quinta-feira, 23 de novembro de 2017

# ON PHILOSOPHICAL METHODOLOGY

This is only a DRAFT. The final version will be published in the book PHILOSOPHICAL SEMANTICS, by Cambridge Scholars Publishig, in 2018/1 



AGAINST THE METAPHYSICS OF REFERENCE:
METHODOLOGICAL ASSUMPTIONS


Eine Art zu philosophieren steht nicht neben anderen wie eine Art zu Tanzen neben anderen Tanzarten ... Die Tanzarten schließen sich nicht gegenseitig aus oder ein … Aber man kann nicht ernsthaft auf eine Art philosophieren, ohne die anderen verworfen oder aber einbezogen zu haben. In der Philosophie geht es demgegenüber wie in jeder Wissenschaft um Wahrheit.

[A way of philosophizing is not one way among others, like one way of dancing among others … Ways of dancing are not mutually exclusive or inclusive … But no one can seriously philosophize in one way without having dismissed or incorporated others. In philosophy as in every science, the concern is with truth.]
Ernst Tugendhat
Philosophy has no other roots but the principles of Common Sense; it grows out of them, and draws its nourishment from them. Severed from this root, its honours wither, its sap is dried up, it dies and rots.
Thomas Reid
Given the commonsense assumptions involved when we take the social role of language as a starting point, at least part of this book must be critical. The reason is clear. The new orthodoxy that dominates much of contemporary philosophy of language is based on what we could call a metaphysics of reference and meaning. Its views often focus on reference more than on meaning, or on something like reference-as-meaning, displaying a strong version of semantic externalism, hypostatized causalism and anti-cognitivism. I call these views metaphysical not only because they oppose modest common sense, but mainly because, as will be shown, they arise from sophisticated attempts to unduly ‘transcend’ the limits of what can be meaningfully said.
  One example of the metaphysics of reference is the position of philosophers like Saul Kripke, Keith Donnellan and others on how to explain the referential function of proper names and natural species terms. According to them, it is not our cognitive access to the world but rather the mere appeal to external causal chains beginning with acts of baptism that really matters. On the other hand, what we may have in mind when using a proper name is for them secondary and contingent.
  Another example is the strong externalist view of Hilary Putnam, John McDowell, Tyler Burge and others, according to whom the meaning of an expression, its understanding, thought, and even our own minds (!) in some way belong to the external (physical, social) world. Using a metaphor always hinted at but never spelt out, it is as if these things were floating outside, determined by the entities referred to with words, in a way that recalls Plotinus’ emanation, this time not from the ‘One’, but in some naturalistic fashion, from the ‘Many.’ By writing this, I am not mocking, but only trying to supply the right images for what is explanatorily wanting. In fact, externalism is an unclean concept. After refinements, externalism is defined in a vague way as the general idea that ‘certain types of mental contents must be determined by the external world’ (Lau & Deutsch 2014). This is an obvious truism, insofar as we understand the expression ‘determined by the external world’ as saying that any mental content referring to the external world is in some way or the other causally associated with things belonging to an external world. As Leslek Kolakowski noted, ‘if there is nothing outside myself, I am nothing’ (2001). But this is trivial enough to be accepted by a reasonable internalist like myself (or by a very weak externalist, which in my view amounts to the same thing). Nonetheless, externalists have proposed in their most central and radical writings to read ‘determined’ as suggesting that the locus of meanings, beliefs, thoughts and even minds is not in our heads, but somewhere in the external world… However, this sounds very much like a genetic fallacy.
  A third example is the view accepted by David Kaplan, John Perry, Nathan Salmon and others, according to whom many of our statements have as their proper semantic contents structured propositions, whose constituents (things, properties, relations) belong to the external world alone, as if the external world had any proper meaning beyond the meaning we give to it. As a last example – which I examine in the present chapter – we can take the views of John McDowell and Gareth Evans. According to them, we cannot sum up most of the semantics of our language in tacit conventional rules that can be made reflexively explicit, as has been traditionally assumed. Consistent with causal externalism, their semantics tends to take the form of things that can be understood chiefly in the third person, like the neuronal machinery responsible for linguistic dispositions unable to become objects of reflexive consciousness.
  Notwithstanding the fact that most such ideas are contrary to the semantic intuition of any reasonable human being who hasn’t yet been philosophically indoctrinated, they have become the mainstream understanding of specialists. Today many theorists still view them as ‘solid’ results of philosophical inquiry, rather than crystallized products of ambitious formalist reductionism averse to cognitivism. It is true that they have in the meantime rhetorically softened their extreme views, though still holding them in vaguer, more elusive terms. However, if taken too seriously, such ideas can both stir the imagination of unprepared thinkers and, more seriously, limit their scope of inquiry.
  In the course of this book, I intend to make plausible the idea that the metaphysics of reference is far from having found the ultimate truth of the matter. This is not the same, I must note, as to reject the originality and philosophical relevance of its main arguments. If I did reject them on this ground, there would be no point in discussing them here. Such philosophical arguments usually cover insight under their suggested illusion, and remain of interest even if they are in the end-effect flawed. If so, they would ultimately require not additional support, but careful critical analysis. In the process of disproving them, we could develop views with greater explanatory power, since philosophical progress is very often dialectical. For this reason, I think we should judge the best arguments of the metaphysics of reference in the same critical way we value McTaggart’s argument against the reality of time or Berkeley’s remarkable arguments against materialism. Consider Hume’s impressive skeptical arguments to show there is nothing in the world except flocks of ideas – an absurd conclusion that was first countered by Thomas Reid. What all these arguments surely did, even if we are unable to agree with them, was to draw illusory consequences from insufficiently known conceptual structures, presenting in this way real challenges to philosophical investigation, useful insofar as they force us to answer them by more deeply analyzing the same structures, as they really are. Indeed, without the imaginative and bold revisionism of the metaphysicians of reference, without the challenges and problems they presented, it is improbable that corresponding competing views would ever acquire enough intellectual fuel to get off the ground.

1. Common sense and meaning
To contend with the metaphysics of reference, some artillery pieces are essential. They are methodological in character. The first concerns the decision to take seriously the so often neglected fundamental principles of common sense and natural language philosophy, respectively assumed by analytic philosophers like G. E. Moore and the later Wittgenstein. According with philosophers of this lineage, we should seek the starting point of our philosophical arguments as much as possible in pre-philosophical commonsense intuitions usually reflected in our natural language. The link between common sense and natural language is easy to understand. We should expect that commonsense intuitions – often due to millennia of cultural sedimentation – will come to be strongly mirrored in our linguistic forms and practices.
  As Noah Lemos wrote, we can characterize commonsense knowledge as:

...a set of truths that we know fairly well, that have been held at all times and by almost everyone, that do not seem to be outweighed by philosophical theories asserting their falsity, and that can be taken as data for assessing philosophical theories (2004: 5).

Indeed, commonsense truths seem to have always reconfirmed themselves, often approaching species wisdom. Examples of commonsense statements are: ‘Black isn’t white,’ ‘Fire burns,’ ‘Material things exist,’ ‘The past existed,’ ‘I am a human being,’ ‘I have feelings,’ ‘Other people exist,’ ‘The Earth has existed for many years,’ ‘I have never been very far from the Earth,’… (e.g., Moore 1959: 32-45). Philosophers have treasured some of these commonsense statements as particularly worthy of careful analytical scrutiny. These include: ‘A thing is itself’ (principle of identity), ‘The same thought cannot be both true and false’ (principle of non-contradiction), ‘I exist as a thinking being’ (version of the cogito), ‘The external world is real’ (expressing a realist position on the external world’s existence), and even ‘A thought is true if it agrees with reality’ (adequation theory of truth).
  The most influential objection to the validity of commonsense principles is that they are not absolutely certain. Clearly, a statement like ‘Fire burns’ isn’t beyond any possibility of falsification. Moreover, science has truly falsified many commonsense beliefs. Einstein’s relativity theory decisively refuted the commonsense belief that the length of a physical object remains the same independently of its velocity. But there was a time when people regarded this belief as a self-evident truth!
  This latter kind of objection is particularly important in our context, because metaphysicians of reference have made this point to justify philosophy of language theories that contradict common sense. Just as in modern physics new theories often conflict with common sense, they feel emboldened to advance a new philosophy whose conclusions depart radically from common sense and natural language. As Hilary Putnam wrote to justify the strangeness of his externalist theory of meaning:

Indeed, the upshot of our discussion will be that meanings don’t exist in quite the way we tend to think they do. But electrons don’t exist in quite the way Bohr thought they did, either. (Putnam 1978: 216)

One answer to this kind of comparison emphasizes the striking differences between philosophy of meaning and physics: the way we get meanings is much more direct than the way we discover the nature of subatomic particles. We make meanings; we don’t make electrons. We find subatomic particles by empirical research; we don’t find meanings: we establish them. We don’t need to read Plato’s Cratylus to realize that the meanings of our words are dependent on our shared semantic customs and conventions.

