Historical Epistemology or History of Epistemology? The Case of the Relation Between Perception and Judgment
- First Online:
- Cite this article as:
- Sturm, T. Erkenn (2011) 75: 303. doi:10.1007/s10670-011-9338-3
- 285 Views
This essay aims to sharpen debates on the pros and cons of historical epistemology, which is now understood as a novel approach to the study of knowledge, by comparing it with the history of epistemology as traditionally pursued by philosophers. The many versions of both approaches are not always easily discernable. Yet, a reasoned comparison of certain versions can and should be made. In the first section of this article, I argue that the most interesting difference involves neither the subject matter nor goal, but the methods used by the two approaches. In the second section, I ask which of the two approaches or methods is more promising given that both historical epistemologists and historians of epistemology claim to contribute to epistemology simpliciter. Using traditional problems concerning the epistemic role of perception, I argue that the historical epistemologies of Wartofsky and Daston and Galison fail to show that studying practices of perception is philosophically significant. Standard methods from the history of epistemology are more promising, as I show by means of reconstructing arguments in a debate about the relation between perception and judgment in psychological research on the famous moon illusion.
Historical epistemology (HE for short) is an approach to the study of (primarily scientific) knowledge predominantly pursued by historians of science today.1 The history of epistemology (HoE), in contrast, is a way of studying knowledge and theories thereof that derives more from philosophical epistemology. In what follows, I shall compare HE and HoE, identifying their advantages and disadvantages. Such a comparison may be viewed suspiciously: Are there not too many different versions of both HE and HoE, such that one version of HE might differ more from another one than from a particular version of HoE? And is the distinction between HE and HoE not simply contingent on the disciplines to which people belong, rather than on a well-justified distinction between two different forms of inquiry into human knowledge (cf. Garber 2005)? To some extent, such worries are not unwarranted. However, while the differences in the programs and practices of HE and HoE are not always easy to detect, some can be identified. We are dealing here with a spectrum that has a gray section in the middle, but also clear black and white end points. Highlighting the differences between HE and HoE is useful for better understanding both approaches and for evaluating what speaks for and against them.
This paper attempts to achieve this in two main steps. First, I identify versions of HE and HoE that are distinguishable and yet similar enough to allow for an interesting comparison and evaluation. As I argue, the most interesting differences between these versions of HE and HoE primarily involve neither subject matter nor goals, but method. These versions both aim to critically reflect current agendas in epistemology, or even to contribute to current epistemological issues, but differ in terms of tools. The question is, then, which method is more promising (Sects. 2.1–2.3). In the second part, I turn to a thorny epistemological problem that can be found in various forms throughout the history of philosophy and science: How are perception and judgment related and what does their relation mean for the foundation of empirical knowledge (Sect. 3.1)? I then discuss two versions of an HE of perception, namely Wartofsky’s, and Daston and Galison’s. Neither of their contributions enables us to better deal with epistemic problems of perception, nor how to state those problems differently (Sects. 3.2–3.5). Finally, and more constructively, I show how HoE fares better. I shall use an age-old debate: How to explain the famous moon illusion? (Sect. 3.6). In the conclusion (Sect. 4), I highlight limits of my arguments. While HoE is more promising than HE in thinking about the epistemic problems of perception, and while there are probably many other epistemological issues where similar arguments work, counterexamples are possible.
2 Historical Epistemology Versus History of Epistemology
2.1 Historical Epistemology and Its Philosophical Ambitions
The terms ‘HE’ and ‘HoE’ are used in various ways. We may begin by looking at the differences indicated by the adjectival construction ‘historical X’, and the composition of substantives ‘history of X’. Terminological observations admittedly merely provide starting-points. But they may help prevent certain confusions.
In his defense of naturalism in epistemology, Goldman (1986) uses the term ‘historical epistemology’ to designate ideas and works of epistemologists of the past. He limits the term ‘history of epistemology’ to designating works by current philosophers and historians about past epistemologists and their writings. Thus, in Goldman’s usage of ‘history of epistemology’, ‘history’ denotes not the object—the past—but beliefs or claims about the past. That contains nothing illogical. Yet, it is odd that the expression ‘historical epistemology’ should refer most naturally to ideas and works of epistemologists of the past. Rather, ‘historical’ is better used to qualify the term ‘epistemology’ in a certain way, as in ‘naturalistic epistemology’. Such a construction thus denotes a certain approach, or family of approaches, to epistemology. This is consistent with how those who coined the terminology used the expression ‘historical epistemology’ (or the French epistémologie historique), and with the meaning intended by those who use it now (e.g., Lecourt 1969; Wartofsky 1979; Daston 1994; Galison 2008; Renn 1995, 1996; Rheinberger 2010). This is also compatible with the more plausible of Goldman’s claims, namely that ‘history of epistemology’ refers to the discipline that studies the works and ideas of past epistemologists. (Of course, ‘HoE’ is also a suitable expression for that past itself, since ‘history’ is generally ambiguous in this way. In this paper, however, I mostly don’t use ‘HoE’ in this sense, though.)
What qualification, then, is indicated by the adjective ‘historical’ in ‘historical epistemology’? It might specify the goal, the subject matter, or the method of such an epistemology, or some or all of these. For instance, most adherents of what is dubbed ‘naturalistic epistemology’ stick to standard goals of epistemology, but alter the method. They accept that a central aim of epistemology is to develop a theory of epistemic justification—and a theory that does not only factually state what we take to be epistemic justification, but that tells us which epistemic justifications to accept and which to reject. As is well known, naturalists reject the armchair methods of conceptual analysis and thought experiment used by many anti-naturalists in epistemology, arguing that any identification of the right standards for justification cannot be achieved independent of empirical science (e.g. Goldman 1986). Perhaps the adjective ‘historical’ in ‘HE’ functions similarly. History would then help us to accomplish some standard goal of epistemology.
