Although a colligatory concept is not a copy-like re-presentation of historical reality, it could still be taken to be justified if one could reduce its content to historical evidence. However, it is immediately clear that this brings one to a dead end since colligatory concepts go beyond empirical data. And this is their most important feature. Colligation is a uniting of seemingly separate data into a whole, and the principle that unites is not inherent in the data. The concept of the 'Thaw' subsumes heterogeneous material under it. and the organizing principle - the atmosphere of relaxation and 'thawing' in comparison to Stalin's freeze - cannot be said to be found in historians' empirical evidence in any unmediated manner.
Imagine a historian collecting evidence from the period of 'Khrushchev's Thaw': the historian might be reading archive material on the numbers of Gulag prisoners released and notice that many were freed - certainly more than when Stalin was in power. He or she might also note that books and magazines that would most likely have been censored earlier were allowed to appear. The historian might find some notes concerning discussions in the Politburo regarding the limits of what can be published and remark that Khrushchev favored a more liberal line. Then he or she might browse newspaper stories about the Soviet leaders' trips to previously hostile countries like Yugoslavia and the USA as well as about return visits by cultural delegates from these countries. And so on.
There is no prospect that the historian would ever come across the concept or idea of the 'Thaw' in this very diverse material1 and certainly no possibility that she would stumble upon a unifying link between phenomena in the source material itself. The dream of such direct empirical inference resembles Baconian naive inductivism. We can safely conclude that the 'Thaw' does not emerge 'from below' in any direct or unmediated manner. Detecting a specific 'pattern' in historical data and colligating them under the 'Thaw' requires some extra-evidential factors. Indeed, the idea of the 'Thaw' seems to imply another metaphor, the 'freeze' of Stalin. On the basis of that, exactly as the idea of colligation implies, the historian may construe the unifying principle of 'thawing' and impose it 'from above' on historical material.
What if one considers the question of empirical justification from the other side? In other words, assuming we already possess the concept of the 'Thaw', is it reasonable to think that historical data vindicates it as a uniquely applicable colligatory concept? This also seems unlikely. First, there are strong general grounds to doubt that any theoretical construct or abstraction can be uniquely determined on the basis of empirical evidence alone. The thesis of the underdetermination of theory by data is one of the key lessons from the twentieth- century philosophy of science (Quine 1953). It is worth paying attention particularly to what has become known as 'contrastive underdetermination' rather than to 'confirmation holism' or 'holist underdetermination'.2 The former says that a body of evidence that confirms or justifies a theory can equally well confirm a number of alternative theories. In other words, one cannot be expected to choose and find the correct theory from among a series of alternatives on the basis of empirical evidence alone. Contrastive underdetermination is sometimes illuminated by the following abstract example. Imagine a data set of points drawn on a surface. How many possible lines could we draw between them? The answer is that it is possible to connect the points in infinitely many ways. If we add more data points, we eliminate many curves, but infinite possibilities always remain.
Contrastive underdetermination means that there is always a logical possibility of alternative theories: although we may not be aware of an alternative theory, it is always logically possible to construe or imagine an empirically equivalent theory. In other words, even if we did not have an alternative colligatory concept to 'Thaw' for the period after Stalin's reign, this does not mean that no other relevant colligatory concept is conceivable and equally warranted in light of the historical evidence. One might object that the 'data' in history research are rather different to the 'points on paper' of the analogy, however. Almost any historical 'data' contain a wealth of information and meanings, and are thus not as radically open as 'data' in the physical sciences. But I think this conclusion would be a red herring.
If one thinks of an archival document concerning the release of prisoners, the books and magazines published in the 1960s, the discussion of the members of Politburo or even actual newspaper stories from the 1960s, one is reminded that they all are loaded with information and suggest practically unlimited lines of interpretation. Indeed, the abundance of information only emphasizes the problem of underdetermination since unlimited lines of interpretation remain open even if many of them seem implausible. One might say that, with respect to any one interpretation or colligation, historical evidence is radically underdetermined: due to the nature of historical evidence and the historical imagination, it is always possible to construct an infinite number of alternative interpretations.3 The openness of interpretation can be illustrated with the example of a list of prisoners in the Gulag, a very simple document in comparison to many others in historiography. What can be concluded on its basis? Or more generally, in what kinds of reasoning can it be used as evidence? Arguably in an endless array of inferences: the number of prisoners, the sex of the prisoners, the nationalities of the prisoners, the nutrition of prison camps, the quality of ink used by prison guards, the management of prison camps, the system of archiving, the politics of the Soviet regime, the 'Thaw', etc. There is really no end to interpretative lines. And this further supports the central idea of and need for colligation: in order to make sense of historical evidence and view the past as containing meaningful patterns, a historian needs to highlight some aspects and link historical phenomena into colligated wholes.
