Berkeley’s central arguments for immaterialism

Introduction

George Berkeley (1685-1753) is notorious for making the startling claim that minds were the only ‘substances’, i.e. the only things capable of independent existence. This position is called ‘immaterialism’, because it rejects material substance (which is by definition non-mental; following Berkeley, I will call any non-mental substance ‘matter’, although causal laws might count as material substances by this definition, and post-Berkeleyan science has posited substances unlike either mind or matter, ordinarily-conceived).

Immaterialism involves denying that things other than minds – houses, mountains, rivers and the like – can exist ‘mind-independently’, i.e. independently of minds existing (though it does not involve denying that these things exist at all – Berkeley has an account of their existence, though I shall not explore it in this essay). As Berkeley himself admitted, the contrary view “prevail[s] amongst men”; though he thought his philosophical system ultimately preserved more of common sense than any other, he was consciously arguing against ordinary assumptions on this topic. In this essay, I will explore his arguments for immaterialism specifically (leaving aside the other elements of Berkeleyan idealism).

Berkeley offers many such arguments, and selecting any as ‘central’ is somewhat arbitrary, but I shall focus on the sections of the Principles of Human Knowledge which he explicitly deems sufficient to establish immaterialism – P1-6 and P22-23 – and his defences elsewher of the premises and assumptions employed in these. While Berkeley’s other arguments purport to show only that there’s no reason to believe in matter, or that there’s better reason to accept immaterialism (because it has superior theoretical virtues, like ontological economy, preserving more deeply-held ‘common sense’ beliefs, and solving scientific, philosophical, mathematical and theological problems), the arguments in P1-6 and P22-23 promise to show more. P3 claims that “the absolute existence of unthinking things without any relation to their being perceived … seems perfectly unintelligible”; P22-23 are supposed to demonstrate that “the absolute [i.e. substantial] existence of unthinking things are words without a meaning, or which include a contradiction”. If materialism (here meaning belief in material substance, not denial of mental substance) is contradictory, it must be false. If it’s unintelligible (or “without a meaning”), it’s unintelligible to say it’s true or false, which stops it being a position we can hold – even if this were only because our understandings are limited. So P1-6 and P22-23 promise to end the debate.

The Attack on Materialism’s Intelligibility

What is intelligibility?

Beginning with P1-6, I shall focus on Berkeley’s attack on materialism’s intelligibility, which I think has the best prospects for success. To understand this attack, we must understand what Berkeley did and didn’t require for intelligibility.

What he did not require was that matter be intelligible in a specific way. He thought it unintelligible in every way (perhaps this is what “perfectly unintelligible” meant), running through all the possibilities he could think of: what I shall later call imagistic, relative and negative conceivability. Only by stopping at P6 could one think he required matter to be conceivable imagistically.

What he did require was that we be aware of intelligible claims’ meaning: in modern jargon, intelligibility was an ‘internalist’ notion, dependent only on those properties of our representations of which we’re aware (their ‘narrow content’). This can be seen from his choice of ‘internalist’ terms like “conceivable” and “intelligible”, and the criticisms he levels at materialism’s intelligibility: that we do not understand the relation involved in our concept of substratum, that if materialists accept that they don’t know matter’s nature materialism will be empty, and so on. Semantic externalists could count the term ‘matter’ as meaningful even if we lacked any understanding of matter’s nature, but Berkeley evidently required more. This was defensible, because he needed only to show that materialism could not be believed, and it seems highly implausible that one can believe what one does not understand (though some externalists maintain this). Some externalist theories would deem ancient aborigines to have believed that the planet Venus existed, because their accepting of the coming of the goddess ‘Barnumbirr’ was responsive to Venus’ appearance, but having this belief seems to require grasping the concept of planethood. Responses to Berkeley which tacitly or explicitly presuppose semantic externalism thus do nothing to show materialism’s intelligibility.

Another thing Berkeley may, like Locke and Hume, have presupposed was that we lack ‘intellectual’ concepts not derived from our experiences, or direct awareness of, e.g., our minds. Since I am unaware of any good reasons for this requirement, and sceptical of our ability to account for all thought and representation with it in place, I shall not presuppose it while evaluating Berkeley, and suggest that we posses a number of ‘intellectual’ concepts.