The adjective ‘historical’, however, might have a different purpose. Compare ‘genetic’ in ‘genetic epistemology’ or ‘evolutionary’ in ‘evolutionary epistemology’. Very often, these expressions do not simply stand for particular methods for attaining some standard goal of epistemology. Genetic and evolutionary epistemologies are not theories of what knowledge is, what its sources and limits are, what should count as epistemic justification or how we can improve our epistemic situation. Rather, they are empirical theories about stages and mechanisms in the development of human cognitive abilities. Thus, the methods used by these approaches differ from those widespread in philosophical epistemology and redefine the goals and/or subject matter of epistemology.
The question, then, is whether HE and philosophical epistemology (be it naturalistic or not) differ only in method, or also in their aims and/or their subject matter. Are we still playing the same game or not? This is no easy matter, and it is certainly no longer merely a terminological issue.
How do conceptions of HE look in this light? In this volume we have identified three main versions of HE that study different subject matters (Feest and Sturm 2011): (1) epistemic concepts that guide and evaluate research (e.g. Wartofsky 1979, 1987; Daston 1994; Galison 2008), such as ‘objectivity’ (Daston and Galison 2007) or ‘observation’ (Daston and Lunbeck 2011); (2) the objects of scientific knowledge or, perhaps better, the concepts scientists employ to refer to their research objects (Rheinberger 2010), such as ‘protein’ (Rheinberger 1997) or ‘heredity’ (Müller-Wille and Rheinberger 2007, 2009); and (3) the structures of long-term scientific developments, such as the origins of mechanics or the relativity revolution (e.g. Renn 1995, 1996, 2004; Renn et al. 2004). While these versions of HE differ in subject matter, they exhibit important common methodological features. For instance, all claim that one ought to study the local contexts of science as well as draw comparisons across time and place—combining micro history and macro history. Also, they all aim to study researchpractices that lead to the introduction or change of concepts of objects, of epistemic concepts, and of shifts in the development of scientific theories. These methodological assumptions are rarely if ever made by those who pursue HoE—about which I shall have more to say in Sect. 2.2.
What about the goals of HE? Clearly the above-mentioned general topics of the three versions of HE have counterparts in traditional epistemology and philosophy of science. First, the study of epistemic concepts is described as a Kantian inquiry—by historical means–into the “a priori forms of our perception and cognition” (Wartofsky 1979, p. xxiii), the conditions of the possibility of certain forms of knowledge and evidence (Daston 1994, p. 282f.), or the “organizing concepts” or “categories of thought” such as “knowledge, belief, evidence, good reason, objectivity, probability” (Hacking 1999, p. 58). Second, studies by the French philosopher-historians Bachelard and Canguilhem have exerted a strong influence on historical epistemologies of the objects of scientific knowledge (Rheinberger 2010). At the same time, they also touch on debates concerning realism and anti-realism about theoretical concepts in philosophy of science. Finally, the topic of long-term models of scientific development can be traced to post-Kuhnian debates on the dynamics of research.
Thus, central versions of HE are motivated by, or try to contribute to, standard philosophical topics and issues. Indeed, Hacking (1999, p. 58) says that his interest in the “categories of thought … is not so different from theirs [Aristotle’s or Kant’s] although widely different in manner of execution”; he aims to put historical material to “philosophical purposes” and wants to “make use of the past for the present” (ibid., p. 55f.). But how? And for which current philosophical debates? Hacking does not state that explicitly. Daston, again, justly warning against committing the genetic fallacy, rejects the view that to historicize X is to relativize it, and also notes that “it is not enough to reveal the contingent or accidental character of our current conceptual categories in order to abandon them. We must put something better in their place.” (Daston 1994, p. 284; cf. Daston and Galison 2007, p. 376). More positively, she thinks that history can provide us with “conceptual alternatives that may significantly transform” current debates. For instance, historical analyses of scientific research practices may show that factual statements are theory-laden in various ways, “from the relatively innocuous selection of topics, in the sense of Max Weber, to the sinister extreme of ideology, in the sense of Karl Marx”. Likewise, HE may reveal that the language of the “neutrality” and “contamination” of statements about facts discloses a problematic epistemology, namely the epistemology of the “given” (Daston 1994, p. 284f.). Does it, however, take thorough inquiry into the practices of early modern science to arrive at such results, interesting as they may be? Investing in good old-fashioned armchair epistemology seems sufficient here—and a better bargain. (Later on (Sects. 3.4–3.5), I shall argue against a more elaborate example given by Daston and Galison for how HE might be philosophically useful.)
Let us ignore both Rheinberger’s assumption that epistemology is an inquiry into “what makes knowledge scientific” (a characterization many epistemologists would reject as being too restrictive) and his limiting of epistemology to theories of objects of knowledge. What I wish to focus instead on is the claim that epistemology studies the “process of generating knowledge”. This certainly makes it appear as if his project were purely descriptive/explanatory and not normative. Renn (2008, p. viif.) likewise states that the “main goal” of an HE of mechanics “is to explain the development and diffusion of mechanical knowledge” or “to explain structural transformations of systems of knowledge” (see also Renn et al. 2004).
My use of the term epistemology requires a brief explanation. I do not use it as a synonym for a theory of knowledge (Erkenntnis) that inquires into what it is that makes knowledge (Wissen) scientific, as was characteristic of the classical tradition, especially in English-speaking countries. Rather, the concept is used here, following the French practice, for reflecting the historical conditions under which, and the means with which, things are made into objects of knowledge. It focuses thus on the process of generating knowledge and the ways in which it is initiated and maintained. (Rheinberger 2010, p. 2f.)
But things are not that easy. For starters, it is not clear whether all issues in epistemology are normative—e.g. the task of defining the term ‘knowledge’. Nor are questions such as: Do we possess any knowledge at all? Under what conditions is the knowledge of objects possible? Such issues are at least not normative in the straightforward sense as questions about standards of rationality or about proper methods of belief revision are. Moreover, proponents of HE do not always evade competition with philosophical projects. Thus Rheinberger (2010, p. 1) claims that HE overcomes the well-known distinction between discovery and justification (without explaining how it is done), and insists on the philosophical significance of HE (ibid., 2f. et passim). Renn, again, claims that HE can help “overcome the problems of the traditional philosophy of science to establish universal norms of scientific rationality” (Renn 1996, p. 2), that it can “reverse the trend of the philosophy of science towards content-independent methodology … which began with Neo-Kantianism and continued with analytical philosophy” (Renn 1995, p. 251), and that it may lead toward a new synthesis of history and the philosophy of science. Due both to the inherent complexity of epistemology as well as the ambiguities of the claims of HE proponents, we would need to hear more about the ambitions of HE, and about how these relate to tasks and projects in philosophical epistemology. But we don’t.
