1. Introduction
Sometimes, our scientific advances demonstrate that our ordinary ways of talking about the world are misguided. Other times, however, they simply give us more detail as to what makes our everyday talk true. Unfortunately, it’s not always clear which of the two cases we’re in.
Spurred on by the empirical successes of scientific theory, we’re often tempted to privilege our scientific ways of talking, treating our everyday talk as convenient but false. Bernard Williams labelled this temptation scientism and warned against it throughout his work, especially in ‘Philosophy as a Humanistic Discipline’. Unfortunately, many who share Williams’ suspicions are often tempted to overreact, ignoring the fact that sometimes science does prove our intuitive conceptions incorrect.
In this paper, I propose a concrete framework for sailing between these two temptations. I suggest that we focus on how scientific advances reveal the indeterminacy of our ordinary language. Once we accept that, we can, by reflecting on what we want them to do, go on to mould our words in a way that accommodates the relevant scientific advance. This allows us to heed Williams’ warning, whilst acknowledging how scientific advances can change our view of reality. As a case study, I’ll discuss an exchange between Arthur Eddington and Susan Stebbing on the significance of the Rutherford-Bohr model of the atom for our conception of ordinary objects.
The major consequence of this proposal is that to answer the question, we must consider not just empirical theory but also our ways of living, and what we want from our everyday thought.Footnote 1 This, indeed, is another major lesson from Williams. For in essence, my proposal is that to answer this question we must practice philosophy as a humanistic discipline as Williams suggests we do. The importance of reflecting on how our concepts serve our lives is a well-recognised part of Williams’ moral philosophy. My proposal demonstrates that the idea applies to our ordinary language concepts more broadly.
2. A Question of Two Images
Often scientific theory presents a picture of the world quite unlike our ordinary conception of it. This provokes a question: how should we think about our ordinary conception under such circumstances? The puzzle is commonly put using Sellars’ (Reference Sellars1963) notion of the manifest and scientific images. The manifest image is roughly the world as it appears to us pre-theoretically. As we pursue inquiry, we build a new, theoretically informed, picture of the world – the scientific image.
To make the question precise, let’s think of the manifest and scientific images linguistically.Footnote 2 The manifest image is the collection of our ordinary claims about the domain. The scientific image is simply our best scientific theory of the domain. So understood, the question is: assuming that our scientific theory is correct, are our ordinary claims systematically mistaken? I’ll call this the question of the two images.
This is, of course, an oversimplification in multiple ways. Our ordinary conceptions don’t obviously cohere into one clear picture.Footnote 3 The manifest image is a rational reconstruction of our ordinary conception.Footnote 4 Our scientific theories don’t obviously cohere into one image either.Footnote 5 As Ryle (Reference Ryle1954: 3) put it, ‘there is no such animal as ‘Science’. There are scores of sciences’. It’s best to talk about the issue relative to domains (not that it’s always clear how these divide). I’ll use a case study for this reason.
Generally speaking, however, the scientific image can make the manifest image look inadequate, incomplete, and/or false.Footnote 6 The scientific image is often more precise, systematic, and detailed. Additionally, the former enjoys rigorously tested empirical support. Considering this, one might be tempted to adopt error theory: in the face of the apparent incompatibility between our ordinary and scientific claims, view the former as systematically failing to track reality.Footnote 7 Error theory is the combination of two claims: (i) the relevant terms and claims are meant to refer to and describe part of reality but that (ii) no part of reality answers to that description.Footnote 8 The discourse is therefore, in J. L. Mackie’s (Reference Mackie1977) sense of the term, systematically erroneous/mistaken.Footnote 9
Call the alternative position compatibilism. Compatibilism is defined by the rejection of error theory. The two are therefore exclusive and exhaustive. Compatibilism is not the claim that we cannot be mistaken. Rather, it is simply the view that our ordinary discourse is not systematically mistaken in the above sense.
As I’ll argue in §5.3, error theory is sometimes appropriate. That said, all too often it is motivated by scientism. ‘Scientism’ is neither a well-defined term, nor is the phenomenon it picks out a homogenous one. Midgley (Reference Midgley2001: 40; Reference Midgley2003: 22) and Haack (Reference Haack2003: 18–19) rightly think of it as a general attitude with which one thinks about the place of scientific thought in inquiry. The key characteristic of scientism, as Williams (Reference Williams2000: 479) highlights, is the thought that non-scientific intellectual inquiry ought to either get out the way of, emulate, or even be subsumed by, science. This is a kind of scientific methodological imperialism: all inquiry ought to be conducted, methodologically, as scientific inquiry is.Footnote 10 The result is a privileging of our scientific representations over others.
There is much to be lost by adopting scientism. This can be seen if we turn to Williams’ notion of a local concept.Footnote 11 Some concepts, and some representations, are local in that they display the idiosyncrasies with which we apprehend the world. There are many ways this can happen. Williams often focuses on thick ethical concepts like liberty or truthfulness.Footnote 12 He claims that these concepts combine descriptive and evaluative elements in such a way that they betray the evaluative outlook of the group that uses them.
This is not the only way in which a concept can be local, however. More generally, a concept is local if (and only if) one cannot use it to directly represent without being a part of the relevant group.Footnote 13 An explanation of someone having knowledge which utilises a local concept also requires reference to them being part of the relevant group (e.g. for liberty, a particular socio-cultural group). Williams contrasts this with scientific concepts and representations which abstract away from our representational idiosyncrasies, thereby describing the world in a minimally perspectival manner.