2. Critical common-sensism
Nonetheless, a key question remains unanswered: how certain and indisputable are our commonsense intuitions? C. S. Peirce, undermining Thomas Reid’s unshakeable belief in common sense and based on his own thesis of the inevitable fallibility of human knowledge, proposed replacing traditional common-sensism with what he called critical common-sensism. According to this theory, slow changes really do take place in commonsense beliefs, even if not in our most central beliefs. This change can occur particularly as a response to scientific progress. Consequently, common sense in general, though highly reliable, is not beyond any possibility of doubt. Still, for heuristic reasons we should maintain a critical attitude and always be ready to submit commonsense views to the scrutiny of reasonable doubt (cf. Peirce 1905: 481-499).
  The idea that our commonsense views are open to revision has been attacked from various standpoints. As we will see, one argument against it is that scientific progress has not altered the most proper forms of commonsense views. Another is that we cannot use falsification to disprove the claims of common sense, because this would require a criterion to distinguish true sentences from false ones. This criterion, however, could not itself rest on common sense, since this would involve circularity…
  The answer to this last objection is that it isn’t necessarily so. First, because commonsense beliefs have different levels of reliability and form a hierarchy (for instance, ‘I exist’ is clearly more reliable than ‘fire burns’). Thus, it seems possible to employ the most trustworthy commonsense beliefs, possibly in combination with scientific beliefs, to falsify or at least restrict the domain of application of some less reliable commonsense beliefs. Moreover, it could very well be that the most fundamental commonsense beliefs can be so intrinsically justified that an analytical justification may be all that philosophy requires.[1] Anyway, insofar as some commonsense beliefs seem vulnerable to refutation, it seems advisable to preserve the attitude of critical common-sensism.

3. Ambitious versus Modest Common Sense
I do not have the ambition to end debates over the ultimate value of common sense. However, I believe I can demonstrate that two deeply ingrained objections against the validity of commonsense principles are seriously flawed, one based on the progress of science and the other based on changes in our worldviews (Weltanschauungen). The first is that science defeats common sense. This can be illustrated by the claim attributed to Albert Einstein that common sense is a collection of prejudices acquired by the age of eighteen… (Most physicists are philosophically naïve.) Changes in worldviews are transformations in our whole system of beliefs, affecting deeply settled ideas like moral values and religious beliefs. In my view, these two charges against common sense are faulty because they arise from confusion between misleading ambitious formulations of commonsense truths and their authentic formulations, which I call modest ones.
  I intend to explain my point by beginning with a closer examination of objections based on the progress of science. With regard to empirical science, consider the sentences:

(a)    The Earth is a flat disk with land in the center surrounded by water.
(b)   The sun is a bright sphere that revolves around the Earth daily.
(c)    Heavy bodies fall more rapidly than light ones, disregarding air resistance.
(d)   Time flows uniformly, even for a body moving near the speed of light.
(e)    Light consists of extremely small particles.

According to the objection, it is widely known that science has disproved all these once commonsense statements. Already in Antiquity, Eratosthenes of Alexandria was able not only to disprove the Homeric view that (a) the Earth is a flat disk rimmed by water, but was even able to measure the circumference of the Earth with reasonable precision. Galileo showed that (b) and (c) are false statements, the first because the Earth circles the sun, the second because in a vacuum all bodies fall with the same acceleration. And Einstein’s relativity theory predicted that time becomes exponentially slower as a body approaches the speed of light, falsifying statement (d). Bertrand Russell once pointed out that the theory of relativity showed that statement (d), like some other important commonsense beliefs, cannot withstand precise scientific examination (cf. Russell 1925, Ch. 1; Popper 1972, Ch. 2, sec. 2). Finally, statement (e), affirming the seemingly commonsense corpuscular theory of light (defended by Newton, but already suggested in the Antiquity), has been judged to be mistaken, since light consists of transverse waves (Huygens-Young theory), even though under certain conditions it behaves as though it consisted of particles (wave-particle theory).
  A point I wish to emphasize, however, is that none of the four above-cited statements legitimately belongs to correctly understood common sense – a sense I call ‘modest.’ If we examine these statements more closely, we see they are in fact extrapolations grounded on statements of modest common sense. These extrapolations are of speculative interest and were made in the name of science by scientists and even by philosophers who projected ideas of common sense into new domains that would later belong to science. In my view, true statements of common sense – the modest statements for which (a), (b), (c), (d) and (e) could be the corresponding non-modest extrapolations – are respectively the following:

(a’) The Earth is flat.
(b’) Each day the sun crosses the sky.
(c’) Heavier bodies fall more rapidly than lighter ones.
(d’) Time flows uniformly for all bodies around us, independently of their motion.
(e’) Light consists of rays.

Now, what is at stake is that these statements have been made for thousands years and have always been confirmed by everyday observation. It is obvious that (a’) is a true statement if we understand it to mean that when we look at the world around us without having the ambition to generalize this observation to the whole Earth, we see that the landscape is obviously flat (discounting hills, valleys and mountains). Statement (b’) is also true, since it is anterior to the distinction between the real and the apparent motion of the sun. Because of the distinction between the apparent and the real motion of the sun, we know that the sentence ‘The sun crosses the sky each day’ can be true without implying that the sun revolves around the Earth. All it affirms is that in equatorial and sub-equatorial regions of the Earth we see that each day the sun rises in the East, crosses the sky, and sets in the West, which no sensible person would ever doubt.[2] Even after science proved that bodies of different masses fall with the same acceleration in a vacuum, statement (c’) remains true for everyday experience. After all, it only affirms the commonplace notion that under ordinary conditions a light object such as a feather falls much more slowly than a heavy one such as a stone... Statement (d’) also remains true, since it concerns the movements of things in our surroundings, leaving aside extremely high speeds or incredibly accurate measurements of time. (In everyday life, one would never need to measure time dilation, which is detectable only when a body approaches the speed of light, which has nothing to do our daily experience. In everyday life, no one ever comes home from a two-week bus trip to discover that family members are now many years older than before). Finally, (e’) has been accepted, at least since Homer, as is shown by his poetic epithet ‘rosy-fingered dawn.’ And we often see sunrays at dawn or dusk or peeping through gaps in the clouds on an overcast day.
  But then, what is the point in comparing statements (a)-(b)-(c)-(d)-(e) with the corresponding statements (a’)-(b’)-(c’)-(d’)-(e’), making the first set refutable by science, while the latter statements remain true? The answer is that scientifically or speculatively motivated commonsense statements exemplified by (a)-(b)-(c)-(d)-(e) have very often been viewed equivocally, as if they were legitimate commonsense statements. However, statements of modest common sense like (a’)-(b’)-(c’)-(d’)-(e’) are the only ones naturally originating from community life, being omnipresent in the most ordinary linguistic practices. They continue to be perfectly reliable even after Galileo and Einstein, since their truth is independent of science. The contrast between these two kinds of example shows how injudiciously mistaken the claim is that many or most commonsense truths have been refuted by science.[3] What science has refuted are extrapolations of commonsense truths by scientists and philosophers who have projected such humble commonsense truths beyond the narrow limits of their original context. If we consider the aforementioned distinction, we find a lack of conflict between the discoveries of science and the claims of commonsense wisdom, also including ones used as examples by philosophers like G. E. Moore.
  I do not claim modest commonsense truths are in principle irrefutable, but only that no one has managed to refute them. Nothing warrants, for instance, asserting that from now on the world around us will be different in fundamental ways. A statement like (b’) can be falsified. Perhaps for some unexpected reason the Earth’s rotation on its axis will slow down so much that the sun will cease its apparent movement across the sky. In this case, (b’) would also be refuted for our future expectations. But even in this case, (b’) remains true concerning the past, while the corresponding ambitious extrapolation (b) has always been false. In fact, all I want to show is that true commonsense statements – modest ones – are much more sensible than scientifically oriented minds believe, and science has been unable to refute them, insofar as we take them at their proper face value.
  Similar reasoning applies to the a priori knowledge of common sense. To justify this new claim, consider first the case of statements like (i) ‘Goodness is praiseworthy,’ which is grammatically identical with statements like (ii) ‘Socrates is wise.’ Both have the same superficial subject-predicate grammatical structure. Since in the first case the subject ‘Goodness’ does not designate any object accessible to the senses, Plato would have concluded that this subject must refer to ‘goodness in itself’: the purely intelligible idea of goodness, existing in an eternal and immutable non-visible realm only accessible to the intellect. Plato reached his conclusion based on the commonplace grammatical distinction between subject and predicate found in natural language. Under this assumption, he was likely to see a statement like (iii) ‘Goodness in itself exists’ as a commonsensical truth. In fact, according to his doctrine, an a priori truth.
  However, we know that with Frege’s innovation of quantificational logic at the end of the 19th century, it became clear that statements like (i) should have a deep logical structure that is much more complex than the subject-predicate structure of (ii). Statement (i) should be analyzed as saying that all good things are praiseworthy, or (iv) ‘For all x, if x is good, then x is praiseworthy,’ where the supposed proper name ‘Goodness’ disappears and is replaced by the predicate ‘… is good.’ This new kind of analysis reduced considerably the pressure to countenance the Platonic doctrine of ideas.
  However, the suggestion that the subject ‘Goodness’ refers to an abstract idea clearly does not belong to modest common sense, and statement (iii), ‘Goodness in itself exists,’ isn’t even inscribed in our natural language. It again belongs to ambitious common sense. Statement (iii) was a speculative extrapolation by a philosopher based on an implicit appeal to the superficial grammar of natural language, and though it was probably a bad choice, it would be unjust to blame modest common sense and our ordinary language intuitions on subject-predicate grammar. Finally, it is wise to remember that quantificational logic has not undermined the (commonsensical) grammar of our natural language; it has only selected and made us conscious of vastly extended fundamental patterns underlying the representative function of natural language.
  What all these examples do is to undermine the frequently made claim that scientific progress contradicts common sense. Scientific discoveries only refute speculative extrapolations of common sense and natural language made by scientists and philosophers, such as the idea that the Sun revolves around the Earth or that there is a purely intelligible world made up of abstract ideas like that of Goodness in itself. But nothing of the sort has to do with the explanations given by modest common sense, the only ones long established by the shared practical experience of mankind over the course of history.

