Defending an External World to Show What It Takes and Use It Elsewhere


The chief advantage of defending our belief in an external world instead of relying on pragmatic faith is that, besides avoiding the need for faith, it shows that we must accept eight methodological principles. This has widespread philosophical advantages since a number of them are often rejected in other areas. A further advantage of accepting them is that they can also serve for the defense of other minds. These principles include, among others, the following:
  • We must accept judgments about our sensory experiences. They are intrinsically credible because we have direct access to their sensory referents, but they’re fallible and may be challenged by judgments about physical objects supported by other sensory experiences.
  • We must accept concepts though we are only directly aware of their conscious components. We must accept logic to justify logical inferences. Logical truths are intrinsically credible because they are self-evident: the same holds for objective synthetic a priori truths. Since analytic truths are only based on linguistic conventions, we must accept the objective synthetic a priori as well logic to justify mathematics. 
  • To avoid phenomenalism we must accept physically necessary causation because constant conjunction would be happenstance without it. Here as elsewhere we must accept indirect confirmation. 
  • We must accept subjective probability judgments and cope with the extent to which they are vague. 
  • We must also accept memory even though no memory is verifiable or even indirectly confirmable by other memories. It follows from the above that Descartes’ requirement for certainty cannot be fulfilled.

It may seem that since we all believe in an external world, we shouldn’t bother with defending it. We should instead, like pragmatists, devote our attention to the more interesting and controversial issues of Quine’s naturalized epistemology. However, the point in defending it isn’t to show that there is an external world. It is to show that we must accept whatever epistemic principles it requires and, which often comes to the same thing, reject those that are incompatible with it. Besides accepting the fallibility of sensory reports we must abandon various popular epistemic restrictions that have widespread effects in other areas including Quine’s naturalized epistemology. These restrictions include, among others, the rejection of indirect confirmation, the more common acceptance in some areas but not others, Descartes’ requirement for certainty, subjective probability judgments, concepts, memories of our experiences, and therefore (as by Wittgenstein) even statements describing them.
Furthermore, Richard Rorty has used such methodological restrictions as reflected in the works of Wittgenstein, Quine, and Sellars in ways that make it difficult at best to oppose post modernism, including the non-representational theory of truth. This is particularly important since, unlike most philosophical doctrines, post modernism has had a profound effect on the humanities in general. Bypassing the egocentric predicament would also force us, like Hume, Santayana, and now pragmatists, to rely on faith in an external world and in doing so set a precedent for other kinds of faith including the religious faith that Hume vigorously rejected.
While a majority of philosophers are unwilling to accept faith in religious beliefs, others, like William James, have felt free to accept them because they only require beliefs to be useful. However, a second requirement is for beliefs to be plausible, like belief in an external world but not for most philosophers religious beliefs. And since current pragmatists use the lack of favorable evidence to reject religious beliefs, instead of being true pragmatists they rely implicitly on the favorable evidence that makes belief in an external world strike them as plausible. Thus I suggest that instead of trying to have it both ways they should abandon their pragmatism even if they prefer to leave the problem of defending an external world to others in order to focus on “naturalized epistemology.” Furthermore, Hume and Santayana faced the same dilemma in accepting faith in an external world but not religious faith.
I argue that we need to rely on four kinds of intrinsically credible beliefs, two empirical and two a priori, reject the requirement for certainty, ascribe at least some credibility to memory to avoid solipsism of the present moment, accept the indirect confirmation of causal relations since constant conjunction would be happenstance without them, and justify belief in a physical world on the grounds that it provides the best causal explanation for the involuntary arrangement of our sensory experiences. Save for reliance on memory even though no memories are verifiable, on the use of indirect confirmation, and rejection of the requirement for certainty, the kinds of judgment we need to treat as basic are in various degrees intrinsically credible, i.e., credible without support. If instead all judgments were only credible extrinsically, any attempt to justify a judgment would lead to a vicious regress.

Using The Foundation’s Components To Defend a Physical World
1.     Judgments about our sensory experiences are intrinsically credible because we have direct access to them—direct since the judgments and referents are included in the same experiences. Direct access doesn’t, however, preclude errors, and sensory judgments may be less credible than judgments about external objects. For instance, the judgment that I had a sensory image of eighty-seven books is less credible than the judgment that I was looking at between seventy and a hundred books[1]. However, it isn’t their degree of probability that makes them foundational. They are foundational because they are intrinsically credible while all statements about the external world require support.