P1-P3

In An Essay Concerning Human Understanding, published 21 years earlier than Berkeley’s Principles, Locke had influentially claimed that the “objects of the understanding” (for which he used the blanket term ‘ideas’; I’ll use ‘concept’, as Berkeley generally uses ‘idea’ in a narrower sense) either came from sensation or reflective awareness of our mental activity, or were formed from such ideas by processes like abstraction. P1 deliberately echoes this claim, though Berkeley declares it to be “evident” rather than founded on Locke’s painstaking arguments, includes only “memory and imagination” and not abstraction in his list of concept-forming processes, and adds in P2 that we have a concept of our mind (later explaining that this comes through direct awareness, and affords us a concept of other minds). He also talks of “objects of human knowledge” rather than “objects of the understanding”, but that he meant his list to exhaust the latter as well is suggested by his echoing of Locke, his assumption that a readership steeped in Locke would need no argument for his claims, the references to intelligibility which follow, and P83’s claim that “the proper use of words is the marking out [of] … things only as they are known and perceived by us”. As potential objects of awareness, everything P1-2 lists is an internalistically acceptable object of the understanding.

Every concept P1-2 lists is also, P3 declares, of something “evident[ly]” dependent on a mind, and so not a material substance. Is this right? Berkeleyan imagination and memory can supply only concepts which sensation or reflection could supply, and Berkeley was right that “everybody will allow” the mental activities (like “thoughts” and “passions”) of which we are reflectively aware to be mind-dependent. So, if P1-2’s list of our concepts is indeed exhaustive, sensation is the materialist’s only hope.

Now, Locke thought that some of the concepts we received through sensation were of potentially mind-independent qualities [though maybe he thought these had to be derived from sensory experience by abstraction], which could exist in matter. If this were so, Berkeley’s game would be up: matter would have an intelligible nature involving qualities concepts of which we got through our senses. Berkeley was obviously aware of Locke’s position. What was his response? He begins by declaring it “evident” that “ideas imprinted on the sense” are mind-dependent, which is true but irrelevant, since it doesn’t show that they’re ideas of necessarily mind-dependent qualities. Here, Berkeley may have fallen into Locke’s trap of not properly distinguishing ideas and their objects. Fortunately, he soon corrects his error. He claims that the existence of “sensible things” (sensations’ objects) is mind-dependent, and that the unperceived existence of sensible qualities (like “odour” and “figure”) cannot be understood. This would mean that sensation conferred no understanding of qualities which could exist in matter and let us describe it.

But is this response correct? Finding it “intuitive”, Berkeley does not back it up, but we can appreciate how sensible qualities could be found intuitively mind-dependent by considering pain: unfelt pains obviously cannot exist. Though Berkeley thought the same held of all sensible qualities and we may initially not, he can leverage our intuitions about qualities like pain, for when experiencing one quality necessarily involves experiencing another the intuitive mind-dependence of the second would show the mind-dependence of the first (since Berkeley allows no method of conceptually separating what is inseparable in sensation and imagination – remember that we are working within P1’s constraints for now). Robinson calls this leveraging manoeuvre ‘the assimilation argument’, and traces it back to Aristotle. Berkeley applies assimilation arguments throughout 1D and in P10, noting for instance that experiencing extreme temperatures involves experiencing pain (plausibly adding that if extreme temperatures are thus mind-dependent so are mild temperatures, an argument independent of his contentious claim that sensing mild temperatures involves pleasure).

Assimilation arguments are not all that Berkeley has up his sleeve. P10 points out that colour is mind-dependent on the best materialist theory, so if anything’s mind-independent it’s not colour. Similar reductios can be targeted against all phenomenal qualities except shape, including temperature. An assimilation argument then gets us to the mind-dependence of visual and tactile shape, the toughest cases for Berkeley and the supposed sources of Locke’s concept of matter: seeing shapes involves experiencing colours, and feeling shapes involves feeling temperatures (Berkeley did not so far as I know make this last assimilation in the Dialogues or Principles, but it is implicit in N288a). This whole chain of thought is entirely independent of intuitions, though Berkeley did not avail himself of it.