4. Resisting changes in worldviews
Finally, I wish to consider commonsense ideas that are challenged by changes in our worldviews. This is, for instance, the case with the belief that a personal God exists or that we have minds independently of our bodies. The objection is the following. The overwhelming majority of cultures accept a God (or gods) and the soul as undeniably real. In Western Civilization for the last two-thousand years, society has even sanctioned denial of these beliefs with varying degrees of severity, sometimes even resorting to capital punishment. Although they were once commonsense beliefs, today no one would say that they are almost universally accepted. On the contrary, few scientifically educated persons would agree with them. Consequently, it seems that commonsense ideas can change in response to changes in our worldviews...
  My reaction to this does not differ very much from my reaction to the objection contrasting common sense with the progress of science. Beliefs regarding our worldviews lack universality, not really belonging to what I call modest common sense. There are entire civilizations, particularly in Asia, where the idea of a personal God is foreign to the dominant religion. Regarding the soul, I remember a story told by an anthropologist who once asked a native Brazilian what happens after people die. The native answered: – ‘They stay around.’ – ‘And later?’ asked the anthropologist. – ‘They go into some tree.’ – ‘And then?’ – ‘Then they disappear’...[4] The lack of concern was evident. And the unavoidable conclusion is that belief in a personal God and an eternal soul do not enjoy the kind of universality that would be expected of modest common sense; if they are said to belong to common sense, must be an ambitious one. In fact, these beliefs seem to result from the distortion of ordinary views through wishful thinking, which has often happened in Western culture.[5]
  Natural language also supports the view that these beliefs are not chiefly commonsensical: a person holding religious beliefs usually does not say he knows that he has a soul independent of her body… He prefers to claim he believes in these things. And even this belief has a particular name: ‘faith,’ which is belief not supported by reason and observation (against faith there are no arguments). On the other hand, the same person would never deny that he knows there is an external world and that he knows this world existed long before he was born… Modest commonsense knowledge is not a question of wishful thinking or non-rational faith.
  What all these arguments suggest is that modestly understood commonsense truths – together with the very plausible discoveries of real science – can reasonably be said to form the basis of our rationality, the highest tribunal of reason. Furthermore, since science itself can only be constructed starting from a foundation of accepted modest commonsense beliefs, it does not seem possible, even in principle, to deny modest common sense as a whole on the authority of science without also having to deny the very foundations of rationality.
  Not only do science and changes in our worldview seem unable to refute modest common sense, even skeptical hypotheses cannot do this in the highly persuasive way one could expect. Suppose, for instance, that radical skeptics are right, and you discover that until now you have lived in what was just an illusory world… Even in this case, you would be unable to say that the world where you lived until now was unreal in the most important sense of the word. For that world would still be fully real in the sense that people perceived it with maximal intensity, and it was independent of the will, was interpersonally accessible and obeyed natural laws… These are criterial conditions that when satisfied create our conventional sense of reality, a sense indefeasible by skeptical scenarios (see Ch. VI, sec. 29).

5. Primacy of Established Knowledge
The upshot of the comparison between modest common sense and science is that we can see science as not opposed to modest common sense, but rather as its proper extension, so that both are mutually supportive. According to this view, science is expanded common sense. Contrary to Wilfrid Sellars (1962: 35-78), the so-called ‘scientific image of the world’ did not develop in opposition to or even independently of the old ‘manifest image of the world,’ for there is no conflict between them. This conclusion reinforces our confidence that underlying everything we can find commonsense truths, insofar as they are satisfactorily identified and understood.
  In endorsing this view, I do not claim that unaided modest commonsense truth can resist philosophical arguments, as philosophers like Thomas Reid have assumed. One cannot refute Berkeley’s anti-materialism by kicking a stone, or answer Zeno’s paradox of the impossibility of movement by putting one foot in front of the other. These skeptical arguments must be wrong, but to disprove them, philosophical arguments are needed to show why they seemingly make sense, again grounding their rejection at least partially in other domains of common sense if not science, something reached only by the comprehensiveness of philosophical reasoning. So, what I wish to maintain is that the principles of modest common sense serve as the most reliable assumptions and that some fundamental modest commonsense principles will always be needed, if we do not wish to lose our footing in everyday reality.
  I am not proposing that a philosophy based on modest common sense and its effects on natural language intuitions would be sufficient. It is imperative to develop philosophical views compatible with and complementing modern science. We must construct philosophy on a foundation of common sense informed by science. That is: insofar as formal reasoning (logic, mathematics…) and empirical science (physics, biology, psychology, sociology, neuroscience, linguistics...) can add new extensions and elements beyond modest commonsense principles, and these extensions and elements are relevant to philosophy, they should be taken into account. As we saw above, it was through the findings of predicate calculus that we came to know that the subject ‘goodness’ in the sentence ‘Goodness is praiseworthy’ should not be logically interpreted as a subject referring to a Platonic idea, since what this sentence really means is ‘For all x, if x is good, x is praiseworthy.’
  I will use the term established knowledge for the totality that includes modest commonsense knowledge and all the extensions the scientific community accepts as scientific knowledge. Any reasonable person with the right information would agree with this kind of knowledge, insofar as he was able to properly understand and evaluate it. It is in this revised sense that we should reinterpret the Heraclitean dictum that we must rely on common knowledge as a city relies on its walls.
  The upshot of these methodological remarks is that we should judge the plausibility of our philosophical ideas against the background of established knowledge, i.e., comparing them with the results of scientifically informed common sense. We may call this the principle of the primacy of established knowledge, admonishing us to make our philosophical theses consistent with it. Philosophical activity, particularly as descriptive metaphysics,[6] should seek reflexive equilibrium with the widest possible range of established knowledge, the knowledge mutually supported by both modest common sense and scientific results. This is the ultimate source of philosophical credibility.
  Finally, if we find inconsistencies between our philosophical theories and our established knowledge, we should treat them as paradoxes of thought, even if they are very instructive, and should search for arguments that reconcile philosophical reflection with established knowledge. Lacking reconciliation, we should treat philosophical theses only as proposals, even if they are stimulating from a speculative viewpoint, as is the case of revisionary metaphysics superbly exemplified by Leibniz, Berkeley and Hume and in considerable measure also by most American theoretical philosophers since W. V-O. Quine. This does not mean that their results require acceptance as ‘solid’ discoveries, but rather that they deserve attentive consideration, the sort we grant to the best cases of expansionist scientism. To proceed otherwise can lead us down the slippery slope to dogmatism.

6. Philosophizing by examples
We must complement our methodological principle of the primacy of established knowledge with what Avrum Stroll called the method of philosophizing by examples. He himself used this method to construct relevant arguments against Putnam’s externalism of meaning (Stroll 1998, x-xi).
  Stroll was a Wittgenstein specialist, and Wittgenstein’s therapeutic conception of philosophy directly inspired his approach. According to Wittgenstein, at least one way of doing philosophy is by performing philosophical therapy. This therapy consists in comparing the speculative use of expressions in philosophy – which is generally misleading – with a variety of examples, most of them of their everyday usage – where these expressions earn their proper meanings, using a method of contrast to clear up confusion. He thought this therapy was only possible through meticulous comparative examination of various real and imaginary concrete examples of intuitively correct (and even incorrect) uses of expressions. This would make it possible to clarify the true meanings of our words, so that the hidden absurdities of metaphysics would become evident... Since contemporary philosophy of language tends to be unduly metaphysically oriented, and in this way diametrically opposed to the kind of philosophy practiced by Wittgenstein, a similar critique of language, complemented by theoretical reflection, is what much of contemporary philosophy needs to find its way back to truth.
  I intend to show that today’s metaphysics of reference and meaning suffers from a failure to consider adequately, above all, the subtle nuances of linguistic praxis. It suffers from an accumulation of potentially obscurantist products of what Wittgenstein called ‘conceptual houses of cards’ resulting from ‘knots of thought’ – subtle semantic equivocations caused by a pressing desire for innovation combined with a lack of more careful attention to nuanced distinctions of meaning that expressions receive in different contexts where they are profitably used.
  One criticism of Wittgenstein’s therapeutic view of philosophy is that it would confine philosophy to the limits of the commonplace. Admittedly, there is no good reason to deny that the value of philosophy resides largely in its theoretical and systematic dimensions, in its persistence in making substantive generalizations. I tend to agree with this, since I also believe that in its proper way philosophy can and should be theoretical, even speculatively theoretical. Nonetheless, I think we can to a great extent successfully counter this objection to Wittgenstein’s views, first interpretatively and then systematically.
  From the interpretative side, we have reason to think that the objection misunderstands some subtleties of Wittgenstein’s position. The most authoritative interpreters of Wittgenstein, G. P. Baker and P. M. S. Hacker, insisted that he did not reject philosophical theorization tout court. In rejecting philosophical theorizing, he was opposing scientism: the kind of philosophical theorization that mimics science. Scientism tries to reduce philosophy itself to science in its methods, range and contents, as he already saw happening in logical positivism.[7] Instead, he would countenance a different sort of theorization, particularly the ‘dynamic,’[8] the ‘organic’ instead of ‘architectonic’ (Wittgenstein 2001: 43) – a distinction he seems to have learned from Schopenhauer (Hilmy 1987: 208-9). This helps explain why, in a famous passage of Philosophical Investigations, he argued that it is both possible and even necessary to construct surveillable representations (übersichtliche Darstellungen). These can show the complex logical-grammatical structure of the concepts making up the most central domains of understanding. As he wrote:

A main source of our failure to understand is that we do not command a clear view of the use of our words – Our grammar is lacking in this sort of surveillability. A surveillable representation produces just that understanding which consists in ‘seeing connections’; hence the importance of finding and inventing intermediate cases. The concept of surveillable representation is of fundamental significance for us. It earmarks the form of account we give, the way we look at things (Is this a ‘Weltanschauung’?). (Wittgenstein 1984c, sec. 122)

Now, in a sense a surveillable representation must be theoretical, since it must contain generalizations, and this constitutes the ultimate core of what we might call a ‘theory.’ [9] If we agree that all generalizations are theoretical, any surveillable representation, as it must contain generalizations, must also be theoretical.
  Moreover, the addition of intermediate connections already existent but not explicitly named by the expressions of ordinary language enables us to make explicit previous conventions that serve as links connecting a multitude of cases. It is possible that because of the generality and function of these links, they never need to emerge in linguistically expressible forms (consider, for instance, our MD-rule for proper names). Expositions of these links are properly called ‘descriptive’, insofar as they are already present under the surface of language. But it is acceptable to call them ‘theoretical’ – in the sense of a description of general principles inherent in natural language – if they are intended to be the right way to assure the unity in diversity that our use of expressions can achieve.
  The addition of intermediary connexions helps to explain why normal language philosophy, as initially developed by Gilbert Ryle and J. L. Austin gradually transformed itself into far more liberal and theoretical forms of philosophy inspired by natural language that we can already find in works of P. F. Strawson and later in H. P. Grice[10] and John Searle. It also helps to justify the introduction of new technical terms to fill the gaps in natural language. Terms like ‘language-game,’ ‘grammatical sentence’ and even ‘surveillable representation’ support this point in Wittgenstein’s own writings. In fact, even Austin, the chief defender of a quasi-lexicographical ordinary language analysis didn’t eschew the creation of new technical terms. Expressions like ‘locutionary act’ (composed of ‘phonetic’, ‘phatic’ and ‘rhetic acts’), ‘illocutionary act’ and ‘perlocutionary act’ (1962, Lect. VIII) were created as the only way to express – guided by reasoning on interactive linguistic activity – fundamental deep structures totally unexpressed in our normal usage.
  Now, from the systematic argumentative side, we can say that independently of the way we interpret Wittgenstein, there are good reasons to believe theoretical considerations are indispensable. An important point is that philosophy can only be therapeutic or critical because its work is inevitably based on theoretical (i.e., generalized) assumptions that make possible its therapeutic efficacy. Usually Wittgenstein did not explicitly stated or developed the assumptions needed to make his conceptual therapy convincing. He was an intuitive thinker in the style of Heracleitus or Nietzsche who all too often did not develop his insights beyond the epigrammatic level. In any case, such assumptions are inevitable, and the result is the same: The critical (therapeutic) and the more constructive (theoretical) searches for surveillable representations can be understood as two complementary sides of the same analytical coin (Costa 1990: 7 f.). Theoretical assumptions were the indispensable active principle of his therapeutic potions.
  Recapitulating, we have found two main methodological principles for orienting our research in this book:

A.   The principle of the primacy of established knowledge (our principle of all principles), according to which modest common sense, complemented by scientific knowledge, constitutes the highest tribunal of reason in judging the plausibility of philosophical views.
B.    The method of philosophizing by examples, according to which the best way to orient ourselves in the philosophical jungle is to test our ideas in all possible cases by analyzing a sufficient number of different examples. If we do not use this method, we risk losing ourselves in a labyrinth of empty if not fallacious abstractions.

Oriented by the two above-considered methodological principles, I intend to perform tasks. The first one is to revive some old and unjustly dismissed philosophical ideas, like descriptivism, the role of facts as the only proper truthmakers, the view of existence as a higher-order property, the verificationist view of meaning, the correspondence theory of truth… The second is to offer some linguistic criticism. I intend to show that the most positive and challenging theses of the metaphysics of reference – even if original and illuminating – are no more than sophisticated conceptual illusions.

7. Tacit knowledge of meaning: traditional explanation
I will assume the practically indisputable notion that language is a system of signs basically governed by conventionally grounded rules, including semantic ones. Linguistic conventions are rules obeyed by most participants in the linguistic community. These participants expect other participants to comply with similar or complementary rules and vice-versa, even if they aren’t really aware of them (cf. Grice 1989, Ch. 2; Lewis 2002: 42). According to this view, the sufficiently shared character of language conventions is what makes possible the use of language to communicate thoughts.
  One of the most fundamental assumptions of the old orthodoxy in philosophy of language is that we lack awareness of the effective structures of semantically relevant rules governing the uses of our language’s most central conceptual expressions. We know how to apply the rules, but the rules are not available for explicit examination. Thus, we are unable to command a clear view of the complex network of tacit agreements involved. One reason is the way we learn expressions in our language. Wittgenstein noted that we learn the rules governing our linguistic expressions by training (Abrichtung), that is, through informal practice, imitation and correction by others who already know how to use the words properly. Later analytic philosophers, from Gilbert Ryle to P. F. Strawson, Michael Dummett and Ernst Tugendhat, have always insisted that we do not learn the semantically relevant conventions of our language (i.e., the semantic-cognitive rules determining referential use of expressions) through verbal definitions, but rather in non-reflexive, unconscious ways. Tugendhat wrote that we learn many of these rules in childhood through ostension by means of positive and negative examples given in interpersonal contexts: other speakers confirm them when correct and disconfirm them when incorrect. Hence, the final proof that we understand these rules is interpersonal confirmation of their correct application. (Tugendhat & Wolf 1983: 140) For this reason, it is often so hard if not impossible to obtain an explicit verbal analysis of the meaning of an expression that reveals its meaning-rules. Using Gilbert Ryle’s terms, with regard to these meaning-rules we have knowing how, skill, competence, an automatized ability that enables us to apply them correctly; but this is insufficient to warrant knowing that, namely, the capacity to report what we mean verbally (1990: 28 f.).
  This non-reflexive learning of semantic rules applies particularly to philosophical terms like ‘knowledge,’ ‘consciousness,’ ‘understanding,’ ‘perception,’ ‘causality,’ ‘action,’ ‘free will,’ ‘goodness,’ ‘beauty,’ which are central to our understanding of the world (Tugendhat 1992: 268). Because of their more complex conceptual structure and internal relationships with other central concepts, these concepts are particularly elusive and resistant to analysis. This insight certainly also applies to conceptual words from philosophy of language, like ‘meaning,’ ‘reference,’ ‘existence’ and ‘truth,’ which will be examined later in this book. Finally, complicating things still more, relevant concepts are also in some sense empirically grounded and not completely immune to additions and changes resulting from the growth of our knowledge. For instance: until recent advances in neuroscience, bodily movement was considered essential to the philosophical analysis of the concept of action. Now, with sensitive devices able to respond to electrical discharges in our motor-cortex, we can even move external objects using sheer willpower. Intentions unaided by bodily movements are now sufficient to produce external physical motions intended by the agent (see neuroprosthetics and BCIs).
  However, lack of semantic awareness can become a reason for serious intellectual confusion when philosophers try to explain what these terms mean. Philosophers are very often under the pressure of some generalizing purpose extrinsic to that required by the proper nature of their object of investigation. Consider theistic purposes in the Middle Ages and scientist purposes in our time, which can easily produce startling but erroneous magnifications hinged on minor real findings. Wittgenstein repeatedly these metaphilosophical views throughout his entire philosophical career. Here are some of his best quotes, in chronological order, beginning with his Tractatus Logico-Philosophicus and ending with his Philosophical Investigations:

Natural language is part of the human organism and not less complicated than it. ... The conventions that are implicit for the understanding of natural language are enormously complicated (Wittgenstein 1984g, sec. 4.002).
  Philosophers constantly see the method of science before their eyes, and are irresistibly tempted to ask and answer questions the way science does. This tendency is the real source of metaphysics, and leads the philosopher into complete darkness. (1958: 24)
  We can solve the problems not by giving new information, but by arranging what we have always known. Philosophy is a battle against the bewitchment of our intellect by language (Wittgenstein 1984c sec. 109).

  The aspects of things that are most important for us are hidden because of their simplicity and familiarity. (One is unable to notice something – because it is always before one’s eyes.) The real foundations of his enquiry do not strike a person at all. Unless that fact has at some time struck him. – And this means: we fail to be struck by what, once seen, is most striking and most powerful. (Wittgenstein 1984c, sec.129).
  Contrary to empirical statements, rules of grammar describe how we use words in order to both justify and criticize our particular utterances. But as opposed to grammar book rules, they are not idealized as an external system to be conformed to. Moreover, they are not appealed to explicitly in any formulation, but are used in cases of philosophical perplexity to clarify where language misleads us into false illusions … (A whole cloud of philosophy is condensed into a drop of grammar.) (Wittgenstein 1984c, II xi).