2.     We must abandon the quest for certainty. Though C.I. Lewis (1946) held instead that sensory judgments must be absolutely certain to provide a foundation, their intrinsic credibility entitles us to rely on them in various degrees despite the fact that we can’t even ascribe precise probabilities to them. This is compatible with the use of mathematical calculations, since, as I show later, there are ways of taking intuitive probabilities into account in our overall probability assessments as we do in perceptual as well as sensory judgments.
3.     We must also accept concepts: among other reasons, we need them to make judgments about our sensory experiences and to interpret them in our judgments about physical objects. Though it would be difficult to explain our behavior without concepts, there is no need for an hypothesis since we are aware of them. There can be uninterpreted feelings and sensations, but intentionality is a readily detectable feature of most experiences. Sentience does not require intentionality, but consciousness does.
The claim that we have concepts is intrinsically credible. For instance, the experience of perceiving a shape as that of a bat differs from that of perceiving a similar shape in an abstract painting since you associate different concepts with the two images. This also shows that concepts involve more than an image. Nor could paradigm images replace concepts: any image has numerous qualities so you would need a concept to pick the relevant ones.
If we lacked concepts thinking would be limited to linguistic images and feelings with no cognitive significance. A number of obvious facts show that this is not the case:
                        i.         The experience of reading sentences you understand is markedly different from that of reading sentences that you could only pronounce, nor is the difference the associated images, since it still applies when, as is usual, images are not involved.
                       ii.         Like many others, I am aware of periods when I think without words or, like characters in Joyce’s Ulysses, use far too few words to express my thoughts.
                     iii.         We sometimes express the opposite thought from the one we have in mind and intend to express.
                     iv.         We may have to struggle to find words to express our thoughts, in which case our thoughts precede our words.
                       v.         We often settle for approximations.
                     vi.         We can’t always say what we think. Sometimes we don’t use a language’s resources. Sometimes it doesn’t provide them.
Concepts also involve dispositions. We lack direct access to dispositions, but we can learn about them by asking ourselves how we would respond to possible cases. Unlike Wittgenstein, I believe that the conscious part of my notion of the addition of one matches the associated disposition. Though I am almost certain that it does, I can confirm my belief by considering how I would respond to the addition of one to any number however high. Furthermore, Alan Goldman (1991) argues that we can ascertain the nature of dispositions associated with terms standing for salient qualities such as greenness as opposed to grueness.
4.     Though we seldom make judgments about the contents of our beliefs, we must when we need to defend beliefs’ logical or empirical relations to one another. Such judgments’ credibility is intrinsic insofar as we have direct access to their conscious parts, and we’ve seen that there are ways of determining the nature of the dispositional part.
5.     We must accept logic to justify logical inferences. Logical truths are intrinsically credible because they are self-evident.
6.     We must accept both logic and I believe the synthetic a priori to justify mathematics. I mean the objective synthetic a priori, like Edmund Husserl, Gottlob Frege, Bertrand Russell, Roderick Chisholm, Barry Smith, A.C. Ewing, Arthur Pap, and Laurence BonJour truths that are objective rather than, like Kant, truths that are mentally imposed in the phenomenal realm but do not apply to things in themselves[2]. Synthetic a priori truths are also self evident and therefore intrinsically credible.
The reason for my belief is that save for logical inferences, mathematical truths are synthetic a priori rather than analytic because analytic truths are only based on linguistic conventions. We need conventions to determine the meanings of the terms standing for numbers and such relations as addition, subtraction, and multiplication, but while we also have conventions for determining the results of adding, subtracting etc., the conventions must be in accord with objective synthetic a priori truths.
7.     Though we have to rely on memory both here and in general, memories of past experiences are neither self evidently true nor accompanied by the experiences to which they refer. Instead we must rely on present memories even though we can’t return to the past to verify them or even to use other memories to confirm them. This is often ignored. Wittgenstein (1953) deserves credit for recognizing the problem in arguing that we don’t describe our experiences because we would have to rely on unverifiable memories. However, he accepted memories of publicly observable events because he failed to see that the same objection applies to all memories, and once this is recognized it becomes obvious that we must ascribe some credibility to memory. As I’ve observed, the alternative is solipsism of the present moment.