When, in P8, Berkeley explicitly attacks the Lockean suggestion that sensible quality-instances resemble mind-independent quality-instances by saying “an idea can be like nothing but an idea”, he presents this ‘Likeness Principle’ as intuitive: “If we look but ever so little into our thoughts, we shall find [this]”. I suggest that he based this on the supposedly intuitive mind-dependence of individual sensible qualities. Just as a pain’s nature is essentially experiential, and so the only thing with a like nature can be an experience, P8 echoes P3 in suggesting that other sensible qualities have essentially experiential natures: it denies that it makes sense “to assert a colour is like something which is invisible”, i.e. something which is not an experience, and this must be because our idea of colour is just an idea of a sort of experience.

But even if our intuitions ultimately differ from Berkeley’s here, P10-style reductios will imply that we have no concept of matter if P1-2’s limited, radically empiricist list of our concepts is exhaustive. I shall examine whether it is later, but first show how our exposition of P8 throws light on P4’s discussion of perception.

P4

P4 argues against the common view that “sensible objects” like houses, mountains and rivers are mind-independent. For all it says, something else may be mind-independent, but sensible objects are the obvious candidates. The argument is short enough to quote in full: “what are [sensible] objects but the things we perceive by sense, and what do we perceive besides our own ideas or sensations; and is it not plainly repugnant that any one of these or any combination of them should exist unperceived?” In other words:

(1) Sensible objects are perceived by sense.

2. Only sensations are perceived by sense.

∴ Sensible objects are sensations or collections thereof.

(3) Sensations are mind-dependent (they can’t exist unperceived).

∴ Sensible objects are mind-dependent.

Berkeley implies that the premises are obvious, but the first is obvious only if indirect perception (defined here, as in 1D174, only as perception dependent on perceiving something not even partly identical with what is indirectly perceived – not e.g. as requiring an inference to justify belief in the object of perception) is allowed, and the second couldn’t be obvious if it were. So Berkeley is often accused of equivocating. But P8’s Likeness Principle would rule out indirect perception of matter by sense if only sensations are directly perceived and if sensations can only let us perceive objects they represent, because the Likeness Principle stops sensations from resembling (and thus, for Berkeley, representing) anything but mind-dependent qualities. So the argument can be construed thus:

(1) Sensible objects are (directly or indirectly) perceived by sense.

(2) Only sensations are directly perceived by sense.

(3) Only mind-dependent qualities can be indirectly perceived via sensations. [P8]

∴ Sensible objects are sensations/mind-dependent qualities or collections thereof.

(4) Sensations are mind-dependent (they can’t exist unperceived).

∴ Sensible objects are mind-dependent.

This argument is valid. Premises 1 and 4 are obvious, and I accept premise 2 because I see no way of perceiving an object by sense except through having a certain sort of sensation. So everything depends on premise 3. Since sensations can only represent what qualitatively resembles them (because I don’t see how sensations could by themselves represent objects as having qualities not involved in those sensations), and since we are assuming the Likeness Principle’s intuitiveness, to question premise 3 we must question whether sensations can only let us perceive objects they represent.

To do so, we must propose an alternative way in which sensations can let us perceive objects. One possible way involves either the inference of those objects from those sensations or what Berkeley calls ‘suggestion’ (an unconscious, automatic association of objects with sensations). Berkeley rather contentiously claims that this is not perception, properly so-called (even though people often call it just this), but even if we disagree with him here, his arguments that we lack a concept of matter would preclude its existence from being inferred by us or suggested to us. The only other ways I can devise involve causation: one is the claim that we perceive whatever is appropriately causally connected to our sensations; the other is the claim that that we perceive whatever we know is thus-connected. The second claim lets us perceive matter only if we can understand it (which Berkeley denies), and the implausibility of saying that someone could perceive something without understanding anything about it undermines the first claim (which allows for this possibility), but such judgements of implausibility are neither universal nor very reliable. Both claims fail if Berkeley is right that only minds can cause anything, but the most Berkeley’s arguments for this show is that we can only understand mental causation, which (because it does not rule out the existence of non-mental causation) would only tell against the second claim. So the claim that we perceive certain causes of our sensations has yet to be ruled out, making P4’s argument fail.