We arrive, then, at some preliminary results. On the one hand, if one assumes that proponents of HE do not claim to contribute to genuine philosophical discussion, they fail to clearly demarcate their approach from philosophical epistemology. On the other hand, it looks more as if historical epistemologists do claim to contribute to philosophical discussion, but they fail to make clear how they do so. I shall next consider another historical undertaking, namely HoE, which has also often stated its philosophical ambitions.
2.2 The Ambitions of History of Epistemology: Two Questions and Four Possible Answers
Not surprisingly, there also exists controversy over the subject matter, goals and methods of HoE. Most important for the present discussion, views on whether or not the history of philosophy is necessary or useful for philosophical thinking itself keep changing (Krüger 2005; Mash 1987; Sorell and Rogers 2005). To create some clarity, let me distinguish between two HoE-related questions.
First, should we study HoE systematically, with the intention, for instance, of judging whether a claim made by Plato or Hume or Kant is (fairly) justified? Second, should we assume that the problems of epistemology are perennial, and that trying to reconstruct and evaluate Plato’s or Hume’s or Kant’s arguments is meaningful for current debates, too? These questions are often conflated, but should be kept apart. One can discuss an assumption taken from Plato’s Timaeus, rationally reconstructing2 and evaluating his arguments for it using all the standard tools (e.g., examine the meanings of central terms, distinguish premises and conclusions, analyze the logical form of the argument, consider its formal cogency as well as the truth of its premises, and so on) without this evaluation in and of itself contributing to any current discussion. The same holds for many topics in HoE.
Now, in principle, four responses to the above two questions are possible: Yes&Yes, Yes&No, No&Yes, and No&No. I shall briefly discuss these options.
As to the double denial (No&No), surely HoE may be understood as mere description. One may simply desire to discover what Kant or Hume meant, what caused their beliefs, and so on, without bothering with whether they had (fairly good) reasons, or whether we might learn from them for solving current problems. Such an approach evades anachronism and presentism—at least literally, because we know that even when simply attempting to interpret a past author correctly it is easy to import concepts and views that are our own and not the author’s. One often reads, for example, that Descartes or Kant held this or that view on consciousness which is relevant for our present-day debates; but that may be quite misleading and distort Descartes’ or Kant’s statements (e.g. Kemmerling 1996; Sturm and Wunderlich 2010). What speaks against the No&No response, however, are two other points. First, a disregard of the method of rational reconstruction is not recommended for HoE. Nozick (1993, p. xi) has noted that while ‘philosophy’ means ‘love of wisdom’, what philosophers really love is reasoning. To adequately understand a past philosopher’s claim, it is poor procedure to ignore his arguments for that claim. Second, guidance by systematic aspirations does not necessarily imply anachronistic treatment of the past. Such guidance may even be necessary. For instance, a standard issue in HoE is whether a past thinker developed a new claim or argument. This can only be answered adequately if one compares the relevant claim or argument with those of that thinker’s predecessors and contemporaries, examining their reasons in order to determine the precise meaning of their statements. This can require extensive rational reconstruction of past debates (see Sturm 2006, 2009).
The doubly affirmative respondents (Yes&Yes) say that we should reconstruct past epistemological views not only with an eye towards their justification, but also assume continuity between past and present epistemological aims, problems, and methods. This stance treats classical authors like great contemporaries that help us avoid repeating mistakes (e.g. Kenny 2005) or that raise our own discussion to a higher level. Bennett (1966) treated Kant in this way, as has Burnyeat (1990) Plato, and Williams (1978) Descartes. This approach has come under attack even by some analytically trained philosophers who argue that the very formulation of philosophical problems and methods often depends on the history of the sciences (e.g. Krüger 2005; Hatfield 1996, 2005) or, as in Garber’s (2005) “antiquarianism”, on even broader intellectual and social contexts. The more such contextualization we want, however, the more difficult it may become to see how to differentiate between HoE and HE with its typical call for rich contextualization. Where to draw the line here is difficult to determine. However, I do think that some histories of epistemology succeed in delimiting their tasks such that they avoid excessive contextualization and concentrate on epistemological issues. In any case, insofar as these points are combined with the idea that one can try to reconstruct past views with an eye towards their justification, these critics at least approximate the Yes&No position.
So what is preferable, the Yes&Yes or the Yes&No position? Both have advantages and limits. The Yes&No respondents are at least right in emphasizing that not all epistemological problems are perennial, that many current ones have origins worth studying, and that doing HoE in this way can help to reflect or reform today’s agendas in epistemology (e.g. Stroud 2011). As noted above, Daston holds a similar view on the function of HE—which is another reason why we must discuss which tools or methods are better suited to address epistemological questions. More on this later. For now I want to stress that I see no principled reason for claiming that no philosophical problems and methods are (relatively) permanent. Is it not striking how parts of ancient texts by Aristotle or Plato still speak immediately to us? Insofar as this is the case, the double affirmative (Yes&Yes) can surely be right.
To sum up, I think one should never respond to both questions in the negative, should sometimes respond to both in the affirmative, and should on other occasions say Yes&No. To complete the list, a No&Yes position would obviously be irrational: it makes no sense to claim one can learn from a claim of a past thinker but deny the need to examine the meanings of that claim and the rationale for it. We sometimes speak of Kant’s or Aristotle’s great insights, but we cannot intuitively or immediately grasp those insights; we must reason our way towards them.
2.3 General Remarks on the Comparison of HE and HoE
Anyone endorsing the Yes&No view need not be worried by the objection, often raised by historians of science, that philosophers assume their problems do not change over time. A similar point may be made regarding, say, philosophical concepts, methods, standards, or goals—an objection illustrated by Renn’s above-mentioned remark that analytic philosophers seek universal (i.e. unchanging) norms for scientific rationality. For instance, a defender of the Yes&No view can question whether it is advisable to reconstruct a past thinker’s argument using logical or semantic tools that were not at that thinker’s disposal, and instead examine how the argument looks in terms of the tools that were at his disposal. Even when an interpreter finds the past logic flawed and claims that the argument can be rescued by using a modern logic, that does not commit him to claiming that modern logic is valid from here to eternity. The Yes&No view of HoE is compatible with a good dose of historicity in epistemological problems, standards, and solutions.