8.     We must accept indirect confirmation. Among other reasons we need causal necessity to defend an external world and the existence of causal necessity can only be confirmed indirectly. We can, however, use a priori mathematics, to show that it is, to say the least, extremely improbable for the regularity of causal relations to be happenstance instead of physically necessary. The need to accept indirect confirmation has widespread implications for the rest of philosophy, which is plagued by problems resulting from its rejection.
Using the foundation’s eight components we can accept the concepts needed to make judgments about our sensory experiences. We can recognize that these judgments instead of being certain are in varying degrees intrinsically credible. We can use judgments about physical objects to confirm or question sensory reports. We can defend mathematical truths on the grounds that they are logically necessary or, for advocates of the synthetic a priori, metaphysically necessary. (Something is metaphysically necessary if it is a synthetic a priori truth that it is necessary.) We can use memory plus mathematical probability theory to determine that there is too much regularity in our sensory experiences for the causal relation to be happenstance instead of physically necessary[3]. Relying again on memory and mathematical probability theory we can see that there isn’t sufficient regularity for a causal explanation without adding physical objects[4]. We can make judgments about the contents of our beliefs as in defending beliefs’ logical or empirical relations to one another. And we can abandon the quest for certainty.

The Use of Foundational Beliefs Provides Further Credibility For Other Foundational Beliefs
One sort of foundational belief may be used to provide further credibility for another as in the case of empirical confirmations of both particular a priori beliefs and a priori knowledge in general. In addition a kind of belief may need the support provided by another kind of belief as in the need to accept concepts both in defending sensory reports and a priori judgments, and the need to accept mathematical probability judgments to defend physically necessary causal relations on the grounds that constant conjunction would be extraordinarily improbable without them. However, the need for this sort of support is consistent with treating a small group of basic beliefs as the foundation for belief in an external world.
The foundation’s eight components make the following kinds of reinforcement possible:
1.     We can accept the concepts needed to make judgments about our sensory experiences. This includes by indirect confirmation concepts’ dispositional components.
2.     By rejecting the quest for certainty, we can recognize that these judgments instead of being certain are in varying degrees intrinsically credible.
3.     We can use judgments about physical objects to confirm or question sensory reports.
4.     We can defend mathematical truths on the grounds that they are logically necessary or, for advocates of the synthetic a priori, metaphysically necessary.
5.     We can use memory plus mathematical probability theory to determine that there is too much regularity in our sensory experiences for the causal relation to be happenstance instead of physically necessary[5].
6.     Relying again on memory and mathematical probability theory we can see that there isn’t sufficient regularity for a causal explanation without adding physical objects[6].
7.     We can defend beliefs’ logical or empirical relations to one another by making judgments about their contents including by indirect confirmation concepts’ dispositional components.

Susan Haack (1993) distinguishes her foundherentism from foundationalism in a work for which I have considerable respect, arguing that foundationalism postulates a privileged class of basic beliefs by requiring that relations of support be essentially one directional. However, instead of being a foundationalist in this sense, I have long been in her terms a foundherentist[7]. Consider for example, sensory reports. Though they are intrinsically credible, both criticism and support can be based on judgments that need support to be credible. Suppose that I reported a nineteen-sided after image. My confidence would increase or decrease depending on whether I was told that the figure I had observed had nineteen sides. The statement about the physical sides would need support, but it could be provided by the kinds of judgments needed to justify belief in an external world.
I used this sort of example against the claim that incorrigibility is the mark of the mental[8]. A more mundane example would have shown that the chance of error is often high, for we are very likely to be mistaken if our reports are detailed and quite determinate as to the qualities in question (say, shades of color). Accordingly, intrinsically credible judgments may be less credible than supported judgments. For instance, the judgment that a figure has three physical sides is far more likely to be true than the judgment that a visual image has forty-eight phenomenal sides.
The key difference between our views is instead that I rely on beliefs about our experiences, while Haack (1996) rejects them on the grounds that “though experience can stand in a causal relation to beliefs, it can’t stand in logical relations to propositions.” 