Around the mid-twentieth century, a number of analytical philosophers were in significant ways directly or indirectly influenced by Wittgenstein. They thought clarification resulting from the work of making explicit the tacit conventions that give meaning to our natural language was a kind of revolutionary procedure: We should identify most if not all philosophical problems with conceptual problems that could be solved (or dissolved) by means of conceptual analysis.
  Notwithstanding, except for the acquisition of new formal analytical instruments and a new pragmatic concern leading to more rigorous and systematic attention to the subtleties of linguistic interaction, there was nothing truly revolutionary in the philosophy of linguistic analysis and the critique of language associated with it. Analysis of the meaning of philosophically relevant terms as an attempt to describe the real structure of our thinking about the world is no more than the resumption of a project centrally present in the whole history of Occidental philosophy. Augustine wrote: ‘What, then, is time? If no one asks me, I know; if I wish to explain it to him who asks, I know not.’ (Augustine, 2008, lib. XI, Ch. XIV, sec. 17). In fact, we find the same concern already voiced by Plato. If we examine questions posed in Plato’s Socratic dialogues, they all have the form ‘What is X?’, where X takes the place of philosophically relevant conceptual words like ‘temperance,’ ‘justice,’ ‘virtue,’ ‘love,’ ‘knowledge’… What always follows are attempts to find a definition able to resist objections and counterexamples. After some real progress, discussion usually ends in an aporetic way due to merciless conceptual criticism. Philosophy based on analysis of conceptual meaning has always been with us. It is the main foundation of our philosophical tradition, even when it is hidden under its most systematic and speculative forms.[11]
  Finally, by defending the view that philosophy’s main job is to analyze implicit conceptual knowledge I am not claiming that philosophy in this case cannot be about the world, as some have objected (Magee 1999, Ch. 23). Even if through our conceptual network, philosophy continues to be about the world because the concepts analyzed by philosophy are in one way or the other about the world. Moreover, in a systematic philosophical work central concepts of our understanding of the world are analysed in their internal relations with other central concepts, with the same result that philosophy is indirectly also about the world – about the world as it is synthetically reflected by the central core of our conceptual network.[12] Indeed, even if the philosophical analysis of our conceptual structures does not depend on empirical experience as such, empirical experience has already in one way or the other entered in the production and change of such conceptual structures.

8. A very simple model of a semantic-cognitive rule
We urgently need to clarify the form of semantic-cognitive rules as it is meant here. However, it is not very helpful if we begin by attempting to analyze a conceptual rule constitutive of a philosophical concept-word. Not only are these concept-words usually polysemic, but the structures of central meaning-rules expressed by them are much more complex and harder to analyze and in this way to elucidate or define. Anyway, although philosophical definitions can be extremely difficult to achieve, the skeptical conclusion that they are impossible can well be too hasty.
  To get a glimpse into a semantic-cognitive rule, I strategically begin with a very trivial concept-word that can be used as a model, since its logical grammar is correspondingly easier to grasp. Thus, I wish to scrutinize here the standard nominal meaning of the concept-word ‘chair,’ using it as a simple model that can illustrate my approach to investigating the much more complicated philosophical concepts that shape our understanding of the world. We all know the meaning of the word ‘chair,’ though it would be not so easy to give a precise definition if someone asked for one. Now, following Wittgenstein’s motto, according to which ‘the meaning of a word is what the explanation of its meaning explains’ (1984g, sec. 32), I offer a perfectly reasonable definition (explanation) of the meaning of the word ‘chair.’ You can even find something not far from it in the best dictionaries. This definition expresses the characterizing ascription rule of this concept-word, which is the following:

(C) Chair (Df.) = a moveable seat with a backrest, designed for use by only one person at a time (it usually has four legs, sometimes has armrests, is sometimes upholstered, etc.).[13]

In this definition, the conditions stated outside of parentheses are necessary and together sufficient: a chair must be a seat with a backrest designed for a single person. These criterial conditions form an essential (indispensable) condition, also called the definitional or primary criterion for the applicability of the concept-word, to use Wittgenstein’s terminology. What follows in parentheses are complementary (dispensable) secondary criteria or symptoms: usually a chair has four legs, often it has armrests, and sometimes it is upholstered. These indications can be helpful in identifying chairs, even though they are irrelevant if the definitional criterion isn’t satisfied. A chair need not have armrests, but there cannot be a chair with armrests but no backrest (this would be a bench). Thus, with (C) we have an expression of the conventional ascription rule for the general term ‘chair,’ which should belong to the domain of what Frege calls sense (Sinn).[14]
  I find it hard to oppose this definition. Table-chairs, armchairs, easy chairs, rocking chairs, wheelchairs, beach chairs, kneeling chairs, electric chairs, thrones… all conform to the definition. Car, bus and airplane seats are not called ‘chairs’ because they are made to be fixed inside a proper place and in this way are not free to be moved, though they are quasi-chairs. It can be difficult to remove electric chairs and thrones from their places, but it is not impossible. Moreover, we can always imagine borderline cases. There could be a seat whose backrest is only 20 cm. high (is it a stool or a chair?), a chair with a seat raised only 10 cm. above the floor (is it even a seat?), a chair whose backrest was removed for some hours (did it become a backless chair or provisionally a stool?). Suppose we find a tree trunk in a forest with the form of a chair that, with some minor carving and painting, is now being used as a chair (it was not manufactured as a chair, but minor changes turned it into something we could call a real chair, depending on the relevance of the changes). Nevertheless, our definition is still reasonable despite vague borderline cases. Empirical concepts all have some degree of vagueness, and one can even argue that vagueness is a metaphysical property of reality. Indeed, if our definition of a chair had overly sharp boundaries, it would be even inadequate, since it would not reflect the desired flexibility of application belonging of our normal concept-word ‘chair,’ tending to diminish the extension of the concept. An often overlooked point is that what really justifies a semantic-cognitive rule is its practical applicability to common cases. That is, what really matters are cases to which we can apply the ascription rule without much hesitation and not those rare borderline cases where we cannot know if the ascription rule is definitely applicable, since the rarity of these cases makes them irrelevant from a practical point of view. Accordingly, the function of a concept-word is far from being discredited by a few borderline cases where we are at a loss to decide whether it is still applicable.
  Furthermore, we need to distinguish real chairs from ‘so-called chairs,’ because in such cases we are making an extended or even a metaphorical use of the word. A child’s toy chair, like a sculptured chair, is a chair in an extended sense of the word. In Victor Hugo’s novel Toilers of the Sea, the main character ends his life by sitting on a ‘chair of rock’ on the seashore, waiting to be swept away by the tide... But it is clear from our definition that this use of the word is metaphorical: a real chair must be made by someone, since it is an artifact; but the immoveable stone chair was only a natural object accidentally shaped by erosion into the rough form of a chair and then used as a chair.
  There are also cases that only seem to contradict the definition, but that on closer examination do not. Consider the following two cases, already presented as supposed counterexamples (Elbourne 2011, Ch. 1). The first is the case of a possible world where some people are extremely obese and sedentary. They require chairs that on the Earth would be wide enough to accommodate two or three average persons. Are they benches? The relevant difference between a bench and a chair is that chairs are artifacts made for only one person to sit on, while benches are wide enough for more than one person to sit on at a time. Hence, in this possible world what for us look like benches are in fact chairs, since they are constructed for only one sitter at a time. If these chairs were ‘beamed’ over to our world, we would say that they remained chairs, since the makers intended them to be chairs, even if we used them as benches. The second counterexample is that of a social club with a rule that only one person at a time can use each bench in its garden. In this case, we would say they continue to be benches and not chairs, since they are still artifacts designed for more than one person to sit on, even if they are now limited to single sitters. Elbourne also asked if a chair must have four legs. Surely, this would be a rough mistake, since according to our definition having four legs isn’t a defining feature: there are chairs with no legs, like an armchair, chairs with three legs, and we can imagine a chair with a thousand legs. The property of having four legs is what we have called a symptom or a secondary criterion of ‘chair-ness,’ only implying that a randomly chosen chair will probably have four legs.
  One can always imagine new and more problematic cases that do not seem to fit the definition, but if we look at the definition more carefully we discover that the difficulty is only apparent or that these ‘exceptions’ are borderline cases or that they are extensions or metaphors, or even that the definition indeed deserves some refinement, remembering that refinement isn’t a change to something other.
  Finally, the boundaries of what we call a ‘chair’ can also undergo changes from language to language and over time; in French an armchair (easy chair) is called a ‘fauteuil’ in contrast to a ‘chaise’ (chair), though a French speaker would agree that it is a kind of chair. I suspect that thousands of years ago, in most societies one could not linguistically distinguish a stool from a chair, since a seat with a backrest was a rare piece of furniture until some centuries ago.

9. Criteria versus symptoms
To make things clearer, it is already worthwhile to broaden our consideration of Wittgenstein’s distinction between criteria and symptoms. A symptom or a secondary criterion is an entity E that – assuming it is really given – only makes our cognitive awareness A of E more or less probable. In contrast, a definitional or primary criterion is an entity E (usually appearing as a complex criterial configuration) that – assuming it is really given – makes our cognitive awareness A of E beyond reasonable doubt (Wittgenstein 1958: 24; 2001: 28).[15]
  For instance, if we assume I can see four chair legs under a table, this is a symptom of a chair, since it greatly increases the probability that a chair is behind the table. But if we assume that what is visually given to me is ‘a moveable seat with a backrest made for only one person to sit on,’ this puts my cognitive awareness of a chair beyond doubt. The definition (C) expresses a definitional criterion, understood as such because its assumed satisfaction leaves no possibility to doubt that we can apply the ascription rule for the concept-word ‘chair.’
  We cannot guarantee with absolute certainty that entity E (criterion or symptom) is ‘really given’ because I accept that the products of human experience are inevitably fallible. Nonetheless, using this ‘assumed given-ness’ based on experience and an adequate informational background, we can find a probability when a symptom is satisfied and a practical certainty when a criterion is satisfied. In this last case, we claim there is a probability so close to 1 that we can ignore the possibility of error in the cognitive awareness A that entity E is given. (Correspondingly, one could also speak in this sense of a presumed necessity.)
  Symptoms or secondary criteria can help us identify entity E using cognitive awareness A, even if we cannot regard E as necessary. However, symptoms are of no use unless definitional criteria are also met. Four legs and armrests that do not belong to a chair would never make a chair.[16]
  Terms like ‘criteria’ and ‘symptoms,’ as much as ‘conditions’ have so-called process-product ambiguity. We can see them as (a) elements of the rule that identifies what is given, but we can also see them as (b) something satisfying the rule that is really given in the world. Our semantic-cognitive rules are also criterial rules, able with the help of imagination to generate criterial configurations belonging to them internally as (a). Hence, we could say that definition (C) is the expression of a semantic-criterial rule with the form: ‘If we accept that E is really given, we must conclude A,’ where the conclusion A is our awareness with practical certainty that E is given.
  One problem here is to know what this awareness means. My suggestion will be that we can equate this cognitive awareness with our acceptance of the existence and applicability of a network of external inferential relations once a semantic-cognitive rule is satisfied. The concept of chair, for instance, consists of internal relations expressed by a definitional rule (C). But our awareness of the application of this concept arises as a maze of external relations resulting from the satisfaction of (C). For example, if I am aware that a chair exists, I can infer that it has a particular location, that I can sit on it or ask someone to sit on it, that I could possibly damage it, borrow it, loan it, etc. I can do this even if I have no real consciousness of the structure of the rule I applied to identify the chair.