In addition, the Yes&No view comes in different varieties. At one end of the spectrum we have epistemological problems as discussed by philosophers only; at the other end we find connections to past science. To some extent, a tight dovetailing of epistemology and science in HoE is desirable. Consider Galileo, Descartes, and Leibniz, or Hobbes and Hume, or Helmholtz, Wundt and Frege: which elements from their works count as scientific and which as philosophical? Any study of such authors and their inquiries must set its own limits and sort relevant material from that which is irrelevant. Accordingly, Carl’s (1994) work on Frege’s theory of sense and reference, Michael Friedman’s (1992) work on Kant’s reflections on natural science, Hatfield’s (1990, cf. 1996) work on the development of theories of spatial perception from Kant to Helmholtz, and Wilson’s (1995) work on scientific instruments in the early modern era and their relevance for philosophical debates over essences and realism all lie at different points along the spectrum. But despite such difficulties, the Yes&No view of HoE makes a good competitor for HE—and the more so when it brings into play relations between past epistemology and science.
So, when I compare HE and HoE in this paper, I consider only versions of both approaches that share a broad philosophical goal: those that ultimately hope to solve epistemological problems, or at least to alter existing epistemological agendas. These approaches, however, differ in method. HoE deals with epistemological questions by means of the reconstruction and evaluation of past arguments as presented in classical philosophical texts or in the more philosophical parts of scientific writings. This can be done for its own sake (or for an improved understanding of past thinkers) but it also at least makes it possible for us to compare them with current arguments—although such comparison often involves bringing into play not only the historical arguments themselves but also the problems, methods and broader goals behind them, given that these may have changed, too. In contrast, HE focuses on epistemic aspects of past science, such as the emergence of new epistemic concepts and objects, the varieties and competition of epistemic standards, the structure of knowledge development but always does so by examining what the historical sources tell us about the actual research practices of science. Certainly proponents of HE do not rationally reconstruct past epistemological arguments in the way typical of HoE. Some defenders of HE even reject philosophical tools such as logic or linguistic analysis (e.g., Renn 1995, p. 2).
If a goal is shared, then one can discuss which of the proposed methods is better suited to achieving that goal. Of course, one might compare HE and HoE even if these approaches differ in both methods and goals, by saying that goals, too, can be compared and criticized. For example, if historical epistemologists should have no philosophical ambition after all, or should confine themselves to purely descriptive or explanatory goals of the history of science, we might question such goals. It is, however, too demanding to do so here.
So, which of the methods better suits the shared goal? In the second half of this essay, I shall present arguments that speak for those of HoE. One might try to develop a general argument in favor of this claim; but presumably, this would not convince adherents of HE, given their skepticism about such arguments. It is better to work with a striking example, a specific epistemological topic where there exist contributions from practitioners of HE and HoE. I’ve chosen a problem with an interesting and rich history in the history of philosophy and the sciences: the relation between perception and judgment, and its role in epistemological questions.
3 Perception and Judgment
3.1 How Perception’s Role in Knowledge Leads to Epistemological Problems
The concept of perception has a long history in theories regarding the justification of empirical knowledge. To know about the world around us, we must perceive it. We expect that perception can play such a role because it is passive: it gives us information about the world whether we like it or not. However, familiar problems arise from this commonsense assumption. Here are two:
The first problem derives from the existence of misleading perceptions, such as hallucinations and illusions. How things are and how they appear to us might not be the same—a point involved in puzzles, some of which have been discussed since ancient times (see e.g. Burnyeat 1979). A common suggestion is that we must sort out reliable or veridical perceptions using certain fundamental judgments—particularly judgments about our perceptual relation to reality. The second problem, however, threatens to undermine this solution. It pushes further the claim that perception must be closely related to judgmental activities if it is to play an epistemic role: To empirically refute or justify a knowledge claim requires that the content of one’s relevant perception share the judgmental content of the relevant knowledge claim. Otherwise, we cannot view perceptions as being inferentiallyrelated to knowledge claims (McDowell 1994; cf. Schumacher 2004b). My knowledge that there are more than three persons in a certain room requires someone’s perception of them, that someone being me or a reliable expert who transfers his knowledge to me. But it is also necessary that my perception be structured in a certain way, namely in such a way that the perception can be used to justify the knowledge claim. When concepts, judgments, or inferences enter, however, the perceptual window to the world may become smudged or scratched. Also, if and insofar as judgments fully permeate perceptions, it may be that we are not checking empirical claims by means of perception but only reinforcing what we already believe. It would be nice if there were some non-conceptual or non-judgmental content of perception after all. This, however, leads back to the first problem.
In various ways, the two problems have been a longstanding source of perceptual theory and epistemology. Descartes, Kant, Helmholtz, the Logical Empiricists, Popper, Hanson, Kuhn, and many others have given different answers to these two questions. Even when terminology, basic assumptions, and goals differ somewhat, the differences between today’s epistemology and that of the past are rather minor. How do proponents of HE deal with these epistemological problems of perception? In the following, I shall discuss two major proposals.
3.2 Wartofsky: “Perception has a History”
In an essay written in 1973, Marx W. Wartofsky claimed that “perception has a history” and, moreover, that “several traditional philosophical characterizations of epistemological questions are wrong, and […] what is needed to replace them is an historical epistemology” (Wartofsky 1979, pp. 189, 191).3 Two questions arise here: First, how is the claim that “perception has a history” related to the idea of an historical epistemology of perception? Second, might an HE of perception solve, prevent, or help redefine central epistemological problems about perception, such as those mentioned above?
As to the first question, Wartofsky has two suggestions: On the one hand, such an HE could be a history of perception itself; on the other hand, it could be a history of theories of perception. He prefers the first solution and considers the second option a “meta-theory” to HE (Wartofsky 1979, p. 198). One might think here of work in the history of perceptual theory and its relation to epistemology (e.g., Hatfield 1990, 2002; Schickore 2006; Turner 1994; Wade 2005).
