheses.
Thus the "Quine-Duhem thesis" has come to be the name
for the (presumably true) view that what is actually required for a test
is not merely one hypothesis, but the conjunction of that hypothesis
with a wide variety of often tacit auxiliary hypotheses
The occurrence of an observation which refutes a prediction derived from
the conjunction of the theory under test with all the necessary
auxiliaries only shows that at least some one member
of that conjunction is false, not that the false assumption is necessarily
the theory under test. Whether the false prediction resulted from
the falsity of the tested hypothesis or the falsity of some other one or
more of the auxiliary assumptions is not known. Thus, according
to the Quine-Duhem thesis, by suitable modification or buttressing of the
proper auxiliary hypotheses, any theory can always be "saved" from potential
refutation. Furthermore, it is often claimed that historical research shows
that scientists do frequently do precisely this. Hence it would seem that,
contra Popper, theories cannot be definitively refuted any more than
they can be confirmed.