I agree that the relation isn’t logical and is causal. However, a causal relation can also be a cognitive relation, and since all beliefs are caused and some are justified, some kinds of causal relation provide justification. This can apply not only when beliefs are caused by apprehending logical or inductive relations, but also when they are caused by and are about sensory experiences. It is particularly hard to reject beliefs about such sensations as aches, pains, and itches that we have ordinary ways of describing. I’ve already rejected Wittgenstein’s claim that we can’t describe any of our experiences because we would have to rely on unverifiable memories of similarities with past experiences to do so, and I deal later with Austin’s attempt to show that we can’t describe our visual experiences.
Though I grant that sensory judgments can be supported or challenged by non-foundational beliefs, I hold that they are in various degrees intrinsically credible. Haack rejects coherentism[9] on the grounds that we need beliefs caused by experiences to support our other beliefs, but to avoid coherentism she needs to assume that beliefs about physical objects caused by sensory experiences are intrinsically credible, i.e., to some extent credible without support. Since she holds that causal relations to sensory experiences are not cognitive, she needs to give a different reason for trusting beliefs caused by sensory experiences.
Another difference is that while Haack’s counterpart of a foundation is limited to the causal and, for her non-cognitive, role of experiences in our judgments about objects, my foundation includes eight kinds of judgments including judgments about experiences. Nevertheless, setting aside my response to her objections to foundationalism, Evidence and Inquiry is the best example of naturalized epistemology that I know of.

Using The Foundation’s Components To Defend Other Minds
Consider next our belief in the mental part of an external world. Although many philosophers accept a physical world but reject the argument from analogy for other minds, the same epistemic premises suffice in both cases.
The chief objection is that you cannot use a single person to justify an inference about a vast number of persons. However, the objection to reliance on a single case doesn’t work. As in purely physical matters you would have to reject causal regularity and therefore an essential part of the defense of a physical world to believe that you were the one exception in a world of creatures with bodies like yours in all respects regarded as causally relevant to your own consciousness. Scientists might discover that they lack a tiny organ that was crucial for you, but empirical hypotheses are never certain.
Consider too the task of explaining why the brain should be causally related to consciousness in one case but unrelated to it in all others. Here too one would have to reject causal regularity since this would require among other things two radically different causal explanations for rational behavior, motivation, and the role of our sensory organs.
Though it may also be objected that we are not directly aware of other persons’ experiences, the same applies to physical objects. In both cases we must rely on indirect confirmation. And though we are twice removed from other persons’ experiences but only once removed from physical objects, inferring the existence of one kind of thing entitles us to use it to infer the existence of another. Furthermore, we have knowledge by acquaintance of experiences, but not of physical objects.

Rejections of the Predicament
I turn now to rejections of the predicament, four attempts to overcome it, and twelve to show that it can’t be overcome. In doing so, I respond to objections to several components of the foundation. I conclude by stressing again the value of addressing the predicament because of the importance of accepting the foundation’s components in other areas.
1.     Rejection: Heidegger (1962) showed that there can’t even be a coherent formulation of the predicament in the following passage: “The question of whether there is a world at all and whether its Being can be proved makes no sense if it is raised by Dasein as Being-in-the-world; and who else could raise it?”
Response: First, his claim that the question could only be raised by a Dasein, a being-in-the-world, assumes the consequent. This is enough to invalidate his argument. Second, for the question to be incoherent you must be convinced that you are a being-in-the-world. You could, however, both be a being-in-the-world and normally believe that you were but still, like Descartes, come to have doubts on the grounds that it would be external to your consciousness.
Heidegger would have replied to the second objection that a Dasein is alongside things in the world instead of external to them. “When Dasein directs itself towards something and grasps it, it does not somehow first get out of an inner sphere in which it has been proximately encapsulated, but its primary kind of Being is such that it is always outside alongside entities that it encounters and which belong to a world already discovered.”
Even if, waving the first objection, we assume that we are in the world so our bodies can be alongside entities in the world, it doesn’t follow that there is a “primary kind of being” in which our consciousness can be alongside them. The acceptance of an external world is an intentional feature of perceptual experiences. We automatically interpret their sensory aspect as identical with external objects, but interpretations are not external objects.