The term ‘minimally’ is crucial.Footnote 14 The difference between local/non-local concepts and representations is a difference in degree.Footnote 15 The less local a concept is, the more appropriate it is for the task of creating minimally perspectival representations.Footnote 16 Scientism, then, can be seen as the view that the less perspectival a representation is, the better it is.
It is an underappreciated feature of Williams’ picture that many (the vast majority) of non-normative ordinary language concepts are local (the consequences of which I’ll return to in §6.3). For example, consider colour concepts. As Williams (Reference Williams1985: 139) discusses, the perceptual idiosyncrasies that are central to possessing our colour concepts make them local. The ordinary language concept solidity, which be central to my case study below, is another example. It is, to use Williams’ phrase, rooted in our perceptual faculties (our mechanosensory system specifically) such that creatures radically unlike us couldn’t use it to directly represent.Footnote 17 There are many ways in which an ordinary concept can be local without considering normativity. Williams (Reference Williams1985: 139) even cites grass as a local concept.
Williams shows us is that there are multiple reasons why we need or would want local concepts alongside scientific ones. The failure of scientism is the failure to recognise this. I discuss them in turn.
First, there is much that minimally perspectival concepts cannot be used to represent. For us to discuss the colours of the world we require local concepts like green. Judgements about the colours of objects would never make it to a minimally perspectival representation of reality. Yet certain colour judgements are true, and others are false. We can provide different information about one and the same entity by using ordinary language concepts and the concepts of physics, respectively and this can lead to new knowledge.Footnote 18 Minimally perspectival true representations do not exhaust all we can know (Williams Reference Williams1985: 139; and Ryle Reference Ryle1954:74). Maligning our ordinary language representations for their local character involves missing out on knowledge of the world.
Second, our local concepts help us to interact with the world. For example, it is much easier to pick out, track, classify, and interact with birds by using our colour concepts than it is by using the concepts of atomic theory or fundamental physics. Finally, local concepts are much more convenient tools of communication. Again, descriptions of birds featuring colour concepts direct the attention of others to our intended subject matter better than those featuring atomic theory. We have need of local concepts. Endorsing scientism involves losing sight of this.Footnote 19 I’ll go on to argue that scientism also involves a loss in self-knowledge as well (§6.2–3).
That said, there is an observe problem. That is, responses to scientistic thinking frequently involve a reactive global compatibilism. I’ll argue this represents just as much of a failure as scientism and show that we can sidestep both if we embrace my reframing of the question of the two images (§5.4), and thereby practice what Williams calls philosophy as a humanistic discipline (§6). To do all this, I’ll need a case study.
3. Atomic Structure and Solidity: A Case
The Rutherford-Bohr model of the atom was produced in 1913 after Rutherford’s scattering experiments overturned Thomson’s plum pudding model. The Rutherford-Bohr model features a dense nucleus at the centre of the atom with orbiting electrons. According to the model, therefore, the atom’s structure features empty space. In contrast, Thomson’s model posited that atoms were continuous spheres of positive charge (conceptualised akin to liquid) with floating electrons. The Rutherford-Bohr model suggested that, contrary to previous expectation, everyday objects are mostly constituted by empty space.
In his 1927 Gifford Lectures, Arthur Eddington (Reference Eddington1928: 342) claimed that the discovery of the model showed that everyday objects such as wooden floors have ‘no solidity of substance’. He added that to step on a wooden floor ‘is like stepping on a swarm of flies’ that, somewhat marvellously, ‘give me a boost up’ each time I step such that I may continue walking. From this discovery about the structure of the atom, Eddington inferred that nothing possessed the property of solidity, and so our ordinary ascriptions of solidity are systematically erroneous. Eddington thus adopted an error theory about solidity.
A decade later Susan Stebbing published Philosophy and the Physicists in which she provided a compatibilist response.Footnote 20 Stebbing claimed that the discovery in question was not that everyday objects aren’t solid. Nor was this implied by said discovery. Rather, the discovery was that solid objects are constituted by atoms within which there is empty space. The discovery neither altered the extension of the concept of solidity, nor demonstrated the systematic falsity of our judgements. Rather, it gave us a more detailed conception of that in virtue of which objects, like tables, are solid (Wittgenstein Reference Wittgenstein and Rhees1958: 45, and Stebbing Reference Stebbing1937: 51).
Stebbing argued that this follows from reflections on how ‘solid’ gets its meaning. The meaning is determined by its everyday use, most centrally by the paradigmatic applications of the term. Stebbing (Reference Stebbing1937: 52) writes:
“solid” just is the word we use to describe a certain respect in which a plank of wood resembles a block of marble, a piece of paper, and a cricket ball, and in which each of these differs from a sponge, from the interior of a soap-bubble, and from the holes in a net.
As such, Stebbing argued, we cannot make sense of the error theory regarding solidity. If it is right that the Rutherford-Bohr model shows that no objects are solid, then there are no correct paradigmatic applications of the term ‘solid’. Yet, since such applications are what give the term its meaning, an error theory would rob the term of its meaning. In Stebbing’s (Reference Stebbing1937: 52) words ‘we cannot understand what the denial of solidity to the plank amounts to’.
