Saturday, 25 August 2012

Tacit knowledge and its antonyms


Tacit & Explicit Knowledge characterises tacit knowledge through the antonym ‘explicit’. But this is linked to what is explicable and hence explainable which blurs the distinction between knowledge of worldly processes and those processes, themselves: between sense and reference. Archaeological investigation of the antonym of tacit in Collins’ early work provides a diagnosis: a misunderstanding of Wittgenstein’s rule following considerations.


No one has done more to investigate both the conceptual underpinnings of the genus ‘tacit knowledge’ and potential sub-species of that genus than Harry Collins. I hope that this critical examination of a small area of his work will be taken in the spirit of frank admiration with which it is intended.

Any account of tacit knowledge faces an apparent dilemma which stems from the combination of an immediate question and an attractive methodological approach. The immediate question is ‘how can something be both tacit and knowledge?’ The attractive methodological approach resembles the via negativa in theology: approaching the nature of god by describing what god is not, the (finite and limiting) properties god does not have. In the case at hand, it is characterising what is tacit by selecting a suitable antonym. Tacit knowledge is not explicit, for example, on a suitable understanding of ‘explicit’. The apparent dilemma then stems from how the antonym is selected or understood. Selected so as to focus on the claim that tacit knowledge is tacit, the contrast with the antonym tends to threaten its status as knowledge and vice versa. Hence the seeming dilemma is that an account can justify calling tacit knowledge either ‘tacit’ or ‘knowledge’ but not both.

One might, for example, propose that ineffable knowledge of what an experience is like, in an intransitive sense in which it is not matter of being like another experience, is an instance of tacit knowledge. That might earn the right to a tacit status – as what the experience is like is, on this approach, not something that can be expressed or communicated – but puts under threat, at least, the idea that there is some content known and hence knowledge at all. (It remains a matter of debate whether Frank Jackson’s Mary really does acquire knowledge when she emerges from her black and white room [Jackson 1982].) Responding to that concern by stressing that any genuine case of tacit knowledge has such a content, something with a conceptually articulable structure for example, threatens the claim that it is tacit after all since if it has a conceptual content, why is that not explicit?

Following the lead of Michael Polanyi, most authors worry more about underpinning the tacit status and ignore worries about knowledge. (Polanyi himself talks of tacit knowing rather than knowledge. Whilst this marks an important stress on practical activity, which I endorse, I will ignore his scruple in what follows and talk of tacit knowledge.) He starts his book The Tacit Dimension with the following slogan:

I shall reconsider human knowledge by starting from the fact that we can know more than we can tell. [Polanyi 1967b: 4]

The broad suggestion is that knowledge can be tacit when it is, on some understanding, untellable. Tellable knowledge is a subset of all knowledge and excludes tacit knowledge. But as Polanyi – like many authors since – immediately concedes, the slogan is gnomic. Does it carry, for example, a sotto voce qualification ‘at any one particular time’? Or does it mean: ever? Polanyi continues:

This fact seems obvious enough; but it is not easy to say exactly what it means. Take an example. We know a person’s face, and can recognize it among a thousand, indeed among a million. Yet we usually cannot tell how we recognize a face we know. So most of this knowledge cannot be put into words. [Polanyi 1967b: 4]

The suggestion is that tacit knowledge is tacit – is ‘more than we can tell’ – because we cannot tell how we know it.

Given that ‘tacit knowledge’ is a semi-technical term, a neologism introduced to capture aspects of knowledge that had not been properly attended to – and hence a matter for plausible stipulation rather than description – one exegetical route from this suggestion is to question what arguments Polanyi himself offers for this, connecting eventually with his idea of ‘personal knowledge’ as the ‘active comprehension of things known, an action that requires skill’ [Polanyi 1958: vii] and the idea that denotation itself is an art. But my focus in this paper is with the way that Collins selects antonyms for ‘tacit’ using the methodological approach described above.