2.     Rejection: There isn’t a predicament. We perceive physical objects directly. The argument from illusion was designed to show that we do not, but it failed.
Response: Setting aside the question of whether the argument from illusion succeeds, the key point remains that even if there were no illusions sensory experiences would still be distinct from the objects we perceive so we do not perceive them directly. We are directly aware of our sensory experiences but not of physical objects. Furthermore, suppose that we become convinced that we perceive them directly; once we accept a physical world we would be forced to change our minds since there is ample evidence that there must be intervening light rays.
3.     Rejection: There isn’t a predicament. Direct realism is true. There are no such internal entities as sense data.
Response: Though use of the technical term “sense data” makes the rejection seem more plausible, it is clear that we have sensory experiences and that we are directly aware of them.
4.     Rejection: Ordinary language philosophy shows that there isn’t a predicament. The fact that we not only have terms standing for external objects, but make successful prediction using them, shows that there are external objects.
Response: First, the fact that we have the term “miracle” doesn’t show that there are miracles. Second, we would also make successful predictions if phenomenalism were true.

 Attempts To Overcome the Predicament
1.     Attempt: A chair is a paradigm external object. There are chairs. Therefore there are external objects.
Response: The argument is persuasive if in agreeing that chairs are external objects, you interpret “external object” in a sense that only places chairs in the same class as tables, rocks etc., but in the conclusion in a sense that requires them to exist independently of our minds.
Suppose it was held instead that (a) there is a chair, and that (b) a chair is a paradigm example of objects that exist independently of our minds, so there are objects that exist independently of our minds. This avoids an equivocation, but premise (b) assumes the consequent.
2.     Attempt: Wittgenstein’s private language argument shows that language requires other people, and there is language, so there must be an external world.
Response: First, Wittgenstein deserves credit for seeing that we can’t verify or even confirm memories of past experiences. Nevertheless, he failed to see that the same objection would force him to reject memories of publicly observable events including memories of the usage of other persons. And since the same would hold for all memories, rejecting memory would force him to accept solipsism of the present moment. Once this is recognized it becomes obvious that regardless of the epistemic barrier we must rely on memory. 
Second even if his argument was successful the most that it could be used to show is that there must seem to be other persons rather than that they actually exist in an external world.
3.     Attempt: It follows from the private language argument that skepticism regarding the external world cannot be intelligibly stated.
Response: Here too it would be sufficient for there to seem to be other persons in the world.

 Arguments That the Predicament Is Insuperable
1.     Argument: Descartes was right. Nothing less than the certainty of an a priori proof would suffice, but an a priori proof is impossible.
Response: Though Descartes founded modern epistemology by calling attention to the egocentric predicament, he barred a solution by requiring certainty. The question of an external world is empirical, and no empirical theory is certain. Furthermore, as Descartes’ ontological argument shows, even a priori arguments can be mistaken.
2.     Argument: You would need to base mathematical calculations on subjective probability judgments. That is impossible because such judgments aren’t precise. I might, for instance, be almost certain that my visual image had seven sides, but I wouldn’t be prepared to assign a probability of say, exactly 99.9996%.
Response: The same problem applies to all empirical judgments: they are never certain and our probability assessments are never precise. Thus probability theory would be useless if the problem were insoluble. Fortunately there are ways of coping with it.
A common way is to take judgments about which you are almost certain and treat them as certain. This makes the probability after the mathematical calculations a bit too high, but you can take that into account. Alternatively you can base your calculations on a slightly lower intuitive probability than the highest that you would definitely not regard as too high[10].
3.     Argument: The only argument for a physical world that is even initially plausible is that our experiences require an external causal explanation, a physical world to cause them. But they do not. Experiences can exist without causes because causation only involves constant conjunction.
Response: If Hume was right, correlations would be happenstance and therefore, as I’ve observed, extraordinarily unlikely without an ontological basis[11]. Hume recognized the problem to the extent of doubting future correlations, but he failed to see the equally pressing problem for past correlations.
4.     Argument: An empirical defense is impossible since it would require direct awareness of external objects as well as experiences. We aren’t directly aware of external objects, so we can’t even begin to confirm the hypothesis that there is an external world.