Note, on this view, it is perfectly coherent to suggest that particular applications of the predicate are mistaken. After all, a wooden bridge may seem solid but in fact be rotten. The point is that the term loses its meaning if there are no correct applications of solidity found in ordinary discourse. Stebbing (Reference Stebbing1937: 52) writes: ‘there could not be a misuse, nor a figurative use, unless there were some correct and literal usages’.Footnote 21 Note, though Stebbing appeals to linguistic considerations, her disagreement with Eddington is about whether wooden planks are solid not about how we should use the term ‘solid’.Footnote 22
4. Reframing the Debate
4.1 Open texture
That is the Eddington-Stebbing debate as it originally panned out. I think it can be improved by a reframing. The reframing will use the notion of open texture heavily, so it is worth spending some time with that idea. Open texture is a kind of semantic indeterminacy that a term displays if and only if, however we define it, there always remain possible cases in which the application of the term is indeterminate. In other words, a term is open-textured iff our extant definition and practice do not fix its correct use for all possible cases.Footnote 23 Open texture isn’t vagueness, wherein paradigmatic uses are indeterminate. Sometimes only far-fetched possibilities demonstrate the relevant indeterminacy of an open-textured term. Open texture is thus much more widespread than vagueness.Footnote 24
There are many ways that open texture can become evident. Friedrich Waismann, who coined the term, often illustrated the phenomenon with thought experiments in which we encounter a new object that we struggle to classify because it reveals indeterminacy in our kind terms. For example:
Now what would you say if a substance was discovered that looked like gold, satisfied all the chemical tests for gold, whilst it emitted a new sort of radiation? “But such things do not happen.” Quite so; but they might happen, and that is enough to show that we can never exclude altogether the possibility of some unforeseen situation arising in which we shall have to modify our definition. Try as we may, no concept is limited in such a way that there is no room for any doubt. (Waismann Reference Waismann1945: 122–123)
Waismann’s point is that it is an open question whether we ought to say this new substance is gold and not because we lack empirical facts. Rather, there is an indeterminacy in the concept gold. Our extant linguistic practices and definitions don’t fix the matter. There are many such examples.
In response to such scenarios, we often close the relevant term – i.e. our continued use of the term fixes whether it applies in these new circumstances.Footnote 25 There are multiple options for closing terms. For example, we could stipulate a new usage of ‘gold’, thereby making it polysemous – i.e. the discovered substance is gold* but not gold. We could replace the term ‘gold’ – i.e. the substance is gold* and we are not to speak of gold. We could take a rigid approach to what we thought being gold requires and claim that the new substance isn’t gold. All these options represent closing the term – i.e. determining the correct application for cases of this type.
It will not always be clear which of these options constitutes changing the meaning of the term, but we needn’t be overly concerned about this.Footnote 26 There aren’t always clear answers about such things; synonymy is a vague concept.
Open texture also forces us to rethink our prior (even paradigmatic) uses of a term, sometimes without a new sample arising. For example, consider the term ‘rodent’.Footnote 27 Guinea pigs are typically thought to be paradigmatic instances of rodents, but for a while a supposed discovery in cladistics questioned this. The supposed discovery was that the narrowest clade that includes guinea pigs as well as other paradigmatic rodents such as mice and rats is very broad. It includes paradigmatic non-rodents such as horses, and indeed, humans. Consensus on this ‘discovery’ has turned, but for my purposes its possibility is enough.Footnote 28
Were this cladistics ‘discovery’ upheld, it would demonstrate that we had made a mistake in thinking that guinea pigs are rodents, that all species in the narrowest clade that includes guinea pigs are rodents, and that humans are not rodents. How are we to respond? Perhaps we could say guinea pigs aren’t rodents, or humans are, or that there are no such things as rodents. The indeterminacy here is a matter of open texture. I’ll leave it to the reader to map them, but just as in the gold case, there are numerous ways we can respond to this open texture. There are more such examples.Footnote 29
Whether a case displays the open texture of a term is a matter of extant practices and definitions. As we carry on in our use of the term, closing it one way or another, the definitions and practices are altered. The kind of case that previously was indeterminate, is no longer. Then new (but only new) types of case will display the open texture of the term. So, when we close a term, we are usually only partially closing it.Footnote 30
It’s worth commenting on the sense in which how we close terms is a choice. By speaking this way, I do not mean to suggest that, when it comes to open texture, anything goes. Our choice answers to something in that there are reasons for closing a term one way rather than another (see §5.3). This doesn’t entail that that there is a unique correct way to close each term. We have reasons to close a term one way or another, but they might not (and often don’t) determine the unique correct closure. There are usually a range of admissible options. Furthermore, I don’t have in mind a particularly voluntaristic picture. Rarely do we consciously sit down to close a term. Rather, our language reacts dynamically, the linguistic community adopts something as correct usage and this closes the term (see §6.1).
4.2 A new frame
We can reframe the debate between compatibilism and error theory in terms of open texture. Of the various ways to respond to open texture, some are error-theoretic response and others are compatibilist. The argument between error theory and compatibilism can be seen as an argument about how to respond to open texture. With regards the solidity case, the thought is that the discovery of the Rutherford-Bohr model demonstrated that the term ‘solid’ is open-textured. Whether you are an error theorist or compatibilist about solidity depends on how you respond to the open texture of ‘solid’.
One easy way to argue that ‘solid’ is open-textured is to cook up a science fiction example. Imagine a bridge made of wooden planks that demonstrate all the standard features we associate with solidity, but which become diaphanous whenever someone named ‘Paul’ uses them. It would be an open question whether these planks are solid, and not because we lacked empirical information.Footnote 31 For my purposes, though, this example isn’t sufficient. The suggestion is that the discovery of the Rutherford-Bohr model made us systematically rethink our prior applications of ‘solid’. So, in that sense, ‘solid’ is more like the rodent example than the gold example.