Abstract concepts of matter

I said that if P1-2’s list of our concepts was exhaustive, then (granting that reflection confers no concepts of mind-independent things) arguments for the mind-dependence of sensations’ qualities (not just their quality-instances!) prove materialism’s inconceivability. As promised, I now turn to examining whether P1-2’s list is exhaustive, and whether we have some other concepts which would let us conceive of matter. I begin with Berkeley’s defence in the Principles’ introduction of his exclusion from the list of concepts formed by abstraction, on which his assimilation arguments for the mind-dependence of sensations’ qualities depend.

Abstraction, for Locke, was a mental operation in which we stripped away certain aspects of a particular idea, leaving behind a general idea which by virtue of its loss of these particularising aspects can fit many different things. Where remembered ideas were copies of ideas of sensation, abstract ideas were partial copies. So forming the abstract idea of a triangle involved stripping away the particular aspects of individual triangles to leave only the essential properties we require to count something as a triangle at all. This would leave an abstract idea of a triangle “neither oblique, nor rectangle... but all and none of these at once”, and so lacking qualities like determinate shape which an experience couldn’t lack: we could never see a triangle of indeterminate shape. Berkeley thought introspection showed that we lacked such ideas, as, unlike remembered or imagined ideas, they would not describe possible experiences. Remembering that Berkeleyan ideas are ‘imagistic’ (experience-like states, as copies or partial copies of sensory ideas must be), I think he is right; I cannot form an image I could not experience.

If your introspection goes differently, and you think such abstraction possible, assimilation arguments will not work, so unless you find shape’s mind-dependence directly intuitive, you will have no reason to accept it. So Berkeley’s denial of abstraction is really vital to his argument, despite his exclusion of it from the Dialogues (which are filled with assimilation arguments).

However, it is natural to suggest that even if Berkeley has shown that we cannot form abstract images of (say) colourless shapes we must have a non-imagistic abstract concept of having a particular shape. This is because we seem to be able to abstract a particular shape we see from the colours it has: we can, after all, tell when a red and a blue image share this shape. And also because, as Berkeley himself allows, we can tell that some features of a particular triangle (e.g. obeying Pythagoras’ theorem) derive merely from its shape, rather than its colour. Berkeley’s own account of general terms – which is that they stand for all the images which fall under them, even when we just call one to mind as representative – cannot account for these abilities, because it explains neither how we decide which images fall under ‘triangle’ nor how we see that all these images will share a feature without separately understanding the property (triangularity) which entails this feature.

Explaining the nature and workings of our non-imagistic abstract concepts is extremely difficult, making it no wonder that Locke and Berkeley opted for a simple imagistic model of thought. But just knowing that we have such concepts of shapes suffices to refute Berkeley’s assimilation argument that we have no concept of shape which could apply to matter (his only good argument for this conclusion). This is not, of course, to explain how our concept of shape can apply to matter. However, in the case of formalisations of geometry like Hilbert’s, we can see how: material reality need only contain a model for the formal system (as do both our visual and tactile field – this provides the best way of explaining what N56 observes: “Mathematical propositions about extension & motion [are] true in a double sense”). This foreshadows the sort of structural representation of the world I advocate later. (Obviously, our everyday ability to ‘intuitively’ work out the properties of types of shapes does not involve formalisations; I confess I do not know what it involves.)

Negative concepts of matter

P80 considers a “negative definition of matter”, which Berkeley anticipates will be the materialist’s last resort because it evades the problems he raises for all the positive definitions he can think of: avoiding commitments is the surest way of avoiding criticisms. This negative definition does not simply refuse to ascribe any positive qualities to matter; it denies that matter has several qualities, deeming it “neither substance nor accident, spirit nor idea, inert, thoughtless, indivisible, immoveable, unextended, existing in no place”. So it is not ‘empty’ in the sense of failing to establish what it would take to count as material. However, Berkeley clearly considered it empty in some sense, saying that it produced no “effect or impression” on our minds “different from what is excited by the term nothing”. This is true if Berkeley is referring to (imagistic) ideas, as the word “impression” suggests he is, because this negative definition of matter would then simply be an extreme case of abstraction in which all imagistic properties (which we are taking as intuitively mind-dependent) were stripped away, leaving behind not an impossibly indeterminate image but no image at all.

