Introduction: Underdetermination and Empirical Equivalence
It is a truism that empirical data, as actually collected in scientific experiments, do not deductively fix scientific theories. For one thing, scientific theories are universal with respect to space and time, whereas empirical results are finite in number and associated with 0.
the locations and instants at which they were gathered. For another thing, measurement results and outcomes of experiments belong to the domain of observable phenomena and do not immediately inform us about the “theoretical parts” of theories. Both considerations illustrate that theories contain more than what is logically implied by empirical data. The amount of information contained in empirical data is consequently insufficient to make deductions leading to one unique theoretical scheme. The difficulty is comparable to the mathematical problem of solving equations when there are more unknowns than equations: in such cases solutions will generally not be unique, and usually there will even be infinitely many solutions. Similarly, one may expect that there will be many scientific theories that agree with any given amount of experimental results. However, these theories may well lead to very different pictures of the parts of the physical world that are not directly visible in experiments.A distinction should be made at this point. The empirical equivalence that we just discussed consists in agreement with the same finite set of experimental data available at a certain instant of time. This is a limited equivalence, which may be broken by any new incoming piece of evidence. One may also consider the more drastic case of equivalence between different theories with regard to all data that will ever be collected, or with regard to all data that could possibly be obtained (even including those that require experiments that will never be actually done). Even the existence of the latter, very strong kind of empirical equivalence seems a priori plausible, since the equivalence is still only on the “surface of observable phenomena”, which does not exclude that the theories may say very different things about unobservable objects and processes. The underdetermination argument therefore still applies.
It is no wonder that empirical equivalence of theories has often been discussed in the philosophy of science literature, as it appears to pose a problem for evidencebased choice between theories. Writers on the subject usually perceive the danger that the objectivity of science may be called into question and that this in turns threatens the very rationality of science. Accordingly, much effort has been directed at down-playing the significance of underdetermination in scientific practice. In a seminal paper, Laudan and Leplin (1991) (see also Laudan (1990)) have provided what has become the paradigm of pertinent argumentation, concluding that the ordinary practice of science will have no problem of principle with breaking any deadlock of empirical equivalence that might arise. They argue that empirically equivalent rivals will often be artificial and not genuinely scientific; and in the rare cases of contenders that are to be taken seriously, they will generally be supported differently, via relations with other theories or background information. Moreover, if new
evidence comes in it should be expected that hitherto empirically equivalent theories will soon cease to be equivalent.
We shall argue, however, that these Laudan-Leplin style arguments underestimate the significance of empirical equivalence and underdetermination within present-day science. Indeed, differently from what is often assumed in the philosophy of science literature,[110] theoretical underdetermination and empirical equivalence of a non-trivial kind actually occur in modern scientific practice and are frequently discussed in the foundational scientific literature. It will be interesting to compare these cases with the philosophical arguments for the practical insignificance of empirical equivalence, and to see whether these cases pose a threat to scientific rationality.
2