There were several features we associated with solidity including (but not limited to) being impenetrable to macrophysical objects, not changing shape under pressure, being constituted by completely continuous matter. The discovery of the Rutherford-Bohr model showed us that these features don’t come together. That is, it demonstrated that objects which display the first two features do not display the third. Since these features (alongside others) characterised the concept of solidity, this gives us pause to reconsider the concept. This is an unexpected discovery that raises systematic questions about the notion of solidity – unlike the diaphanous planks! The question becomes: how central to the notion of solidity is the thought that solid things are constituted by completely continuous matter with no empty space? The more central it is, the more tempting error theory is.
Reframed, the debate becomes meta-linguistic negotiation.Footnote 32 It’s a question of how we ought to use the term ‘solid’ and about what conceptual practices we wish to keep. This compares to a first-order dispute about whether wooden planks are solid.
Ultimately, the claim that the Rutherford-Bohr model displayed the open-textured nature of the term ‘solid’ is a socio-linguistic empirical one. Its truth depends on the semantic intuitions people had in the face of this discovery. The reason to reframe the debate is primarily the philosophical payoff, however.
Before outlining this payoff, I note that I do not reject the possibility that there could be some conflicts between the manifest and scientific image that do not involve open texture. Considering how pervasive open texture is, these will be rare. My main point is that thinking about whether our ordinary terms are open-textured is useful in cases where it is unclear how to respond to the tension. These are the philosophically interesting cases. They’re best illuminated by being considered through the lens of open texture.Footnote 33
5. The Payoff
This reframing has numerous benefits. It improves the quality of the debate and enables us to develop both error theory and compatibilism.
5.1 A better debate
The first is that it characterises the debate more charitably, illuminating a tension between intuitions that pull in opposite directions in a way that Stebbing and Eddington’s exchange does not.
The intuitions are as follows. First, there is an intuition that something is solid if it’s like a plank of wood not a soap bubble (i.e. that paradigmatic applications of the term fix its meaning). This is hard to square with the thought that nothing is solid. Simultaneously, it’s intuitive that something filled with empty space is not solid. Viewing the debate as I’ve suggested enables us to recognise both intuitions. They are different considerations that go into deciding how to respond to the open texture of ‘solid’. This allows the compatibilist and error theorist to more charitably interpret one another.
Stebbing and Eddington seem unable to recognise the intuitions driving one another. This continues in Eddington’s response.Footnote 34 Stebbing (Reference Stebbing1937: 54) construes Eddington’s error as basic conceptual mismanagement, a basic misunderstanding of what the ordinary concept of solidity amounts to. Eddington’s misstep is not basic conceptual incompetence as can be seen when it’s noted that others responded as he did. Of course, numerous authors can make the same subtle conceptual mistake but a good litmus test for basic conceptual competence is whether others use the concept similarly. Eddington’s position was not uncommon at the time.Footnote 35 For example, Ernst Zimmer (1936: 51) writes that according to physics:
a table, a piece of paper, no longer possess that solid reality which they appear to possess; they are both of them porous, and consist of very small electrically charged particles, which are arranged in a peculiar way.
Furthermore, there are still people who espouse error theory.Footnote 36 In a New York Times interview David Chalmers says:
Nothing is truly solid out there in the world. Things are mostly empty space. (Ramakrishna Reference Ramakrishna2019)
Appearing at HowTheLightGetsIn festival, physicist Bernard Carr claims:
The idea that objects are solid is replaced by the idea that they are mainly empty space (the atomic picture) … Our common-sense reality, that’s what’s been demolished. (Institute of Art and Ideas 2024)
We shouldn’t treat error theory as simple conceptual misunderstanding. In fact, we should drop Stebbing’s claim that the term ‘solid’ has no sense if error theory is true and that as such error theory is incoherent. Though paradigmatic applications of the term in ordinary usage play a substantial role in determining the meaning of ‘solid’, it’s not the case that the term would be senseless if the paradigmatic applications were false. Viewing things through the lens of open texture shows us that we can recognise possible meaning assignments to the term ‘solid’ on which error theory is true (even if we remain compatibilists). I return to this with my discussion of error theory about witches in §5.3.
My reframing enables us to recognise the possibility that in many cases we could have reasonably decided differently and still developed a functioning linguistic practice. For example, we could have decided that there are no rodents, that guinea pigs are not rodents, or that humans are. None of these options represent egregious conceptual error. It may be reasonable to go either way in some cases.Footnote 37
5.2 Improved error theory
The second benefit of this reframing is that it improves error theory by developing it into a meta-linguistic position. The error theorist can suggest that it is the association to continuous matter at the microphysical level that is, or should be, most central to our application of the term. The discovery was that this doesn’t apply to any of the entities that we previously took to be solid.
This reframing also illuminates the variety of different options available to the error theorist. There are various error-theoretic ways one can close a term. These include prescriptive suggestions to change our concepts or form new ones. These prescriptive suggestions focus less on how we do use a term and more on how we should. For example, an error theorist might suggest we make ‘solid’ polysemous by forming a technical notion of solidity (about atomic structure) and an everyday notion of solidity (about, say, continuity of matter). The error theorist would then claim that only solidity* is instantiated.Footnote 38 They could remain error theorists about the ordinary notion, whilst not being so about the technical notion.
More generally, the point is that my reframing turns error theory from what seemed a flat-footed view that ignores the semantics of ordinary language into an approach attuned to such considerations. Such a transformation is made possible by thinking of things in terms of open texture.