If we suggest that a negative definition of matter provides a non-imagistic concept of it, we shall have to explain what sort of concept this is. It cannot involve qualities left over once we exclude the mind-dependent ones unless we independently understand mind-independent qualities, since merely excluding the former can’t confer understanding of the latter. The only concepts which negative definitions can provide which we wouldn’t have independently of them are “relative” concepts based on the relation of ‘not resembling’, a relation we understand through our acquaintance with resembling and non-resembling ideas. P80 denies that negative definitions can provide such relative concepts, but I cannot see why, because though such a concept is slight it is not empty: “X doesn’t resemble Y” and “X doesn’t resemble Z” would be synonymous were they both empty, but they’re not – indeed, one can be true without the other’s being true. Nonetheless, a purely negative definition of matter is slight; to see if we can get something more, I shall examine relative concepts more fully.

Relative concepts of matter

In P16, Berkeley considers a way of forming a concept of matter which does not involve ascribing any qualities to it (and so is unthreatened by his denial that we have concepts of mind-independent qualities). This is to characterise it as standing in a certain relation to something. Such a characterisation (which, following P80, I shall call ‘relative’) relies, he notes, on our having independent concepts of both the relation and the something in question: to define ‘matter’ as what ‘gyre-and-gimbles’ sensations grants no understanding of it, and neither does defining it as what perceives ‘slithy toves’. The suggestion he considers in P16 (that matter “supports” extension) fails this test: we lack an appropriate notion of support here, because the usual one applies only to extended things supporting other extending things, and P15 has just reaffirmed that extension is mind-dependent. (P16 ascribes this suggestion to “the most accurate [materialist] philosophers”; some interpret Locke as holding it, despite his expressions of scepticism about our understanding of this notion of “substratum”, and maybe Berkeley did too.)

What about other relative conceptions of matter? The obvious suggestion (made by Hylas in 2D216) is that matter is what causes our sensations. Berkeley rejects this because he thinks he has no concept of causation besides volition, and mindless matter can’t have volitions. But many philosophers in the Humean tradition conceive of causation as constant correlation between certain types of situation (NB: not just between anything), and even Berkeley must allow that we have a concept of constant correlation. Indeed, N461 says that our concept of volitional causation only amounts to constant correlation between volitions and their intended effects (which seems to be right): “Wn I ask whether A can move B, if A be an intelligent thing, I mean no more than whether the volition of A that B move be attended with the motion of B, if A be senseless whether the impulse of A against B be follow’d by ye motion of B”. This would allow for a parallel concept of material causation if we understood material motion, rather than just motion in sensory fields. Of course, Berkeley would deny that we do understand material motion, and to fit matter into such a causal story we must already have a way of characterising different material situations (like A impacting B, and perhaps the law which makes B then move – such laws could be ‘material’ in the relevant sense of being mind-independent). This would by itself give us a concept of matter as possessing certain determinable features such as spatiotemporal locations (though there might turn out to be different ‘types’ of matter, with different determinable features – e.g. one type might have spatiotemporal locations, and the other might not). Once we had this concept, we would already have showed materialism’s intelligibility; it would only be an added bonus that we could take account of our belief in matter’s causal efficacy by fitting it into a ‘Humean’ causal story.

I think that a structural representation of the world can provide the requisite characterisation of different material situations through the structural positions of their components. This characterisation also offers us an understanding of causation as a particular natural relation or group of such relations if we favour this over a Humean account, as these relations could be characterised through their structural roles. (Although, as with constant correlations between appropriately similar/contiguous/etc. situation-types, there will be nothing to mark these relations out as causal besides their holding between certain situation-types and not others. I am not sure to what extent this is a problem, partly because lacking any direct understanding of causation I do not see how it can be avoided.)

Structural role concepts of matter

Structural representations are not an option that Berkeley considered, and perhaps for that reason seem to be immune to his arguments, which target the possibility of understanding potentially mind-independent qualities: the advantage of structural representations here is that they need not specify the qualities of elements in the structure, but nonetheless characterise these elements by specifying their relations to other elements in the structure. I shall call such a characterisation a ‘structural role concept’. It is similar to a relative concept, except that it does not require that the elements related to the element being characterised are themselves understood independently of their relations to other elements. So all the elements in a structure could be characterised by their structural roles, as in a formal system.