Wartofsky’s point is not that our perceptual abilities have a “natural” history to be studied by biology, neurophysiology, or psychology (1979, 190–194; cf. Wartofsky 1983a). Rather, the history of “human perception begins to develop only” where its natural histories end; it begins with human “praxis” or action. In other words, perception is historical because it is causally influenced by actions and intentions.4 It is part of a complex “feedback loop” in which it causally influences, and is causally influenced by, the two basic human activities of communication and production, and the resulting artifacts and representations. Artifacts and representations are tools for refined action, but only by going through a stage of refined perception. Refined actions lead to more refined tools and representations, which lead to even more refined actions, but only through a stage of even further refined perception… and so on. Wartofsky (1979, p. 205f.) concludes, therefore, that there is no “perceptual neutrality”: “for perceiving organisms, what is ‘there’ is always a product of their activity, and that cross-section of the world which this activity encounters and transforms into an environment.” To use one of his examples, when a hunter hears the crack of a branch, or sees a sudden flight of birds, he “transforms that very sound and sight into an artifact—an instrument—of the hunt itself” (ibid., p. 206). A second conclusion—not distinguished, but in fact different—of Wartofsky’s considerations is that not only are our perceptions caused by our intentions and actions, but “nature itself” is caused in this way (ibid.). Influenced here by Kuhn, he claims that the “plasticity of perception is evidenced, for me, by the fact that, as styles or canons of representations change, historically, the world has seen changes as well” (ibid., p. 206).
3.3 Problems with Wartofsky’s Views
Even if we dismiss the second, more problematic conclusion,5 Wartofsky’s argument for the first conclusion—that perception has a history—is not convincing.
To begin, Wartofsky frequently wavers between saying that perception is causally dependent on human action, and that it is a mode of human action.6 If Wartofsky claims that perceptions are not identical with, but are caused by, actions, and therefore have a history, then this is trivial. No one doubts that perceptions can be caused by actions and that products of conscious human action have a history. If, on the other side, Wartofsky would stick to the identity claim, then this is quite doubtful. To show how murky Wartofsky’s views are, one must realize first that he also alternates between speaking of perception as perceptual processes and as perceptual faculties.7 In which sense could perception be considered a form of action? Obviously, faculties cannot properly be called actions, being processes of a certain kind. Wartofsky must mean perceptual processes. But that is implausible as well. We are given a few examples—a hunter perceiving a flight of birds, say—but no argument for why all perceptual processes are actions. The examples are also highly debatable. When a hunter hears a crack or sees birds fly up, his attention to the events, and therefore his consequent perception, does not show that the perceptions themselves are actions. More specifically, the contents of the hunter’s perceptions will still be something passive. There are strong limits to the plasticity of perception, as is well known from perceptual illusion (Fodor 1984): When I see a pencil bent in a glass of water, I can know that it is not actually bent, and I may even know why the illusion occurs (because of the refraction of light in water). Knowing that a perception is illusory, or why it is so, normally does not eliminate the illusion. This “cognitive impenetrability” of perceptual illusion shows that perceptual content is at some distance from judgmental activity. Similarly, if the hunter realizes that he did not actually hear a flock of birds but was perhaps a victim of some intricate illusion set up by Greenpeace activists, he will nevertheless have had certain perceptions, quite independent of his intentions and actions. And, most likely, even if the subject isn’t a hunter, he will still perceive something looking like a flock of birds.
Furthermore, there is also the question of how an HE of perception would change the agenda of traditional epistemological problems of perception. Of the problems described earlier (Sect. 3.1), Wartofsky seems to have something like the second one in mind. As he says, perceptions can be influenced by conceptual activity, in which case they may vary according to “alternative contexts, or situations, or cultures”, though not historically in his own sense (e.g., 1979, p. 190). But he does not elucidate how, given his historicist view of perception, one can overcome or transform the epistemological problems concerning perception connected to this assumption. So much for Wartofsky.
3.4 Daston and Galison: Research Practices Shape Which Perceptions are Epistemologically Relevant
Here we have two important claims: (A) Daston and Galison want not to study the history of perceptual theories but to focus on, say, “practices of seeing” instead of theories of vision and (B) claim that these practices are nevertheless of “philosophical import” because they (in conjunction with theories) “dictate not just how the world looks but also what it is—what scientific objects are and how they should be known.”
… close consideration of these practices seldom enters into the ancient and still continuing philosophical debate about the epistemological status of vision per se. Whether vision is repudiated as a false guide, leading the unwary astray with the gleam of mere appearances, or defended as the noblest or most intellectualized of the senses, it is conceived abstractly in this debate, as the same faculty for Plato and George Berkeley, René Descartes and Arthur Schopenhauer. Proponents and opponents treat theories and valorizations of vision historically and with discerning attention to nuance, but they rarely address the activity of seeing. … we have focused on practices of seeing, rather than theories of vision. We nonetheless hold these practices as well as theories to be of philosophical import. They dictate not just how the world looks but also what it is—what scientific objects are and how they should be known (Daston and Galison 2007, p. 368f.).
Daston and Galison investigate the history of the ideals and practices of atlas-making from the eighteenth to the twentieth centuries. They identify three epistemic ideals that emerged successively and that shaped practices of scientific objectivity as illustrated by atlas images (ibid., pp. 18, 42–50, 371): “truth-to-nature” (the researcher selects and synthesizes observable features in order to then visually represent the general essences of objects), “mechanical objectivity” (the researcher tries to represent the individual particularities of observed objects independent of any preconceptions, and typically uses mechanical devices to achieve those representations), and finally “trained judgment” (the expert interprets certain patterns of objects in order to group them as families). These ideals, so the argument goes, lead to a difference in the training of perceptual powers in scientists. Moreover, they determine (at least in part) what counts as empirical evidence. Certain kinds of perceptions and the related empirical claims may be acceptable for one scientist but not for another. Standards of evidence can vary with epistemic goals. In other words, we cannot take methods for granted, but must evaluate them relative to certain research goals.