Response: Confirmation needn’t be direct. Predictions based on the ascription of a causal role to external objects provide ample grounds for accepting indirect confirmation. Nor is it plausible to hold that fictitious objects could be casually efficacious. Also I know of no philosophers who reject indirect confirmation without sometimes tacitly accepting it. For instance phenomenalists reject indirect confirmation for physical objects but accept if for both the existence and nature of other persons’ experiences. And though Wittgenstein rejected indirect confirmation of the nature of other persons’ experiences, he implicitly accepted it for the existence of their experiences. It is obvious that the practice of accepting indirect confirmation in some areas but rejecting it in others needs a defense, but I have never seen or even heard of one, and without a defense, it is time to abandon it.
5.     Argument: If, as you hold, we need physical necessity to avoid the happenstance objection, we can’t avoid the objection because we can’t verify or even directly confirm that there is physical necessity so the claim that it exists isn’t even meaningful. For the same reason the claim that there is an external world is also meaningless.
Response: First, here too no one is consistent in regarding all statements that aren’t verifiable or directly confirmable as meaningless. Nevertheless, of the numerous refutations the verifiability and confirmability criteria of meaningfulness my favorite is the late C. B. Martin’s observation that we must understand a statement’s meaning to even begin to determine whether it is either verifiable or at least confirmable. Meaningless statements are unconfirmable, but we don’t show that they are meaningless by showing that something other than their lack of meaning blocks their confirmation. In contrast we can show that statements about other minds can’t be directly confirmed because we are only directly aware of our own experiences, and we can explain why some statements aren’t even indirectly confirmable.
The criteria were originally designed to discredit such statements as that the absolute is omnipresent. Such statements are indeed meaningless. But since they are meaningless, that is the only reason we can give for saying that they are. To do more, we would have to know their meanings. Otherwise we wouldn’t know what would have to be verified or confirmed. Thus the criteria are useless in regard to meaning.
6.     Argument: We would need to recognize objective similarities to discover causal relations but similarities are imposed by our minds rather than objective.
Response: Though we attend to the similarities in which we are interested, that doesn’t prevent them from being objective, and we can be interested in similarities relevant to causal relations.
7.     Argument: Your argument for an external world assumes that we can describe our sensory experiences. Even if Wittgenstein failed to show that we can’t, Austin succeeded. Locutions used by phenomenalists and representational realists for this purpose pertain instead in one way or another to external objects.
Response: S.C. Coval has shown that we can use “adjusters” to make such locutions pertain to sensory experiences instead of objects. In addition, since locutions may acquire new meanings when users intend to communicate them and listeners grasp their intent, we can describe our sensory experiences by expressions that originally pertained to external objects. I sometimes intend to describe mine, and sometimes interpret others as describing theirs, and I’m not unique in this respect.
Though the semantic theory underlying this response resembles Paul Grice’s, his view needs to be modified to avoid objections. For Grice you must intend to induce a belief in an audience. It is objected that this isn’t our intention when we use a sentence ironically or metaphorically, and the same applies to parodies, tales, and charades. This objection is avoided by the broader view that we can also take advantage of the context to induce various other propositional attitudes as well as belief.
Davidson objects that Grice doesn’t explain how we communicate a belief, and the same would apply to communicating a proposition as a candidate for any other attitude. But conversational implicature works so well that Davidson should reconsider his position instead of requiring an explanation that would satisfy philosophers like himself and Quine.
Furthermore, it follows from their sort of view that we should reject analyticity, while one can use conversational implicature to defend it from the most plausible argument against it, namely that since the same sentence can almost always be used to express more than one proposition, very few sentences are analytic[12]. For while this is true of sentences, it does not apply to the propositions we can use them to express, since we can intend to use a sentence in a way that makes it express an analytic proposition and others can grasp our intention. I count this as a major advantage for the notion of conversational implicature since I believe that we should think long and hard before accepting both the rejection of analyticity and other highly counterintuitive doctrines. Nevertheless, Quine’s attack was useful for it showed the need to distinguish the view that sentences are analytic from the view that they can be used to express analytic propositions.
Since the modified use of a word may become standard practice in the right context, Grice’s view also explains how words acquire the new meanings that make dictionaries give numerous definitions of the same expression. Furthermore, contexts and intent can be used to explain how the most primitive linguistic conventions are established. Suppose that a context includes something that is likely to interest your audience, say food, and you emit a word-like sound that isn’t yet a word to call attention to it. If you succeed, and use it for the same purpose in the future it will have become a word. And you may use it later when that kind of thing is absent but needs to be obtained, and still later when you want them to think about it for some other purpose.