In contrast to the Quine-Duhem thesis on underdetermination, Tucker (2004) argues that it is possible to use evidence to determine theories, save perhaps in some stubborn historiographical cases. Further, Tucker suggests that we distinguish between determinists ('historians infer from evidence with historiographic theories and methods a single historiographical "output"'), indeterminists ('whatever consistency and regularity we find in historiographic judgment result from political, ideological, or socio-historical factors that influence groups of historians') and underdeterminism ('historians are constrained by the evidence and their theories to choose among a finite range of possible historiographies') (Tucker 2004, 9). This requires some comment. First, Tucker does not address 'determinism' per se. Neither does he talk about 'evidential determinism' - that is, that evidence alone determines historiographical interpretations, which one might expect in a discussion of the underdetermination thesis. He is interested in what might be called 'cognitive determinism', as it is both evidence and theoretical matters in the fashion of good old internalism that determine the 'output'. I think that evidential determinism does not hold and I am also agnostic with respect to cognitive determinism, but do think that it would valuable if one managed to develop a theory of cognitive determinism. Second, one does not have to be exclusive with respect to cognitive factors and extracognitive factors. It may be the case that they both have a role to play in scientific (and historiographic) decision making, and this state of affairs would be a disaster only for the 'Old Rationalist' (See Bird 2000, 3-9), that is, the one who believes that strong algorithmic kinds of rationality principles alone determine the content of science. If both types of factors have a role to play, it is not possible to develop an algorithm to show how theory choices are determined, but we would have a set of rational and other criteria at our disposal nevertheless. This means that 'cognitive indeterminists' do not have to think that cognitive factors, such as evidence, have no role whatsoever in theory decisions. They only need to think that solely cognitive factors do not determine the output. Furthermore, Tucker's indeterminists are not indeterminists if they think that political, ideological and socio-historical factors determine output. This is a common misunderstanding among philosophers of such schools in science studies as the Sociology of Scientific Knowledge (e.g. Bloor 1991), which leans strongly towards determinism and talk of causality but denies that evidential and cognitive factors decide outcomes (they also leave room for both social and observational input). Third, a more appropriate understanding of underdetermination would be to say that historians are able to choose among (logically) infinite, not finite, possible historiographies, although this stretches the meaning of 'choice', as all historiographies are not available in the same already-formulated manner.
Tucker also claims that there is a discrepancy between the actual history of science and the kind of history that the Quine-Duhem thesis would lead us to expect because scientists have agreed to prefer some theories over others. On this basis, he concludes that 'the evidence ... from the historiography of science is sufficient to disprove Duhem-Quine's underdetermination thesis' (2004, 145). This is a hasty conclusion, however, as the core of the underdetermination thesis is that empirical equivalence between various theories with respect to shared evidence is compatible with there being other rational or non-rational reasons for preferring some theories over others. It is strange that Tucker draws his conclusion because he also recognizes both empirical equivalence as the essence of the Duhem-Quine thesis and the possibility of extra-empirical criteria in theory-choice. Tucker admits that some 'parts of historiography are underdetermined' because some disputes have remained unresolved for a long time (2004, 151). But it is similarly odd that he both bases this conclusion on empirical evidence and suggests determining the extent of underdetermination in historiography empirically. The underdetermination thesis is a logical claim regarding the relation between theories and evidence and empirical investigations can, at best, make it more or less plausible. Naturally, if one theory appears to be more strongly supported than its rivals and manages to widen its evidential scope more than any other, for example, this provides a pragmatic reason to choose it in that situation. This state of affairs does not show the underdetermination thesis to be right or wrong, however. To show that it is wrong, one would need to prove that there is logically only one possible theoretical construction from a given set of evidence. But so far philosophers have not managed to find any such proof. In other words, if Quine and Duhem are correct, even if scientists had only one theory, this does not show the thesis to be wrong, since there may be yet unimagined empirically equivalent and more justified theories.
Consensus may simply reflect a lack of imagination in a community or perhaps the existence of repression (which Tucker has well analyzed).