From a historical point of view, the interest in research practices provides a more comprehensive picture of past science; it concretizes how scientists understood their goals and methods; it helps check whether they could live up to ideals of evidence; and it reveals when ideals were influenced by new instruments, or how instruments or images were adapted to new ideas regarding empirical evidence. So, there is a good deal to be said in favor of claim (A).
3.5 Problems with Daston and Galison’s Views
There is, first, the claim (B*) that “practices of seeing … dictate not just how the world looks but also what it is—what scientific objects are and how they should be known”. The last ten words of this statement can be read as referring to what scientists, based on their “practices of seeing” should believe about objects and how those objects should be known. Since Daston and Galison maintain that their historical analysis of these practices is of “philosophical import”, the statement cannot be merely an expression of what scientists believe as a matter of fact about their practices, irrespective of whether these beliefs are right or wrong. But what might that import be? One cannot simply use historical evidence for certain research practices and their development to contend that this evidence already shows what the world actually (or at least on our best considered judgments) looks like or what objects it contains. That would be committing a genetic fallacy, against which Daston (1994, p. 284) justly warns.
One might also simply concentrate on a certain part of the claim as follows: (B**) “practices of seeing … dictate … what scientific objects are and how they should be known”. Here the problem is that it is unclear how this is related to the further claim about the historical variance of standards of evidence that is so important to Daston and Galison’s work. One can accept that standards of evidence develop historically, namely relative to changing epistemic ideals, without accepting that “practices of seeing” sufficiently determine what counts as a scientific object and how it should be known. The careful reader of the long passage quoted above may have noted that Daston and Galison do not maintain that practices by themselves would “dictate … what scientific objects are and how they should be known”. Rather, in the previous sentence the authors state that they take both “these practices as well as theories to be of philosophical import”, followed by the claim that “They dictate not just how the world looks but also what it is…”. And yet practices of perception do not even play a necessary role in this. According to Daston and Galison’s own argument in favor of the variance of standards, it is changes of epistemic ideals—truth-to-nature (get at the essences!), mechanical objectivity (record all minute particulars!), and trained judgment (discover family resemblances!)—that explain why standards of evidence change. Plausibly, practices follow.8
It is unclear how a history of practices of perception can help to solve standard epistemological problems about perception and judgment (Sect. 3.1). In various ways, Daston and Galison identify versions of these problems: e.g., as the attempt of those who subscribed to the ideal of mechanical objectivity to avoid distorting judgments even if in fact they could not (2007, p. 320f.) or as a “worry about variable observers” (ibid., p. 369; cf. p. 278). But they do not pretend to solve these problems. Rather, they want to emphasize the plurality and sometimes even the conflict of epistemic ideals or virtues: “a plurality of visions of knowledge … is likely to be a permanent aspect of science” (ibid., p. 371; cf. e.g. pp. 33f., 40, 370). Whether that is true or not, it does not help to solve any epistemological problems with perception by analyzing research practices.
Finally, if the variance of epistemic ideals does the main job in explaining why standards of evidence have changed, then it seems reasonable to tell a different kind of story: namely, one ought to look more closely at why scientists came to accept certain goals or ideals, and how they developed views about scientific methods and standards from them. One could—and probably should—reconstruct the fine-grained details of the epistemological arguments and debates in which the scientists were involved. Of course, this would mean using a standard method of HoE rather than of HE.
3.6 Perception and Judgment in the Moon Illusion
Now I would like to further strengthen the view that the reconstruction of the reasoning, that is, the precise arguments used in the debates among past authors, is better suited to address ongoing epistemological debates than examining research practices and how they have led to the emergence of certain epistemic concepts. I use a topic that remains linked to problems of perception (Sect. 3.1), but—in order to reach historical epistemologists as well—in the context of the history of science: perceptual illusions. Much past research on these illusion was motivated not only by the wish to explain them but also to learn how to avoid drawing misleading inferences from them (Schickore 2006). We need to understand why a pencil placed in water looks bent even though we know that it is straight, why visual sensations of rapidly moving bright objects (lightning, comets, or the like) persist for some time after their external causes are gone, and so on. Only when we know how such illusion is to be explained can we know which perceptions are reliable. Moreover, by studying whether or not higher-order cognitive processes are involved in illusion, one can also learn whether these processes are involved in perception in general.
One of the most familiar of all perceptual illusions has puzzled astronomers since antiquity, and later also physiologists, opticians, and psychologists (Plug and Ross 1989). The horizon moon appears larger than the zenith moon, while its physical size, its distance from the Earth, and even its visual angle to the retina (about 0.5°) remain unchanged. Several competing explanations for this phenomenon exist, and sophisticated experiments in recent decades have not helped to narrow down the number of theories (Hershenson 1989; Ross and Plug 2002, p. 180). Among the still popular explanations of this illusion are two: the so-called perceived-distance theories, and the angle-of-regard theories. The first—defended by Al-Haythem, Descartes, and Robert Smith, among others—states that any object seen through filled space (such as the moon across terrain at the horizon) appears larger than an object at the same visual angle but seen through empty space (such as the zenith moon). The second theory—held by Ptolemy, Berkeley, and others—claims that the zenith moon looks smaller because the viewer lifts his head and eyes to an unusual position.
In the twentieth century, the angle-of-regard theory was experimentally defended by the psychologists Edwin G. Boring et al. in the 1930s and 1940s (Boring 1943; Holway and Boring 1940a, b; Taylor and Boring 1942). Against their views, Lloyd Kaufman and Irvin Rock launched objections in the 1960s, defending the perceived-distance theory (Kaufman and Rock 1962a, b; Rock and Kaufman 1962a). They used devices to measure virtual “moons” and added or eliminated terrain between viewers and those “moons”. The effect of terrain was established in both directions: Excluding terrain from horizon “moons” eliminates the illusion while adding terrain to zenith “moons” creates the illusion. These data remain uncontested to this day.