A.C. Grayling (1997) holds that Davidson points out the chief difficulty in objecting that Grice fails to take account of compositionality, the fact that a sentence’s meaning depends on the meanings of the constituent expressions and the way they are combined. For instance, the meaning of “There are four lakes among these mountains” is quite different if one replaces lakes by towns. Presumably the difference is that lakes, unlike towns, must be low in relation to the mountains so the difference lies in the way we interpret “among.”
However, the use of common knowledge and the context within a sentence to select the meaning is still in accord with Grice’s key insight. In compositionality there is a choice between standard meanings and the choice depends on the context within the sentence and common knowledge about the entities in question. Given the role of context and shared knowledge, Grice’s theory can be expanded to include compositionality.
It may still be objected that the semantic theory I’ve described is incomplete. It is, but a complete theory should include it.
8.     Argument: Even if there are concepts, you can only acquire a concept by an ostensive definition, and ostensive definitions require a linguistic preparation. This is impossible for private experiences.
Response. If ostensive definitions always required linguistic preparation, we couldn’t even begin to acquire a vocabulary. A child can’t be given a linguistic preparation for the first words it learns. Preparation is in order when examples are likely to be misinterpreted, but we often arrive at the desired concepts without it. This is particularly likely if we are already interested in the sort of thing in question. Consider an ostensive definition of “food.” A single example may suffice, and several examples of different kinds of foods are very likely to suffice. Furthermore, we don’t acquire all concepts by ostensive definitions.
9.     Argument: Your attempt to defend an external world requires judgments about sense data, i.e., uninterpreted sensory experiences, but there are none.
Response: Some sensory experiences aren’t interpreted, but as I’ve noted in defending concepts, the sensory component in perceptual experiences is interpreted. This doesn’t, however, keep us from making judgments about the sensory component and recognizing the similarity of sensory experiences with different interpretations. In any case, acknowledging the possibly misleading effect of interpretations is consistent with overall fallibilism.
10.  Argument: Your argument for an external world fails because it leaves open many possibilities as to its nature.
Response: As Frank Jackson (1972) has observed, it only follows that it is much harder to determine the nature of the external world than to defend its existence.
11.  Argument: Even if, as you have argued, we are entitled to rely on memory to compare present experiences with past experiences, Wittgenstein was still right to the extent that we don’t describe them. Using pain reports as an example, he showed that instead of descriptions the locutions are really non-cognitive substitutes for “the primitive, natural expressions of the sensation and used in their place. A child has hurt himself and he cries: and then adults talk to him and teach him exclamations and, later, sentences. They teach the child new pain behavior (Ibid section 244).”
Response: This would be hard to buy if, when asked if I still had a headache, I responded that it still ached but less than before. It would be harder if responding to a doctor’s questions, they were about a previous pain and therefore not a kind of pain behavior, and I said that it was only in the rear part of the left side of my throat, that it was dull rather than sharp, and was constant rather than intermittent.
And it would be virtually impossible to buy if I said exactly the same thing but was lying.
It is clear instead that we sometimes both intend to talk about our experiences and are interpreted as doing so. Though it is important to show that we could rely on conversational implicatur to do so, it would be surprising if we didn’t have expressions designed for this purpose, and it seems clear that we have such expression even if we shouldn’t have them.
And for the same reasons the meanings of such expressions cannot be replaced by the publicly observable conditions of their use.
12.  Argument: Nevertheless, regardless of our actual practice, in a public language we ought to define pain in terms of publicly observable behavior because we can’t determine the nature of other persons’ experiences. (Ibid sec. 271).
Response: In this case I’ve already provided a response by defending the argument from analogy.

We must accept eight epistemic principles to defend our belief in an external world including other minds. Though the principles can also be defended individually, showing that we must accept them to defend this belief provides considerable further support for those that are sometimes questioned.