5.3 Improved compatibilism
The third benefit of the reframing is that it improves compatibilism in three ways, each of which helps compatibilists avoid problems that have historically plagued the approach. Though I’ll keep my focus on Stebbing, I’ll also highlight how my reframe improves on an array of historical compatibilist positions along the way.
The first improvement is that it helps compatibilists avoid an implausibly stark division between our scientific and ordinary language. Compatibilists often diagnose the error theorist’s misstep as inappropriately applying concepts between the scientific and ordinary domains. For example, Ryle (Reference Ryle1954: 78) writes that the manifest and scientific image are ‘constitutively speechless’ with regards each other. Thomasson (Reference Thomasson2007) and Putnam (Reference Putnam1987) claim that the conceptual frameworks of ordinary language and scientific theory are incommensurable.Footnote 39 Stebbing (Reference Stebbing1937: 51) writes that ‘no concepts drawn from the level of common-sense thinking are appropriate to sub-atomic, i.e. microphysical, phenomena’.
This style of compatibilism blocks error theory by claiming that the conceptual frameworks of the two images are so different that the cross-application of concepts between them is inappropriate (so no conflict arises).
The problem with this is that it prohibits a unified picture by installing an unbridgeable conceptual division between ordinary and scientific language. A commitment to incommensurability makes explaining everyday phenomena using physics difficult. For example, it prohibits an explanation of what it takes at the atomic level for wooden planks to be solid. Our ordinary and scientific representations should unify into a single overall picture of reality.Footnote 40 They cannot if the compatibilist goes this route.
Reframing the issue in terms of open texture helps the compatibilist avoid this pitfall. We can close our ordinary terms in such a way that leads to incommensurability, but we needn’t, and compatibilism needn’t rely on doing so. Only in some cases will cross-application of concepts from the manifest and scientific image be inappropriate. My reframing highlights this as we can close terms differently for different reasons in different cases.
The second improvement of my reframing is that it encourages us to consider linguistic practice widely construed. This is something compatibilists have, historically speaking, failed to do. Many compatibilists (notably Anthony Flew and Norman Malcolm) overemphasise the role of our paradigmatic applications of the relevant term in their attempts to block error theory. This yields a rigid and narrow range of permissible meaning-variations for the given term.Footnote 41
A wider focus involves unpicking the inferential relations, contexts, patterns etc. that contribute to the meaning of the term. This helps the compatibilist discover a wider array of possible meaning-assignments or ways of closing terms, thereby giving them more options for explaining how our ordinary use of a term is compatible with scientific advances. This involves not just looking at how we do use a term, but further considering the ameliorative question of how we could, should, or want to (see §6.2–3).
We must think first about what we currently do with our linguistic practice. Stebbing (Reference Stebbing1937: 52) does this well. Stebbing fails, however, to ask about the ends to which we put our representations of solidity – i.e. what ends do we wish the practice of ascribing solidity serve? This will reveal possible meaning assignments that descriptive conceptual analysis does not.
Framing the issue in terms of open texture encourages the compatibilist to do this. That there are various admissible answers to the question of how to close a term pushes us to think about not just what we currently do, but what we wish to do. It pushes us to consider a wide range of meaning-assignments that fit with both our extant practice and the goals to which we put our representations.Footnote 42
The third improvement of my reframing is that it helps compatibilism avoid over-generation worries which have historically plagued the approach.Footnote 43 The thought is that Stebbing’s approach entails that we ought to be global compatibilists when we oughtn’t. I’ll illustrate this by contrasting the solidity example and another where error theory is clearly the right response.
Consider the concept witch. We have a functioning witch discourse with accepted norms of concept application.Footnote 44 The natural sciences demonstrated that many of the phenomena previously explained by witch-talk were better explained without the notion witch.Footnote 45 The right response to this is to adopt a witch error theory. Yet all the considerations Stebbing provided in favour of compatibilism apply in the witch case. Our ordinary linguistic practice (including correct paradigmatic applications) determines the meaning of ‘witch’. So, if Stebbing’s right, we cannot make sense of a witch error theory. If Stebbing’s reasoning demonstrates that we should be compatibilists about solidity, it does so here too. The argument over-generates.
Compatibilists can avoid this over-generation by adopting my reframing. Reframed, the idea is that scientific discoveries about phenomena we previously explained by reference to witchcraft displayed the open texture of the term ‘witch’. For example, as we learnt to explain disease in naturalistic ways, we learnt that witch curses don’t cause ill health (alternative explanations prevailed). This prompts recognition of the open-textured nature of witch, leaving us with a choice: are we to say that there are no witches or modify the concept to exorcise the associations with cursing?
Once we think in terms of open texture, the compatibilist can claim that how the term is used only partially determines how we should close it. This is key to avoiding the over-generation worry. Stebbing’s reasoning hinged on the fact that no sense can be made of the relevant term if we adopt error theory. Compatibilists should drop this claim.