10. Challenges to the traditional explanation (i): John McDowell
Supporters of semantic externalism have challenged the idea that the meanings of expressions consist in our implicit knowledge of their constitutive rules or conventions. According to their view, the meanings of expressions are predominantly related to physical and social-behavioral worlds, depending in this way only on objects of reference and supposedly also on neurobiological processes involving autonomous causal mechanisms. In this context, there is little room for discussing the conventionality of meaning.
  As evidence for the externalist view, we can adduce our lack of awareness of the structure of semantic rules determining the linguistic uses of our words. If we lack awareness of senses or meanings, it might be that they could as meanings be instantiated to a greater or lesser extent in a non-psychological domain. If this is so, in principle cognitive (also called pre-cognitive) participation in meaning could be unnecessary. Meaning could result solely from autonomous causal mechanisms, not recoverable by consciousness. In opposition to Michael Dummett’s ‘rich’ view of implicit meaning, John McDowell illustrated the externalist position on the referential mechanism of proper names, observing that:

We can have the ability to say that a seen object is the owner of a familiar name without having any idea of how we recognize it. The assumed mechanisms of recognizing can be neural machinery [and not psychological machinery] – and its operations totally unknown to whoever possesses them. (McDowell 2001: 178)[17]

Some pages later, McDowell (following Kripke) asserts that the referential function of proper names would not be explained by conventionally based implicit identification rules that can be descriptively recovered, because:

The opinions of speakers on their divergent evidential susceptibilities regarding names are products of self-observation, as much as this is accessible, from an external point of view. They are not intimations coming from the interior, from a normative theory implicitly known, a receipt for the correct discourse which guides the behaviour of the competent linguist. (McDowell 2001: 190)

This view is in direct opposition to the one I defend in this book, not because it never can be justified, but because it isn’t the standard case. In what follows, I intend to show that usually the implicit application of internal semantic-cognitive rules based on criteria is indispensable for the referential function. Moreover, we have already begun to see that to have reference, a usually tacit and unconscious cognitive element must be associated with our expressions and should be instantiated at least in some measure and at some moment in the language user’s head. For in no case is this clearer than with McDowell’s main focus: proper names (see my Appendix to Chapter I).
  Here is how we could argue against McDowell’s view. If he were correct, an opinion about the given criterial evidence for the application of a proper name found through external observation of our referring behavior should be gradually reinforced by the cumulative consideration of new examples, that is, inductively. Even repetition of the same example would be inductively reinforcing! However, this is far from the case. Consider our characterizing semantic-cognitive rule (C) for applying the concept-word ‘chair.’ We can see from the start that (C) seems correct. We naturally tend to agree with (C), even if we have never considered any examples of the word’s application. And this shows that speakers are indeed only confirming a recipe for correct application that comes from inside, as a matter of tacit agreement among speakers… Admittedly, after we hear this definition, we can test it. Thus, we can imagine a chair without a backrest but see that it is really a stool, which isn’t properly a chair. If we try to imagine a chair designed so that more than one person can sit on it, we will conclude that we should call it a sofa or a garden bench... We can understand supposed counterexamples only as means to confirm and possibly correct or improve the definition, thereby discovering its extensional adequacy in a non-inductive way. This specification of meaning seems to be simply a contemporary formulation of something Plato identified as reminiscence (anamnesis): the recalling to mind of his ideas. We do not need to go beyond this, imagining all sorts of chairs (rocking chairs, armchairs, wheelchairs…) in order to reinforce our belief in the basic correctness of our intuitive definition.
  Now consider the same issue from McDowell’s perspective. Suppose he is right and our knowledge of the meaning of a common name like ‘chair’ were the result of self-observation from an external viewpoint. We could surely acquire more certainty that chairs are seats with backrests made for one person to sit on by observing the similarities among real chairs that we can see, remember or imagine. Inductively, the results would then be increasingly reinforced, possibly by agreement among observers about an increasing number of examples. As we already noted, even examples of people reaching shared agreement by singling out thousands of identical classroom chairs would not enable us to increase our conviction that we have the factually true evidential conditions for applying the concept-word ‘chair.’ Moreover, it is clear that one does not need much reflection to recognize that the idea is absurd of definition (C) capturing a neuronal mechanism and not resulting from an implicit shared agreement. Furthermore, I am sorry to say, the explanation of the implicitly conventional identification rule for the proper name Aristotle investigated in the Appendix of the last chapter is sufficient to make this whole discussion idle.
  We conclude, therefore, that the ascription rule made explicit in definition (C) does in fact have the function of rescuing for consciousness the tacit convention governing the referential use of the word ‘chair’ (as with our earlier definition of Aristotle). It seems from the start intuitive and may only require the help of confirmatory, corrective and improving examples. And what is true for a general term should presumably also be true for other expressions, as we already saw regarding proper names.
  Indeed, if all we have in these cases is a shared convention, then a psychological element needs to be involved, even if only in an implicit way, constituting what could be called a non-reflexive cognitive application of the rule. Definition (C) makes explicit a convention normally instantiated in our heads as an (implicit) non-reflexive application, whenever we make conscious use of the word ‘chair,’ which only confirms the traditional standard explanation.

11. Challenges to the traditional explanation (ii): Gareth Evans
There is another argument against the claim that we have tacit cognitive access to semantic conventions that govern our use of expressions. This argument comes from the philosopher Gareth Evans, who directly influenced McDowell. Evans invites us to contrast a person’s belief that a substance is poisonous with a mouse’s disposition not to consume it. In the case of a human being, it is a genuine belief involving propositional knowledge; in the case of a mouse, it is a simple instinctive disposition to react in a certain way to a certain smell, not a true belief. Proof of the difference is the fact that:

It is of the essence of a belief state that it be at the service of many distinct projects, and that its influence on any project is mediated by other beliefs. (Evans 1985: 337).

If someone believes a certain substance is poisonous, he can do many different things based on that belief. He can test his belief by feeding it to a mouse, or if he is depressed, he can try to commit suicide by swallowing a dose. He can also relate his belief that the substance is poisonous to a variety of other beliefs. For instance, he might believe he will become immune to a poison by consuming small amounts every day, gradually increasing the dose... As our knowledge of semantic rules is not susceptible to such inferences, thinks Evans, it consists not of actual belief states, but rather of isolated states, not very different from those of the mouse. Therefore, they are not cognitive (or pre-cognitive) psychological states in a proper sense of the word. (Evans 1985: 339)
  The characterization of belief proposed by Evans is really interesting and in my view correct, but his conclusion does not follow. Certainly, it agrees with many of our theories of consciousness, according to which a belief is only conscious if it isn’t insular, while an unconscious belief is insular – though there are degrees of insularity. But the crucial point is that Evans’ argument blinds us to the vast gulf between our semantic uses of language and the mouse’s behavioral disposition to avoid consuming poison.
  As a weak but already useful analogy, consider our knowledge of simple English grammar rules. A child can learn to apply these rules correctly without any awareness of doing so; and some adults who have never learned formal grammar are still able to apply these rules correctly to many different words in many different contexts. Moreover, even if our knowledge of these grammar rules is very often pre-conscious, with sufficiently careful examination we can bring them often to consciousness.
  The problem becomes still clearer when we consider our simple example of an implicit semantic-cognitive rule, the criterial rule (C) for the application of the concept-word ‘chair’ to the identification of chairs. Certainly, a person can derive many conclusions from this rule. He can predict that normally five persons cannot sit side-by-side on a single chair. He knows that one can transform a chair into a stool simply by cutting off its backrest. He can know the price and if he would like to buy a similar chair. He knows that by standing on a chair, he can reach an overhead ceiling lamp… He knows all this and much more, even without having ever consciously considered definition (C). And this only means that we can have a belief state enabling us to identify chairs, putting it at the service of many different projects mediated by other beliefs without being explicitly aware of the involved meaning-rule (C).
  We can see a continuum, beginning with more primitive and instinctively determined dispositions and ending with semantic-cognitive rules of our language and their effects. It includes dispositions like those of mice, which cannot be cognitive, because they are instinctive (it is utterly implausible to think that a mouse could be reflexively conscious). There are also more sophisticated ones, like our unconscious beliefs, thoughts and cognitions, which we can consciously scan and reflexively access (presumably through meta-cognitive processes).
  If we accept the view that our semantic rules are usually conventional rules exemplified in the simplest cases by models like (C), then we must reject the radicalism of positions such as those of Evans and McDowell. After all, the application of such rules allows us to make many different inferences and relate them to many other conceptual rules. Rule (C) has greater proximity to the rules of English grammar than to the innate dispositional regularities demonstrated by a mouse that instinctively avoids foods with certain odors. Moreover, it is clear that in such cases, unlike the mouse, for people inferences to other beliefs are always avaliable. This can be so even if we admit that our semantic-cognitive rules do not in themselves possess the widest availability proper of those completely conscious belief states considered by Evans.[18]
  The root of the confusion is that the semantic rules in question, with and because of their apparent triviality, have not yet been investigated in a sufficiently systematic way. In an academic world dominated by science, the procedure that leads to their discovery does not seem worthy of careful investigation. Nevertheless, to proceed more systematically in this seemingly trivial direction is in fact philosophically invaluable, and this is what I will do in the remainder of this book.