But the perceived-distance theory does face objections, one of which Boring was eager to point out: If the mind, being presented with the retinal image of the horizon moon, takes into account the seemingly greater distance to the horizon, and then computes from this the size of the moon in this position, then people would not perceive the horizon moon as closer than the zenith moon. But they do. Subjects frequently report that the horizon moon looks closer—and the zenith moon looks farther away! More dramatically, they report that the horizon moon appears to float somewhat in front of the horizon. This is called the “secondary” aspect (Egan 1998, p. 606) of the illusion: We are faced not only with a size illusion (the “primary” aspect), but with a distance illusion as well. Researchers also speak of a “size-distance paradox” of which the moon illusion is a primary example. The illusion is even considered a Kuhnian anomaly for a basic law of perception, the size-distance-invariance hypothesis (SDIH).9
Kaufman and Rock did not neglect the secondary aspect, but they denied that it refutes their explanation. They distinguished between “levels” of perception (Rock and Kaufman 1962b, p. 908): perception involves judgment, whereas mere registration does not. On the one hand, the mind unconsciously “registers” a certain distance on the basis of a certain cue (visible terrain). Then, by taking account of this registered distance, the mind is said to achieve a representation of a certain apparent size of an object. On the other hand, the reports on the apparent closeness of the horizon moon ought to be understood as statements about consciously “perceived” distance, based on the registration of the moon as particularly large, and merely elicited by the experimenter’s question. This is why their explanation has been described as the “further-larger-nearer theory” (Dees 1966; Plug and Ross 1994, p. 327).
Rock and Kaufman (1962b, p. 908f.) replied that their distinction was not ad hoc, but supported by experiments (e.g., by relating it to the flat-sky illusion, which is independent of the moon illusion). But instead of explaining why the horizon moon does not appear smaller due to its apparent nearness, they just repeated that the distance perception is not a phenomenal experience but only a judgment. They insisted that the apparent nearness of the horizon moon is an experimental artifact.
Under Emmert’s law the moon is thus, because of its remote registry, perceived as large. Looking large, it seems, under Euclid’s principle, near. And might one not add that, under Emmert’s law appearing near, it looks small? That would be the whole circle of the logic of these two principles: the horizon moon, being far, is big; being big, is near; being near, is small (Boring 1962, p. 905).10
Now, Kaufman and Rock used traditional epistemological vocabulary without reflecting on its adequacy. Historical comparison helps here. Would Kant have agreed with Kaufman and Rock’s view that the secondary aspect of the illusion is not sensory but judgmental? He would not. In Kaufman and Rock’s reconstruction, subjects are said to judge or perceive that the horizon moon appears nearer, not, of course, that it is objectively nearer. When Kant applies his distinction between sensory representations and judgments—which he does with reference to the example of the moon illusion (Kant 1900ff., vol. VII, p. 146)—the notion of judgment is used for what subjects state to objectively be the case. Kant uses the perception-judgment distinction to defend the senses against the charge that they mislead us, when in fact a careless use of the understanding is responsible for erroneous judgments. Thus sensory representations as such need not involve any judgments. Of course, our judgments may turn out to be false. For instance, we err if from our perceptions we infer that the moon is actually bigger at the horizon than in the zenith. But we do not err when we simply apprehend the horizon moon as being near, or when we report this perception. And although Kant does not mention the secondary aspect of the illusion, his distinction allows him to view this aspect as being as phenomenally real as the first one. Kaufman and Rock’s distinction does not.
This shift in the meaning of the perception-judgment distinction did not appear out of nowhere. Since the nineteenth century, especially due to the influence of Helmholtz, the idea of unconscious judgments and inferences embedded in perception itself became popular—somewhat ironically, in reaction to a (mis)understanding of Kant’s distinction between appearances and things-in-themselves (Schickore 2006). Helmholtz used the notion of unconscious judgments and inferences to explain the moon illusion as such, which is different from Kant’s attempt to analyze the epistemological role of judgment in making empirical knowledge-claims, namely to make objective assertions. The notion of unconscious inferences remains present in Rock’s theory of perception (Hatfield 2002), but the epistemological aim of objective knowledge no longer grounds usage of the perception-judgment vocabulary in current psychological research.11 However, given problems such as those raised by Boring, we should be careful when transferring this vocabulary from one domain or agenda to another.
What does all this mean for epistemological problems of perception? As to the first epistemological problem—how things appear and how they really are might not be the same—the rational reconstruction of the foregoing historical episode reveals that it is too simple to claim that we should check perceptions by means of reliable judgments or theories about our relation to reality. This does not mean, of course, that the moon illusion does not occur. But we do not know why it occurs. As to the second problem—how to prevent perception from being biased by conceptual or judgmental activities—the episode supports the view that we can distinguish conceptual and non-conceptual content in perceptions. It is only when we take an epistemological interest in perception that we must judge whether or not the perception presents things as they are. Of course, there are other options for dealing with the problem, such as denying that all conceptual activity involved in perception must lead to biases (e.g. Schumacher 2004a). But discussing these options would go too far here. It is interesting, though, that the rational reconstruction of debates and arguments in past science invites us to think about such options.
I have argued against the usefulness of HE, and in favor of HoE, for discussing epistemological problems. HE suffers from ambiguous goals and fails to make clear how it might improve or change epistemological debates.
To support HoE further, I have rationally reconstructed arguments from a short but significant historical debate on research in perceptual illusion. It has become clear that our understanding of the relation between perception and judgment has changed. But that involves no historicity of perception as such. The story is not one of perception relative to practices, but one of understandings of the concepts of perception and judgment relative to problems and arguments that philosophers and scientists face. Also, I have shown how this historical debate can be related to familiar epistemological problems about perception. All this becomes clear only if one invests a sufficient amount of rational reconstruction and critical appraisal of past scientific and philosophical reasoning. It is perhaps obvious why HoE does better than HE: some of its standard methods are not much different from those of epistemology simpliciter.
Still, my conclusions are limited. I did not show that studying research practices is never epistemologically useful. For instance, when asking whether scientific standards are realizable, or whether experiments lead to artifacts, understanding research practices can be quite useful—but, again, only if their analysis is firmly embedded in the reconstruction of arguments. Moreover, my conclusion is limited to one epistemological topic (though an important one). But it should be clear how to develop analogous examples, and how counterexamples would have to be presented to overcome or limit my doubts about HE. For now, I continue to prefer the standard methods of HoE over those of HE.