The key consideration that splits the witch and solidity case involves what Putnam (Reference Putnam1975) calls the ‘stereotype’ associated with the terms. The stereotype is a collection of features of the entity the term picks out. Specifically, those features which a speaker must understand paradigmatic instances of the entity to display for the speaker to count as having correctly acquired the term.Footnote 46
Scientific discoveries highlight the open texture of our ordinary terms by demonstrating that something is afoot with the features of the stereotype. This includes (but isn’t limited to) showing that some features of the stereotype aren’t ever displayed, that the stereotype is missing some important features, that features appear regularly but never all together, that the stereotype collapses together entities with importantly different features.Footnote 47
In the solidity case, one of the stereotypical features (that solid things are constituted by continuous matter without empty space) was shown not to be instantiated by paradigmatically solid entities. In more detail, what Stebbing (Reference Stebbing1937: 54) calls ‘the nineteenth-century view of the ultra-microscopic world as consisting of solid, absolutely hard, indivisible billiard-ball-like atoms, which were assumed to be solid and hard in a perfectly straightforward sense’ was previously part of the stereotype of solidity but was shown not to be instantiated. Since the rest of the stereotype remained intact, there’s room for us to adapt (but essentially keep and correctly use) the stereotype of solidity in the face of this new information – i.e. there is room for compatibilism.
In the witch case, however, it was shown that none of the associated descriptions applied. As such, it is appropriate to adopt error theory about witches. It was then hard to continue to use and make sense of the notion. It’s hard to make sense of someone continuing to insist that someone is a witch whilst accepting that they do not hex people, aren’t in consort with the devil, do not eat children etc. This consideration splits the two cases. In general, the more scientific discovery causes disruption to the stereotype, the more plausible error theory becomes.
Note, this consideration introduces some vagueness. How much disruption to the stereotype calls for error theory? It’s unclear, this is a vague matter. That’s not a problem, however. It’s plausible that there are borderline cases in which whether we ought to be compatibilist is unclear. Additionally, since our practices and stereotypes change, whether we should be compatibilists can change too. For example, the emergence of new norms governing the use of ‘witch’, may have changed the stereotype. Being a Wicca practitioner may be sufficient for being correctly labelled a witch now. The scientific advances that naturalistically explain the spread of disease don’t disrupt this stereotype.
More generally, whether a discovery disrupts our linguistic practice depends on our practices, circumstances and ends and these can change. This is not problematic, open texture is a dynamic phenomenon. A term can be (partially) closed one way only to then be closed another way later.
5.4 Scientism and conservatism: Scylla and Charybdis
The final, and perhaps most important, benefit of my reframing is that it allows us to chart a course between the Scylla of scientism and the Charybdis of what I’ll call conservatism – a blind privileging ordinary language use as it is.
The key step is to recognise that we do different things with our representations and as such we need representations of varying types.Footnote 48 Crucially, then, we ought not simply ask which representations and concepts are best simpliciter, but rather which are best for a purpose. Failure to do this leads us to evaluate our representations according to inappropriate standards and that leads us astray when considering the relationship between the manifest and scientific image.
Eddington fails to ask what our ordinary solidity ascriptions are used for. He (Reference Eddington1928: xiii-xiv) speaks of the manifest image as ‘a traditional picture gallery of conceptions which convey no authentic information of the background and obtrude irrelevancies into the scheme of knowledge’. He doesn’t reflect on what these conceptions are irrelevant to, with the natural reading being the under-specified, and most general, task of gaining knowledge. He fails to appreciate that different representations and representational tools have different virtues that help them serve different ends. He assumes that ordinary language representations are to be evaluated according to the ends of scientific representations. Unsurprisingly, they come up short.
Eddington’s mistake is a scientistic one. He assumes that, using Williams’ (Reference Williams2000: 483–484) words:
if there is … one set of these concepts, those of physical science, [which] are potentially universal in their uptake and usefulness, then it follows from this that they are somehow intrinsically superior to more local conceptions which are humanly and perhaps historically grounded.
This is a mistake because:
even if it were possible to give an account of the world that was minimally perspectival, it would not be particularly serviceable to us for many of our purposes … we need concepts and explanations which are rooted in our more local practices, our culture, and our history.
As discussed in §1, this involves overlooking the contributions our local concepts add to our lives. This goes for solidity too. We can provide different information about one and the same entity by using solidity and the Rutherford-Bohr model of the atom.Footnote 49 Just because all objects are made of atoms, and so we can use an atomic theory to discuss every object, does not mean that atomic theory exhausts all that can be said about objects.Footnote 50
So, the localness of a concept doesn’t exclude it from being used to further Eddington’s most general of ends – contributing to the scheme of knowledge. Furthermore, the concept of solidity helps us interact with the world. For example, when we’re looking for somewhere to put a mug of tea, the information conveyed using ‘solid’ is more useful than any atomic model. It helps us communicate with others about where’s safe for a mug of tea too.
Note, the vices of scientism are not science’s vices.Footnote 51 Even putting aside the technological yield of science, scientific representations and concepts have virtues. For example, they are often precise and are said to carve at the joints of nature.Footnote 52 Error theory is not always scientistic either. That said, the value of scientific concepts is not an argument for error theory.
The key problem is the failure to ask about the ends to which we put our representations. We must ask ourselves: what do we want to with our different representations? Why would carving at the joints, precision, and so on be the virtues we want in a representation? It depends on what ends we wish it to serve. This should be at the forefront of our minds when closing our open-textured terms.
Though the scientistic version of this threat is well-identified by Williams, there is a compatibilist version too. As seen (§5.3), responses to scientism often involve overly quick arguments for global compatibilism. Such responses not only get us the wrong result, but, and this is underappreciated, they involve making the same mistake. Previous blanket approaches to compatibilism have missed out the variety of possible meaning-assignments we can give our terms because of a failure to recognise the variety of ends to which we do and could put our representations.Footnote 53 This leads compatibilists to blindly privilege our ordinary language as it is now.Footnote 54 This is what I call conservatism. It may be a different result, but at its core it involves the same mistake as scientism.