We will need a prior understanding of some of the relations in such a system, as a relation X could only be characterised through its structural role of holding between elements Y and Z (and presumably many more) if we already understood the ‘holding between’ relation. It seems as though we can understand this relation through our understanding of particular relations, of which Berkeley allows that we have “notions” and which can relate ideas to minds (which stops the attack on abstract ideas from applying, because our concepts of relations involving minds can’t be imagistic). Once we have this, most of the relations a structural representation of the material world needs (e.g. instantiation, or various quality-specific resemblance relations) can be characterised through their structural role, the exception being constant correlation (needed if we give a ‘Humean’ account of causation), facts about which simply supervene on other facts (depending on our metaphysics – i.e. on the specific sort of structure we ascribe to reality – these might be facts about which spacetime points instantiate certain properties).

(I am relying on the claim – plausible for many reasons – that the world contains universals, a special class of properties and relations which alone fill certain structural roles in the structure we will end up ascribing to [a substructure of] the world. Were all properties and relations – whichever set of elements/ordered n-tuples of elements was their extension – ontologically equal, any structural representation of the world would be vacuously true – and so vacuous – provided that the world contained enough elements.)

If we add experiences (which of course can be characterised by qualities independent of their structural role) to the structure, linking them to the rest of it through constant correlations with certain mind-independent situations (psychophysical laws perhaps among them), I think we will have characterised material reality as well as we could hope to. This is a double-edged conclusion: Berkeley’s challenge to conceive of matter has been met, but only with a fairly attenuated conception, which I think acceptable but others may not. (Perhaps because reality might have many substructures which fit our structural representation, which would make the reference of some of this representation’s primitive terms indeterminate – indeed, a Putnam’s model-theoretic argument suggests that this will definitely be the case if it can be modified to allow for the interpretation of terms describing experiences being fixed, constraining the interpretation of all other terms. But, like Putnam, I am not sure that this would be unacceptable.)

The Master argument

In P22, Berkeley says that his views can be “demonstrated with the utmost evidence in a line or two”; the lines which follow make the now-familiar claim that we cannot conceive of something existing unperceived, and that this rules out “the bare possibility of [materialism’s] being true”. I have explored some of Berkeley’s reasons for thinking this, but P23 gives another justification, which has come to be called the ‘Master Argument’. I shall examine three interpretations of this: Peacocke’s, Pitcher’s and Prior’s.

Peacocke sees P23 as a continuation of earlier arguments, particularly P5’s claim that to “conceive [sensible objects] existing unperceived” would involve an impossible sort of abstraction. He thinks that Berkeley’s point in both passages is that conception involves imagistic ideas, that these involve a perspective which cannot be abstracted away, and that a perspective presupposes a perceiver. This argument faces problems: (1) it does not by itself show that the perceiver must be more than merely possible (and Berkeley’s other arguments which might do so would render it redundant, rather than supplementing it); and (2) we have seen that Berkeley accepted non-imagistic conception. Anyway, regardless of its merits, this argument does not seem to have been what Berkeley had in mind: (1) neither P23 nor P5 refers (even implicitly) to perspective; (2) when the Principles do discuss perspective, only arguments that we could not know material objects’ precise properties are mentioned – had Berkeley thought of Peacocke’s stronger argument, we would expect it to be mentioned here; (3) it ignores the problem P23 raises about anything we conceive of as unconceived being conceived of by us, which is plainly meant to support Berkeley’s conclusion.

Pitcher (following Russell) offers one way of understanding this problem: he reads Berkeley as conflating representations (what we conceive with) and what they represent (what we conceive of with them – the representations’ content), and assuming that all features of the former must be features of the latter.dOn this reading, P23’s argument is that because the representation is conceived of, it must represent its object as being conceived of. The problems with this, both as an argument and as an interpretation, are: (1) that it implies that I cannot conceive of anything as not conceived by me, now, committing us to a conceptual solipsism of the present moment which is counter-common-sensical and inconsistent with Berkeley’s system; (2) that though Berkeley’s internalism commits him to the view that only representations’ (transparent) qualities determine their content, this does not commit him to the unmotivated view that all of those qualities do so (since there are viable ways of distinguishing among these qualities – only some will be intended to determine the representation’s content, for instance); and (3) that, had Berkeley held this view, he would have made it explicit in one of his many discussions of philosophy of language in Alciphron and other works. Johnston and Hicks try to defuse (1) as an interpretive problem by claiming that Berkeley’s early notebooks show he was once a conceptual solipsist, and that P23 was a holdover from this period, but Berkeley would not have given an argument obviously inconsistent with the Principles’ system pride of place.