Exceptions being, e.g., the philosophers Marx Wartofsky and Ian Hacking (below more on both). The latter prefers to describe his work not as belonging to historical epistemology but “historical meta-epistemology”, arguing that the former notion refers more to historical accounts of the development of scientific knowledge whereas his—and Daston’s—work is more about the historical development of epistemic concepts such as ‘objectivity’, ‘rationality’, or ‘knowledge’ (Hacking 1999). Moreover, “historical meta-epistemology” supposedly falls under “historical ontology” (Hacking 2002, pp. 7–12). Things get complicated here, because at least one other version of historical epistemology, namely that of Hans-Jörg Rheinberger, is about what Hacking calls “historical ontology”, or the history of objects of science. It suffices to note these terminological juxtapositions; from now on, I shall ignore them (a few reasons for considering Hacking an historical epistemologist can be found in Kusch 2011).
Historians resent the term “rational reconstruction”, but they need not. As Kitcher (1993, p. 13) says, “philosophically oriented history should not reconstruct in the sense of drawing lines that would have to be altered in a more detailed presentation”. Admittedly, rationalizations often do not provide the true causes for the adoption of a belief or the acceptance of a method. Whether Kant was truly awakened from his dogmatic slumber by Hume’s reminder about causation is doubtful (Carl 1989). And according to Schaffer (1994), Kekule’s dream story of how he discovered the benzene ring was made up afterwards. But it still is possible that rationalizations are sometimes correct explanations. Moreover, we should distinguish between reasons for the acquisition of and reasons for sustaining a belief. Even when a belief has not been adopted for the reason the philosopher or scientist officially states, it may well be sustained later on for that reason. Finally, rational reconstructions are methodologically recommendable: When an author makes a certain claim only once, or without any premises on which the claim is based, or without any inferences made from the claim, then the claim should perhaps not be taken seriously. In contrast, claims involving many premises and consequences give good reason to believe that the author meant them seriously. So we should look for these kinds of inferential items when trying to identify an author’s beliefs.
The paper is related to others in his work (e.g., Wartofsky 1976, 1983a, b, 1987; cf. Gould 2003, p. x; Dolling 2003), and is probably the first publication where HE was advanced in the English-speaking world. Wartofsky (1928–1997), an editor for the Boston Colloquium for the Philosophy of Science, organized various events during the 1970 s to promote his approach (see e.g., a session on “Historical Epistemology and Scientific Practice”; Anonymous 1975). Another paper, from 1977, reveals that Wartofsky was aware of Bachelard, Foucault, and Lecourt. So, he might have taken the term ‘HE’ from that tradition (Wartofsky 1979, p. 121)—but he does not mention any doctrine he adopted from it.
Wartofsky’s approach does contain Marxist elements. For instance, he claims that “our modes of cognitive practice change with changes in our modes of production, of social organization, of technology and technique” (1979, p. xxii; cf. Gould 2003, p. x). However, his arguments for the historicity of perception do not depend on these elements.
This conclusion even questions some of his own statements, e.g.: “My own view… is an explicitly realist view of perception in… that the ‘objects of perception’ are taken to be independent of perception, though they are mediated by an activity of perception” (Wartofsky 1979, p. 193).
For the claim that perception is to be explained or caused by actions: “… the forms or modes of perception, its structures themselves, are historically variant;… this variation is related to historical changes in the forms or modes of human action” (Wartofsky 1979, 189). Or: “… in its very genesis, perception is linked to that practical interaction with an external world” (ibid., 194). For the claim that perception is itself a kind of action: “… perception is a highly evolved and specific mode of human action (or praxis)” (ibid., p. 189). “I take perception to be a mode of outward motor action” (ibid., 194). “… perception is understood as a mode of human action;… it is therefore… endowed with all the qualities of human action or praxis, namely: effectiveness in the world (causal efficacy); intentionality (as it is involved in the conscious teleology of human action); and, necessarily, a mode of physical or organic activity… and exhibiting… the specific features of reflexiveness or internal activity characteristic of such other organic functions as digestion, emotion, or hormone balance” (ibid., p. 196).
For the faculty view: “I take human perception… as the specifically human faculty which develops only after biological evolution of our sensory system has been completed. That is to say, I take it as an historically evolved faculty…” (Wartofsky 1979, p. 189).
In conversation, Daston has emphasized that much of her claims concern the emergence of epistemic concepts: practices precede the “crystallization” of a concept like objectivity. One can grant this only to a certain extent: (1) If one writes a history of the emergence of X one must already have a preliminary (if minimal) understanding of X, since only then can one sort out historical material properly; also (2) for the material to be sorted out properly, there had to be an at least implicit understanding on the side of the historical agents that what they were doing falls—in a minimal sense—under ‘X'. Daston justly cautioned me that this is true for any history whatsoever, even for biological accounts of species development. The further one goes back, the less the ancestor resembles a current organism (or epistemic concept). Still, without the concepts or ideals of practices we cannot even speak about the latter.
The SDIH describes the relation between the perceived size S of an object with a given visual angle α and the perceived distance D between object and observer: S = tan α × D. This is a function based upon the physical facts about the relation between the size of an object, its distance to a viewer and the visual angle of the object on the retina, formed by the light traveling from the object to the viewer (Hershenson 1989).
Emmert’s (1881) law states the same relation as the SDIH but only for afterimages: Create an afterimage on your retina by looking at some color stimulus for a sufficient time (30 or 60 s will do). Next, look at a white wall at a greater distance. The afterimage appears “blown up”, depending on the distance to the wall.
I thank Alix Hui, John Carson, Uljana Feest, Kyle Stanford, Jens Timmermann, and two anonymous referees for valuable comments and criticism. Cynthia Klohr made helpful suggestions for wording the text. Special thanks go to Lorraine Daston, Hans-Jörg Rheinberger and Jürgen Renn, whose work provoked me to think more clearly about the relation between philosophy and the history of science. Completion of this essay was supported by the Max Planck Institute for the History of Science (Berlin) and by the Spanish Ministry for Science and Innovation, Reference number FFI 2008-01559/FISO.