My approach, on the other hand, makes reflection upon our representational aims central. To ask how to close a concept is to ask what the best conceptual tools are. The admissibility of various answers highlights the importance of asking: tools for what? That is to ask: what do we want to do with this concept, the representations we create with it, and the resulting linguistic practice? Answering this requires a clear understanding of how we currently use language (as Stebbing emphasises) but also a recognition that we can change our concepts to suit our purposes (as Waismann emphasises).Footnote 55 By foregrounding this question, my reframing allows us to sail between the Scylla of scientism and the Charybdis of conservatism.
6. The Scope of the Humanistic Project
I’ve argued that focusing on the ends to which we put our concepts and terms is key to avoiding scientism and conservatism. It’s therefore key to answering the question of the two images. Here, I examine what matters about these ends. I argue that the crucial point is that they are our ends.Footnote 56 This may cause anxiety about the possibility that other groups with other ends may appropriately close terms differently. I’ll explain that we can dispel this anxiety if we adopt Williams’ conception of philosophy as a humanistic discipline.
My proposed approach to answering the question of the two images does more than show us how best to avoid both scientism and conservatism then. It shows that successfully answering the question requires adopting Williams’ picture of philosophy as a humanistic discipline. Moreover, by highlighting that we need a self-reflective humanistic response to the widespread phenomena of open texture, my proposal expands this picture beyond the usual discussions of thick ethical concepts.
6.1 Our ends
I use the term ‘ends’ in a deliberately broad manner. Our ends are multifarious and our ordinary language concepts rise to meet them in an all manner of unplanned, responsive, and dynamic ways. We have use for certain concepts to help us avoid danger, make art, communicate, joke, and organise our lives. There is no bijective mapping between our concepts and ends. Our concepts often serve our ends in a diffuse and indirect manner, such that how they do so is sometimes opaque to us.Footnote 57 More generally, the picture is that using certain concepts establishes practices that serve us as linguistic communities by enabling us to do all sorts of things.Footnote 58
The crucial unifying feature of these ends, is that they are ours. Recognising this, however, can leave us with what I’ll call the anxiety of alternatives: fear that the fact that others may close their terms differently (and adopt different concepts) undermines our closures (and concepts).Footnote 59
Williams shows us that appreciating the sense in which they are ours can in fact dispel the anxiety, however. The God’s eye point of view is neither available nor helpful for deciding which closure of ‘solid’ is best – i.e. to decide between compatibilism and error theory. Williams emphasises that in recognising the value of our local concepts we must recognise that their value comes in them being ours:
We and our outlook are not simply in the same place at the same time. If we really understand this, deeply understand it, we can be free of what is indeed another scientistic illusion, that it is our job as rational agents to search for, or at least move as best we can towards, a system of … ideas which would be the best from an absolute point of view, a point of view that was free of contingent historical perspective.Footnote 60 (Williams Reference Williams2000: 490–491)
The anxiety is driven by the desire that there be a universal stock of best concepts (alongside the recognition that there are none). It is the final throes of scientism.Footnote 61 Once we recognise this, we can rid ourselves it and accept that the appropriate closure of a term depends, at least partially, on what we wish to do with it. Our ends can serve as the source of justification for one closure over another. They do so in virtue of being ours. To take this justification to be insufficient in some way is to cling to the idea that there could be a universal justification – i.e. to cling to scientism. As Queloz (Reference Queloz2025: 122) puts it:
It is not just the concepts that are contingently ours, but also the demands they must meet. We should not seek the concepts that are absolutely best, in that anyone has reason to use them. We should seek the concepts that we have reason to use, given the demands on our conceptual apparatus in the situation that is peculiarly ours.Footnote 62
It’s worth commenting on the extent to which this introduces relativism or pragmatism into the picture. What makes for the appropriate closure of a term is, to some extent, relative to our ends. We choose to shape our concepts however best serves us. What best serves us depends on our ends. That said, once these choices are made, the truths that they are used to represent are not thereby relative to anything.Footnote 63 Additionally, a picture of the world is not true because it serves a particular end. Truth is not identified with practical value. The fact that which concepts we should use is partly relative to our ends does not imply that our concept choice determines how the world is. Similarly, whether a sentence is true depends on what the words mean but this does not imply linguistic idealism according to which linguistic entities make the world thus and so.
The thought is that fully understanding the value and place of ordinary language concepts, and thus how to close them aptly, involves seeing how they serve our ends. That said, it is not totally a matter of our ends. As the witch case (§5.3) demonstrates, sometimes the world makes it quite clear how to close our terms. I note, though, that in the interesting cases which raise the question of the two images (e.g. ‘solid’) the world only decides so much. Beyond this our ends become key in determining the correct closure – i.e. in deciding whether to be compatibilists or error theorists.
6.2 A humanistic project
This focus on us and our ends is very important. It gives the exercise of closing our terms a very interesting dual nature. In closing our terms we are both betraying our conceptual perspective on reality and forming it. We are reflecting on what is so, whilst simultaneously deciding to develop our conceptual repertoire in one direction or another. This involves self-reflection and yields self-knowledge. By understanding how the Rutherford-Bohr model of the atom and our ordinary notion of solidity relate we don’t just learn about reality, we learn about our concepts, about how we cut up the world, about our ends – in short, we learn about ourselves.