I will begin with Tacit and Explicit Knowledge and argue that one of the antonyms Collins selects there undermines the knowledge status of tacit knowledge because it ignores the distinction between mind and world or sense and reference. To understand why he selects this, I will examine the way antonyms selected in his earlier work Changing Order are explicitly influenced by his interpretation of Wittgenstein’s discussion of rules. This interpretation, in turn, sustains the tacit status of his earlier conception of tacit knowledge but the cost of its knowledge status. Finally, I return to show how it is at least plausible that this interpretation motivates Collins’ use of the concept of strings in Tacit and Explicit Knowledge which is what undermines the distinction of mind and world and thus distorts the conception of knowledge in play.

Collins’ recent work and and the antonym of ‘tacit’

In Tacit and Explicit Knowledge, Collins approaches the nature of tacit knowledge in accord with the methodological approach sketched above: through a contrast with what is explicit. He describes this strategy in a pithy summary at the very start of his book: ‘explain “explicit”, then classify tacit.’ [Collins 2010a: 1]. But, as I will argue, his particular interpretation of ‘explicit’ threatens to impale his account on the second horn of the dilemma by undermining its knowledge status.

A clue to the difficulty comes in the first sentence of the first chapter of the book. ‘Tacit knowledge is knowledge that is not explicated’ [ibid: 1 italics added]. This might be terminologically innocent if ‘explicate’ were taken to mean make clear. If so tacit knowledge would be knowledge that is not, or perhaps could not be, made clear, which has echoes of Polanyi’s slogan, we can know more than we can tell. In fact, Collins suggests four distinct meanings of ‘explicable’: elaboration, transformation, mechanization and explanation [ibid: 81]. It is the fourth to which causes trouble.

On that understanding, thus tacit knowledge stands opposed to what can be explained. Even though he allows that this can be a matter of degree, it yields a much stronger claim which puts under threat the knowledge status of tacit knowledge. In this section I will describe this element of Collins’ account before asking, in the rest of the paper, why Collins thinks this.

It will help to contrast two claims Collins makes. He says, on the one hand, that:

[T]he idea of tacit knowledge only makes sense when it is in tension with explicit knowledge, and since cats and dogs and sieves and trees cannot be said to ‘know’ any explicit knowledge, they shouldn’t be said to know any tacit knowledge either. In fact, they don’t ‘know’ anything... [ibid: 78]

This addresses limits on the proper subjects of knowledge. Only those in the space of reasons, as one might say, can have knowledge, whether explicit or tacit. But at the same time, Collins suggests, one of the things that makes the very idea of tacit knowledge seem unduly mysterious is just the separation of rational subjects from the rest of the animal, vegetable and mineral world. Thus it is helpful to his project of demystifying tacit knowledge to suggest the similarities between human and non-human cases.

In all the ways that do not involve the way we intentionally choose to do certain acts and not others, and the way we choose to carry out those acts, the human, per individual body and brain… is continuous with the animal and physical world. We are just like complicated cats, dogs, trees, and sieves… Sometimes we can do things better than cats, dogs, trees and sieves can do them, and sometimes worse. A sieve is generally better at sorting stones than a human (as a fridge is better at chilling water), a tree is certainly better at growing leaves, dogs are better at being affected by strings of smells, and cats are better at hunting small animals… That teaching humans to accomplish even mimeomorphic actions is a complicated business, involving personal contact, says nothing about the nature of the knowledge, per se. [Collins 2010a: 104-5]

So aside from the fact that we can choose to do some things rather than others, and can choose to do them in particular ways, whilst cats, dogs, trees and sieves cannot, the performance of the tasks, which for us is expressive of tacit knowledge, is just the same. In that respect, we are just like those animals, plants and artefacts, according to Collins.