12. Non-reflexive semantic cognitions
I believe contemporary theories of consciousness support the traditional view according to which we have implicit knowledge of our meaning-rules. I will begin by appealing to reflexive theories of consciousness. But first, what are these theories?
  In the philosophical tradition, the idea of reflexive consciousness was already suggested by John Locke with his theory of internal sense (Locke 1690, book II, Ch. 1, §19). Reflexive theories of consciousness were introduced to the contemporary discussion by D. M. Armstrong (Armstrong 1981: 55-67; 1999: 111 f.). We can summarize Armstrong’s view as saying there are at least two central meanings of the word ‘consciousness.’ The first is what he calls perceptual consciousness, which consists in the organism being awake, perceiving objects around it and its own body. This is the simplest sense of consciousness. John Searle wrote that consciousness consists of those subjective states of sentience or awareness that begin when one wakes up in the morning after deep, dreamless sleep and continue throughout the day until one falls asleep at night, or lapses into a coma, or even dies (Searle 2002: 7). By this he meant chiefly perceptual consciousness. This is also a very wide and consequenty not so distinctive sense of consciousness, since less developed species also have it. For instance, we can say that a hamster sedated with ether loses consciousness, because it ceases to perceive itself and the world around it. It seems justified to assume that when a hamster is awake, it has some primitive form of cognition of the world around it, as shown by its behavior. However, the width of this extension only suggests the irrelevance of perceptual consciousness for us. We are aware of the world in the same way a hamster seems to be conscious of it, but in a much more demanding, more human sense of the word. Certainly, a mouse perceives a cat, but it is unlikely to know it is facing its archenemy. This also holds for internal feelings. A snake may be able to feel anger; but we hardly believe a snake is aware of this anger, since it certainly has no reflexive consciousness.
  Now, what distinguishes a mouse’s perceptual awareness and a snake’s anger from our own conscious awareness of things around us and from our own feelings of anger? The answer is given by a second sense of the word ‘consciousness’ which Armstrong considers the truly important one. This is what he termed introspective consciousness and that I prefer (following Locke) to call reflexive consciousness: This is a form of consciousness that we can define as reflexive awareness of our own mental states.
  According to one of Armstrong’s most interesting hypotheses, reflexive conscious­ness emerges from the evolutionary need of more complex systems to gain control of their own mental processes by means of higher-order mental processing. In other words: our first-order mental events, like sensations, feelings, desires, thoughts, and even our perceptual consciousness of the world around us, can become objects of simultaneous introspections with similar content (D. M. Rosenthal called these meta-cognitions higher-order thoughts[19]).
  According to this view, only when we achieve reflexive consciousness of a perceptual state can we say that this state ‘becomes conscious’ in the strong sense of the word. So, when we say in ordinary speech that a sensation, a perception, a sentiment or a thought that we have ‘is conscious,’ what we mean is that we have what could be called a meta-cognition of it. This shows that Armstrong’s perceptual consciousness is actually a kind of unconscious awareness, while reflexive consciousness – the true form of consciousness – is probably a faculty possessed only by humans and a few higher primates such as orangutans.[20]
  Now, let us apply this view to our tacit knowledge of semantic-cognitive rules. It is easy to suggest that we usually apply these rules without having a meta-cognitive consciousness of them and therefore without making ourselves able to consciously scrutinize their structure. In other words, we apply these rules to their objects cognitively[21], and these rules are ‘cognitive’ because they generate awareness of the objects of their application. But in themselves these rules usually remain unknown, belonging to what I above called unconscious awareness. Hence, it seems that we need to resort to some kind of meta-cognitive scrutiny of our semantic-cognitive rules in order to gain conscious awareness of their content.
  One objection to using this kind of theory to elucidate tacit knowledge of our rules is that there are a number of interesting first-order theories of consciousness that do not appeal to the requirement of higher-order cognition. In my view, we can classify most, if not all, of these apparently competing theories as integrationist theories of consciousness. We can do this, because they share the idea that consciousness of a mental state depends on its degree of integration with other mental states constituting the system. This is certainly the case of Daniel Dennett’s theory, according to which consciousness is ‘brain celebrity’: the propagation of ephemerally fixed contents influencing the whole system (Dennett 1993, Ch. 5). This is also the case with Ned Block’s view, according to which consciousness is the availability of a mental state for use in reasoning and directing action (Block 1995: 227-47). This is likewise the case of Bernard Baars’ theory of consciousness as the transmission of content in the spotlight of attention to the global workspace of the mind (Baars 1997). And it is also the obvious case of Giulio Tononi’s theory, according to which consciousness arises from the brain’s incredible capacity to integrate information (Tononi 2004: 5-42). These are only some well-known contemporary first-order theories of consciousness that are historically consonant with Kant’s view. According to him, to be consciously recognized, a mental state must be able to be unified (integrated) into a single Self. From the perspective of such integrationist theories, an unconscious mental state would be one that remains to a greater or lesser extent dissociated from other mental states. And all these views seem to possess a degree of reasonability.
  The objection, therefore, would be that I am trying to explain implicit knowledge of language by relying solely on meta-cognitive theories of consciousness, ignoring all others. However, I believe there is more than one way around this objection. My preferred way is the following: we have no good reason to think integrationist and reflexive views of consciousness are incompatible. After all, it makes sense to think that a mental state’s property of being the object of meta-cognition also seems to be a condition – perhaps even a necessary one – for the first-order mental state to be more widely available and more easily integrated with other elements constituting the system. As Robert Van Gulick wrote in the conclusion of his article on consciousness:

There is unlikely to be any theoretical perspective that suffices for explaining all the features of consciousness that we wish to understand. Thus a synthetic and pluralistic approach may provide the best road to future progress. (Stanford Encyclopedy of Philosophy 2014)

Indeed, we can reinforce our suspicion by reconsidering a well-known metaphor developed by Baars: A conscious state of a mind is like an actor on stage who becomes visible and therefore influential for the whole system because he is illuminated by the spotlight of attention. However, it seems reasonable to think that this could happen only because some sort of searchlight of the will added to some sort of meta-cognitive mental state provides the light for this spotlight. Hence, one could easily argue that the first-order mental state is accessible to the rest of the system and hence conscious due to its privileged selection by some kind of supposedly metacognitive state of attention.
  My conclusion is that our awareness of semantic-cognitive rules and the possibility of scrutinizing them metacognitivey is able to resist integrationist theories, since they all leave room for conscious processes able to be scrutinized by means of reflexive attention. Consequently, assumig some kind of reflexive plus integrationist view, the plausible conclusion remains that we can have some kind of cognitive states that make us conscious of their objects even if they are not in themselves objects of consciousness. Thus, it seems plausible that only if we first order processes to (reflexive, metacognitive) scrutiny of attention we can subject them to conscious analysis. And most of our semantic-cognitive rules belong to such cases
  It seems to me that this assumption could explain why we can have unconscious or implicit tacit cognitions when we consciously follow semantic-cognitive rules without being cognitively aware of the content of these rules and consequently without being able to analyze them. They remain implicit because we rarely pay attention to these rules when we apply them and because even when this occurs, they are not there as objects of reflexive cognition. These rules are there, using a well-worn metaphor, like spectacles. When seeing things through them, we are normally unaware of the lenses and their frame. Assuming these views, we conclude that we can distinguish two forms of cognition:

(i) Non-reflexive cognition: This is the case with cognitions that are not conscious, because they are not accessed by a higher-order cognitive process and/or focused on by inner attention, etc. (e.g., my perceptual consciousness when I use rule (C) identifying a chair.)
(ii)  Reflexive cognition: This is the case of cognition accessed by a higher-order cognitive process and/or focused on by inner attention, etc., being for this reason able to be the object of conscious access and reflexive scrutiny. Any mental states, sen­sations, emotions, perceptions, and thoughts can be called reflexive if they are accompanied by higher-order cognitive inner attention and/or focused on by inner attention. (This is a previous condition needed for the kind of reflexive scrutiny that can make us aware of the semantic-cognitive rule (C) for the identification of a chair as requiring a seat with a backrest, designed for use by only one person at a time.)