Thinking in terms of open texture foregrounds the questions: why we are inquiring, what makes for better or worse inquiry (and why), and, indeed, who constitutes the we that inquire. Scientism and conservatism fail to do this.Footnote 64 Their failure to recognise that we have different uses for scientific and ordinary language concepts and representations is a failure to reflect upon the complexity of our own point of view. Earlier (§2) I argued that scientism involves a loss of knowledge, communicative ability, and possible ways of interacting with reality. We can add to that list that it involves a loss of self-knowledge. As Adrian Moore (Reference Moore and Moore2023: 133) writes:
[if] metaphysical deliberation[s] is to be of value, then it needs to be of a very different sort from the reflection exercised in physics … it needs to have a significant element of self-consciousness, making the identity of the “we” who are engaged in reflecting an unavoidable focus of that reflection. We need to see metaphysics as an enterprise that is completely different from any of the natural sciences, an enterprise that, unlike them, not only betrays “our” point of view but allows for exploration and refinement and creative development of that point of view.Footnote 65
Moore is endorsing Williams’ conception of philosophy as a humanistic discipline (or the humanistic project as I’ll call it). A distinctive feature of the humanistic project is that it recasts human self-understanding as central to philosophical inquiry. This is not because the subject matter of philosophy is human behaviour/thought, as it is for say anthropology, sociology, or psychology. Rather, philosophy involves understanding the world in such a way that we recognise that understanding as our own, as from our point of view. We come to self-understanding in that we learn about our point of view.Footnote 66 In Moore’s (Reference Moore, D’Oro and Overgaard2017: 45) words, it’s philosophy considered as ‘an attempt, by humans, from their unique position in the world, to make sense both of themselves and of that position’.
The humanistic project is the best way to account for the sense in which closing a term betrays and develops our outlook. This is because understanding and responding to open texture requires reflecting on the ends to which we put our representations and recognising those ends as our ends. The two images we are trying to relate are our images. By recentring the subject, the humanistic project captures this in a way neither Eddington nor Stebbing fully manage to.Footnote 67
Adopting the humanistic project is therefore crucial to my proposed approach to the question of the two images. Note, this doesn’t make the question a question about us; the correct answer is not an answer that describes us. Rather, it makes self-understanding a corollary of successfully answering the question. The very inquiry betrays our point of view in such a way that we recognise it as our own. Scientism and conservatism do not and as such they involve a loss of self-knowledge.
This application of the humanistic project provokes a further question: who are we? The answer depends on the ends/concepts under discussion. Perhaps for some concepts it is all rational thinkers, or at least all humans.Footnote 68 For some, as noted, it’ll be a particular socio-cultural historical group. For others it may be a smaller group. There is little general to say about this other than what Williams (Reference Williams1993: 171, n. 7) does in Shame and Necessity:
It is not a matter of “I” telling “you” what I and others think, but of my asking you to consider to what extent you and I think some things and perhaps need to think others
We are required to reflect on the analysis of a given end/concept and to ask whether we ourselves fit into the group of those who have that end/use that concept. What I have in mind, then, is what Williams calls the ‘we of invitation’.
6.3 Expanding the humanistic project
I mentioned in §2 that it is an underappreciated feature of Williams’ picture that nearly all our ordinary language concepts are local. In Williams, and associated secondary literature, discussions of conceptual localness are predominantly kept to thick ethical concepts.Footnote 69 Thick ethical concepts frequently diverge between different communities, and they’re ripe for ameliorative analysis. They are therefore obvious examples for illustrating the point that a group’s ends partially determine the aptness of their conceptual choices.
By emphasising the role our ends play in answering the question of the two images, I’ve shown that Williams’ project extends far beyond thick ethical concepts. In closing the open texture of our terms (whether in response to scientific advances or for some other reason) we, as linguistic communities, participate in unconscious, implicit, ameliorative work. Much ordinary language is ripe for ameliorative analysis.Footnote 70 The failure of authors like Stebbing or Eddington to ask what we could/should/want to do with a term is a failure to see this.
This is fitting since the boundary between thick ethical concepts and our descriptive local concepts is dynamic and not always clear. Recent normatively-guided ameliorative conceptual analysis on social concepts has illustrated this.Footnote 71 For example, Haslanger’s work showed that terms like ‘woman’, which were not obviously evaluative, may well be (or ought to become) thick ethical concepts.Footnote 72
The extent to which our ends, and our lives, inflect the standards with which we evaluate our everyday concepts is underestimated even by those familiar with Williams’ picture. All local concepts are ripe for the kind of ameliorative analysis that both betrays and refines our point of view. My reframing of the question of the two images demonstrates how this applies to many of our descriptive concepts. As such, it demonstrates that the humanistic project applies more widely than is usually thought. Closing open-textured terms involves engaging in the humanistic project. Given the prevalence of open texture, the project applies very widely indeed.Footnote 73
7. Conclusion
It’s not always clear whether a new scientific advance proves our ordinary ways of talking to be misguided or not. We ought to heed Williams’ warning against scientism whilst also being cautious to avoid conservatism. I’ve argued that we can do this by reframing the question in terms of open texture.
Approaching the question this way requires considering the ends to which we put our ordinary language representations. It requires recognising them as our ends and accepting that the fact that they are so partially justifies how we choose to close our open-textured terms.
This amounts to an exercise in philosophy conceived as a humanistic discipline. Though discussions of Williams’ humanistic project are frequently limited to our use of thick ethical concepts, my proposal demonstrates how the humanistic project extends much further than this. Recent conceptual engineering projects have expanded our sense of which concepts are ripe for ameliorative analysis. In a similar spirit, my proposal demonstrates how widely Williams’ humanistic project applies.Footnote 74
Open access