Prior offers another uncharitable interpretation of the problem P23 raises: he thinks that the “contradiction” Berkeley sees lies in someone thinking truly of some specific thing that it is not thought of, and that Berkeley did not realise that this contradiction can be avoided if the person only thinks that some unspecific thing is not thought of. The contradictory claim can be formalised as ∃x∃y(xC*¬Cy), and the uncontradictory one as <∃x(xC*∃y¬Cy), , where Cy means that y (a thing) is thought of, and xC*z means that x (a person) accurately thinks z. z is a proposition (∃y¬Cy) in the uncontradictory claim – which thus involves de dicto thought – whereas in the contradictory claim z just involves de re thought – ascribing a property to a specific thing. This means that that thing is thought of, making ∃x∃y(xC*¬Cy) contradictory because it implies ∃x∃y(xC*¬Cy ∧ Cy ∧ ¬Cy): xC*¬Cy implies Cy, because y is thought of, and ¬Cy, because ¬Cy is thought accurately.

This interpretation sticks close to P23, only ascribing belief in the obvious implications just mentioned to Berkeley. It fits his mention of “contradiction” in P22, though he sometimes called claims he thought inconsistent only with other truths (rather than internally inconsistent) “contradictions”, and in P24 allows that he may just have proved materialism to be “without a meaning” rather than contradictory. And Berkeley’s language in P23 is de re (the problem is supposed to be that we “think of” “certain” things, which we do not do when we think that ∃y¬Cy), and it is not implausibly uncharitable to claim he missed the subtle distinction between de re and de dicto thought, which typically gets lost in English. A greater problem for Prior’s interpretation is that ∃x∃y(xC*¬Cy)’s contradictoriness shows only that we cannot accurately think of some y as unthought of, whereas P22 denied we could not “conceive” of unconceived things at all. And this argument also falls short of proving that y’s existence depends on its being thought of, rather than being constantly correlated with this, although here it does not fall short of P22’s promises. But lacking a better interpretation, I tentatively suggest that Berkeley missed this shortcoming of his argument, since P23 moves so fast, and also note that by the end of P22, Berkeley’s challenge to materialists had become to show “the bare possibility of [their] opinion’s being true”.

Besides Prior’s objection to the argument he ascribes to Berkeley, there is another one: there is no contradiction if ¬Cy means only that y is unthought of in the past or future or in counterfactual situtations, rather than in the actual present. This lets a materialist truly say of y that it was unconceived before she just conceived of it (enough for materialism’s truth, unless we’re presentists), or that it would have been unconceived had she not done so. Most materialists would admittedly want to say more than this, but it makes P23 look like an inconsequential trick, as it allows that there might be many unconceived things which only cease to be unconceived when we turn our thoughts to them; you could equally note that you cannot truly think that you’re not thinking about anything, but this would not prove that you think necessarily or at all times.

So we have been unable to find a good argument in P23 against the possibility of materialist beliefs being true, let alone a proof of the impossibility of our having such beliefs (which P22 initially promises). (Winkler ascribes such a proof to Berkeley, but it cannot be found in P23, and could not have been in Berkeley’s mind when he wrote this as it would render P23 redundant. Winkler mentions only P22 in his ascription, but P22 simply claims that we will find ourselves unable to conceive of matter; perhaps, as Winkler suggests, Berkeley thought this was because we would find ourselves lacking a concept of unconceivedness – for reasons already discussed and dismissed – but this can only be a guess.)

Conclusion

The ambitious promises of P1-6 and P22-23 fail. These were the only arguments which promised to prove that immaterialism is false; so Berkeley has no such arguments. But while the Master Argument’s failure taught us little, the realisation that materialism must involve concepts other than those which P1-2 lists revealed an important problem, and the search for such concepts illuminates many issues within the philosophy of language, and reveals the extent to which our concept of matter must be an attenuated, ‘intellectual’ one which falls short of the sort of full qualitative characterisation Berkeley could give of the entities involved in his metaphysical system.