Once in possession of this distinction, we can better understand the implicit or tacit status of the cognitive meanings or contents or semantic rules present in uses we make of expressions. When we say that the structures of semantic-cognitive rules determining the references of our expressions are often implicit (as in the case of the semantic rules defining the words ‘chair’ or ‘Aristotle’), we are not assuming that they are properly pre-cognitive or definitely non-cognitive, lacking any mentality. Nor that they are completely isolated or dissociated from any other mental states (in the last case, we would lack even the ability to choose when to apply them). What we mean is just that the psychological instantiations of these conventional rules are of a non-reflexive type. That is, although consciously used (we know we are using them), they are not likely to be the subject of some form of reflexive cognitive attention. Moreover, as already noted, there is a reason for this, since the structures of these rules are not the focus of our attention when we use the corresponding concept-word in an utterance; it is so because our real concern is much more practical, consisting primarily in the cognitive effects of applying these rules.
  As an obvious example: if I say, ‘Please, bring me a chair,’ I don’t need to explain this by saying, ‘Please, bring me a moveable seat with a backrest, made to be used by only one person at a time.’ This would be discursively obnoxious and pragmatically counterproductive: it would be almost impossible to communicate efficiently if we had to spell out (or even think of) all such details each time we applied semantic-cognitive rules. What interests us is not the tool, but its application – in this case, to inform my hearer that I would like him to bring me a chair. In linguistic praxis, meaning isn’t there to be scrutinized, but instead to be put to work.
  A consequence of this view is that in principle our inner attention must be able to focus on non-reflexive semantic-cognitive rules involved in normal uses of words and scrutinize them meta-cognitively by considering examples of their application or lack of application. Taking into consideration the variable functions and complexity of our semantic-cognitive rules enables the philosopher to decompose them analytically into more or less precise characterizations. It seems it is by this mechanism, mainly helped by examples, counterexamples, comparisons and reasoning that we become aware of the conceptual structure of our philosophically relevant expressions.

13. Conclusion
Summarizing this chapter, we can say that we have found two main methodological devices: (A) the primacy of established knowledge and (B) the method of philosophizing by examples. We will use them as guides in this book’s analyses. Particularly relevant in this context is the idea that we can still see philosophy as an analytical search for non-reductive surveillable representations of our natural language’s central meaning-rules. It is almost surprising to verify that more than two-thousand years after Plato we still have reason to accept the view that solving some of our most intriguing philosophical problems would require deeper and better analyzed explanations of what some central common words truly mean.





[1] See, for instance, the justification of the external world summarized in Ch. VI, sec. 28.
[2] This is a statement like that by Heraclitus of Ephesus, who noted that, ‘The sun is the width of a human foot.’ We need only lie on the ground and hold up a foot against the sun to see that this is true.
[3] I am unable to find real exceptions. Under normal circumstances fire has always burned. Some say that the idea that trees draw energy from the earth was once a commonsense truth until photosynthesis was discovered… But this idea wasn’t a very basic or modest commonsense truth, since it could easily be refuted by the well-known fact that trees do not grow in complete darkness. The idea that a new sun crosses the sky each new day is surely absurd – but is it a commonsense idea? In fact it was suggested by a philosopher, Heracleitus, going beyond the humble intentions of modest common sense. Modest, humble common sense is not interested in answering such questions, which have no relationship to ordinary life concerns.
[4] Roberto DaMatta, in an interview. (A more forceful example is the obstinate rejection of any kind of theism of the Pirahã tribe in the Amazon rainforest studied by Daniel L. Everett).
[5] It was certainly much easier to believe in the existence of a personal God and an eternal soul independent of the body a thousand years ago, before the steady accumulation of conflicting knowledge discovered by the natural and human sciences.
[6] The expression ‘descriptive metaphysics’ was introduced by P. F. Strawson in contrast to ‘revisionary metaphysics.’ It aims to describe the most general features of our actual conceptual schema, while revisionary metaphysics attempts to provide new schema to understand the world. Strawson, Aristotle and Kant developed descriptive metaphysics, while Leibniz and Berkeley developed revisionary metaphysics (Strawson 1991: 9-10).
[7] As these interpreters wrote: ‘Wittgenstein’s objection to “theorizing” in philosophy is an objection to assimilating philosophy, whether in method or product, to a theoretical (super-physical) science. But if thoroughgoing refutation of idealism, solipsism or behaviorism involves a theoretical endeavor, Wittgenstein engages in it.’ (Baker & Hacker 1980: 489). Anthony Kenny (1986) preferred to think that Wittgenstein actually held two competing views on the nature of philosophy – therapeutic and theoretical. But the here proposed unified interpretation seems more charitable.
[8] As he writes, ‘We have now a theory, a “dynamic” theory (Freud speaks of a “dynamic” theory of dreams) of the sentence, of the language, but it appears to us not as a theory.’ (Zettel 1983b: 444).
[9] Well aware of this, Karl Popper famously called the statement ‘All swans are white’ a theory, adding that this theory was falsified by the discovery of black swans in Australia…
[10] Paul Grice’s sophisticated and ingenious work contains an influential (albeit qualified) criticism of ordinary language philosophy as practicized by Ryle, Austin and Strawson (1989, Chs. 1, 2, 10, 15, 17). According to him, these philosophers often confused ordinary uses of statements resulting from conversational implicatures with their literal meaning. When implicature failed, they mistakenly concluded that these statements had no meaning. This would be the case of statements like ‘This flag looks like red’ (supposedly understood by Austin as showing that sense-data do not exist because this statement is devoid of sense), ‘The present King of France is wise’ (understood by Strawson as a statement without truth-value) and ‘If green is yellow then 2 + 2 = 5’ (understood by him as showing the queer character of material implication). I agree with Grice’s rejection of all these ordinary language philosophers’ conclusions, although I remain suspicious regarding Grice’s own explanations. Material implication, for instance, seems to belong to our practice of truth-functional reasoning, which makes explicit a basic general layer subsumed under our more informative factual language. In this sense, it also provides wide intermediate connections. Anyway, under critical scrutiny, I think that natural language intuitions still provide a valuable guide – a point with which Grice would certainly agree.
[11] Philosophers like Berkeley, Leibniz, Hegel, even Heidegger, can be seen as doing revisionary conceptual analysis, refuting and replacing ambitious forms of common sense.
[12] Rudolf Carnap’s formal mode of speech (1937, part 5, sec. A, § 79) instead of material mode of speech, as much as W. V-O. Quine’s broader semantic ascent (1960, Ch. 7, § 56) point to this same fact, namely, that conceptual analysis is also about the world.
[13] If you wish to avoid the word ‘seat’, you can also define a chair as ‘a moveable piece of furniture with a raised surface and a backrest, made for only one person at a time to sit on.’
[14] As will be frequently recalled, I do not deny that referential meanings include things that cannot be easily captured by descriptive conventions, unlike case (C) – things like perceptual images, memory-images, feelings, smells. However, they belong much more to the semantic level called by Frege illuminations (Beleuchtungen), based on natural regularities more than on conventions.
[15] The correct interpretation of this distinction is a controversial issue that does not concern us here; I give what seems to me the most plausible, useful version.
[16] At first view it seems that these logico-concetual remarks appeal to old-fashioned semantic definitions leading us to the rejection of findings of modern empirical psychology (cf. E. Margolis & S. Laurence, 1999, Ch. 1). But this is only appearance. Consider Eleanor Rosch results. She has shown that we are able to categorize under a concept-word much more easily and quickly by appealing to prototypical cases (cf. Rosh, 1999: 189-206). For example, we can more easily recognize a sparrow as a bird than an ostrich or a penguin. In the same way, an ordinary chair with four legs can be recognized as a chair more easily and quickly than can a wheelchair or a throne. However, this does not conflict with our definition, since for us the psychological mechanism of recognition responsible for the performance is not in question, but rather the leading structure subjacent to it. We can often appeal to symptoms as the most usual ways to identify things. For instance, we identify human beings first by their faces and penguins first by their appearance, even if human faces and a penguin’s appearance are only symptoms of what will be confirmed by expected behavior, memories, genetic makeup, etc. Hence, the ultimate criterion remains dependent on a definition. (In one wildlife film counterfeit penguins outfitted with cameras deceived real penguins. The trouble with these moronic birds is that they are overly dependent on innate, instinctive principles of categorization.)
[17] The expression in brackets appears in the author’s footnote on this passage. In Dummett’s more orthodox position, McDowell sees a relapse into the psychologism justifiably rejected by Frege.
[18] Freud distinguished (i) unconscious representation able to associate itself with others in processes of unconscious thought from (ii) unconscious representation that remains truly isolated, unassociated with other representations, which for him would only occur in psychotic states and whose repression mechanism he called exclusion (Verwerfung). Evans treats the relative insularity of our non-reflexive awareness of semantic rules in a way that suggests Freud’s concept of exclusion.
[19] Cf. Rosenthal 2005. In this summary, I will ignore the dispute between theories of higher-order perception (Armstrong, Lycan) and higher-order thought (Rosenthal), and still others. In my view, David Rosenthal is right in noting that Armstrong’s perceptual ‘introspectionist’ model suggests the treatment of cognitions of a higher-order as if they contained qualia, and that it is implausible that higher-order processes have phenomenal qualities. Armstrong, on his side, seems to be right in assigning a causal controlling role to higher-order experience, since for him consciousness arise from the evolutionary necessity to maintain an unfied control upon more complex mental systems. Aside from that, although Armstrong doesn’t use the word ‘thought’, he would probably agree that there is some kind of higher-order cognitive element in the introspection of first-order mental states, an element that interests us here. I prefer the term meta-cognition for these higher-order cognitions, since I believe that not only Rosenthal, but also Armstrong would agree that we are dealing with a cognitive phenomenon. (For initial discussion, see Block, N. O. Flanagan, G. Güzeldere (eds.) 1997, part X.)
[20] I will pass over the traditional idea that of themselves first-order mental states automatically generate meta-cognitions. This view would make it impossible to have perceptual consciousness without introspective consciousness. However, this view not only seems to lack a convincing intuitive basis; it also makes the existence of unconscious thoughts incomprehensible.
[21] Some use the term ‘pre-cognitive’ for what is implicitly known. I use the word ‘cognitive’ in a broader sense, including what is implicitly or pre-consciously known.