Proponents of naturalized epistemology hold that bypassing the predicament has the practical advantage of enabling us to concentrate on more important issues. However, addressing the problem shows that we must accept kinds of judgment that are often rejected in other areas including naturalized epistemology. Furthermore, the alternative is to rely on pragmatic faith, which sets a precedent for faith in general, unless faith is limited, as by current pragmatists, to beliefs that are plausible because, unlike religious beliefs, there is supporting evidence, in which case, the problem remains.
As I see it, epistemologists have three main tasks. The first is to determine what methodological principles must be accepted to defend such seemingly obvious beliefs as that there is a physical world and that there are other minds instead of either rejecting the beliefs or relying on pragmatic faith to accept them. The second is to determine whether the principles are justifiable. If they are, the third is to call attention to the fact that it follows that we are entitled to apply them to other areas. My main objective in writing this paper was to show that we are entitled to do so.

BonJour, Laurence. 1985. The Structure of Empirical Knowledge. Cambridge, Mass.: Harvard University Press.
BonJour, Laurence. 2002. Epistemology: Classic Problems and Contemporary Responses. Lanham, Md.: Rowman & Littlefield.
Goldman, Alan H. 1988. Empirical Knowledge. Berkeley: University of California Press.
Grayling, A. C. 1997. An Introduction to Philosophical Logic. 3rd ed. Oxford: Blackwell Publishers Ltd.
Haack, Susan. 1993. Evidence and Inquiry: Towards Reconstruction in Epistemology. Oxford, UK: Blackwell.
Heidegger, Martin. 1962. Being and Time, trans. J. Macquarrie and E. Robinson. New York: Harper.
Lewis, Clarence Irving. 1947. An Analysis of Knowledge and Valuation. La Salle, Ill.: The Open Court Pub. Co.
Strawson, Galen. 1989. The Secret Connexion: Causation, Realism, and David Hume. Oxford: Clarendon Press.
Wittgenstein, Ludwig, and G. E. M. Anscombe. 1953. Philosophische Untersuchungen Philosophical Investigations. Oxford: Blackwell.

[1]I’ll add a reference to a more extended defense.
[2]Furthermore, the objective synthetic a priori was accepted as early as Locke and Berkeley. Locke distinguished trivial from interesting a priori truths and Berkeley relied on the distinction in holding that pleasure, pain, and color are necessarily phenomenal. An example is Russell’s observation that it is logically possible but impossible a priori for a surface to be red all over and also green all over, but while its interest is limited to showing the possibility of such judgments, I show elsewhere that there are many synthetic a priori truths to which this does not apply.
[3]Though Galen Strawson (1989) makes this point, he offers persuasive arguments that Hume regarded causal relations as physically necessary. Nevertheless, I find this hard to reconcile with Hume’s doubts about induction.
[4]It is worth adding that the existence of physical objects derives further indirect support from the fact that visual shapes and sizes not only resemble tactual shapes and sizes but are related to tactual shapes and sizes in ways that would be implausibly happenstance without comparable shapes and sizes in external objects.
[5]Though Galen Strawson (1989) makes this point, he offers persuasive arguments that Hume regarded causal relations as physically necessary. Nevertheless, I find this hard to reconcile with Hume's doubts about induction.
[6]It is worth adding that the existence of physical objects derives further indirect support from the fact that visual shapes and sizes not only resemble tactual shapes and sizes but are related to tactual shapes and sizes in ways that would be implausibly happenstance without comparable shapes and sizes in external objects.
[7]The same holds for (BonJour 2003).
[8]Since references to the exchanges on this subject would identify the author, I’ll add them later.
[9](BonJour 1985) provides what is probably the best defense of a coherence theory of knowledge though not truth. However, he later abandoned it (BonJour 2002) in favor a foundationalism. Though his later book was written for students, it merits the attention of philosophers.
[10]For a further defense of subjective probability see Henry E. Keyburg, Jr., and Howard E. Smoker 1964.
[11]For a defense of this view see chapter five in (Strawson 1989). He also argued that Hume accepted a secret physically necessary causal relation and only held that we couldn’t describe it. Nevertheless, Hume’s doubts about induction suggest that he wasn’t consistent in this respect.
[12]Though propositions are sometimes rejected they can be defended as timeless possibilities of thoughts that may or may not be instantiated in actual thoughts. An advantage of this account is that there can be, as there surely are, truths about the world that are not reflected in actual thoughts.