A clue to how Collins addresses the apparent incompatibility between the claims that cats, dogs, trees and sieves know nothing whilst the way they ‘do’ things is just like the way we do things when we use and express our tacit knowledge is his focus on what he calls (in the last quote) the ‘nature of the knowledge, per se’. This does not, however, seem to mean the way humans know how to do the task, their knowledge, after all. Rather, it seems to mean the nature, not of the knowledge, but of the task, or better the process, itself. This is how it can be a common element between humans and non-humans since the same process can be enacted, or brought about, or underpinned by human agency or non-human happening. The cost of this demystifying move is substantial, however. It means that the focus of Collins’ account is not the nature of knowledge, for example of a worldly process, but the worldly process itself. His account is thus pitched at the level of reference rather than sense.

A second consequence is that Collins takes ‘tacit’ to admit of degrees. Having said that ‘tacit knowledge is continuous with that possessed by animals and other living things’, he goes on to say that ‘in principle it is possible for it to be explicated, not by the animals and trees themselves (or the particular humans who embody it), but as the outcome of research done by human scientists’ [ibid: 85]. This comment is relevant – is not a non sequitur – because such scientific explanation tends, on his account, to undermine the tacit status. It renders the examples highlighted via cats, dogs, trees and sieves merely ‘medium degree’ (as opposed to strongly) tacit knowledge.

Elsewhere the opposition between being tacit and being scientifically explicable and the relative status of the former is made even more explicit:

In The Logic of Tacit Inference, Polanyi argues persuasively that humans do not know how they ride, but he also provides a formula: ‘In order to compensate for a given angle of imbalance α we must take a curve on the side of the imbalance, of which the radius (r) should be proportionate to the square of the velocity (v) over the imbalance r~v2/α.’ While no human can actually ride a bike using that formula, a robot, with much faster reactions, might. So that aspect of bike-riding is not quite so tacit after all. [Collins 2010b italics added]

So the fact that the task or process can be explained by others – whether or not they themselves have practical knowledge how to do it – counts against it being fully tacit for a different subject, however he or she thinks about or grasps riding a bike. Explanation by others has a kind of ‘action at a distance’ for the tacit or explicit status of a subject’s knowledge.

The assumption that there is a conflict between being tacit and being scientifically explicable is also operative when Collins notes that, for skilled typists, consciously following the rules they originally learnt by slows them down. He comments that ‘this seems to bear on nothing but the way humans work; it does not bear on the way knowledge works’ [Collins 2010a: 104]. ‘Knowledge’ simpliciter does not denote the knowledge or know how of human typists, then, but rather a thoroughly generalised account of the task or process of typing that could be given. This assimilation is also suggested in a later comment on the limits of human typing:

The constraints on the methods available for efficient typing by humans (by contrast eg with machines) are somatic limits; they have everything to do with us and nothing to do with the task as a task – nothing to do with knowledge as knowledge. [ibid: 104]

That last line makes plains the real subject matter of Collins’ book: not the knowledge a particular subject has but a task or process, whether carried out by humans, animals or even trees or sieves, independently of whether or not any knowledge is actually involved. The same blurring (roughly, of what is in the world and what is in the mind) is present in this idea: ‘the modern world is thought of as driven by explicit knowledge – patterns’ [ibid: 80]. However, when comet Shoemaker Levy 9 hit Jupiter, that exemplified a pattern codified – roughly – in Newtonian physics. But it wasn’t driven along by its, or anyone’s, knowledge. Whilst one might have knowledge of patterns, patterns are not knowledge.

Construing the antonym of tacit the way he does has far reaching consequences for his account and undermines the claim that it is an analysis of a form of knowledge at all. Let me mention just one practical consequence of this. Suppose that there is a culinary task, such as making a white sauce from a roux or example, that can be carried out either mechanically according to a recipe of rules or by judgements of taste, eye and hand. If one thinks of tacit knowledge as characterising the realm of sense rather than reference it will be a substantial question whether a particular chef, on a particular occasion, carries out the task using tacit or explicit knowledge. But if the tacit or explicit status is fixed, at the level of reference, by whether that task or process could be explained or mechanised by scientists elsewhere in the world, then that question will not apply. However the chef thinks of her task, her knowledge will be tacit or explicit in a way which lies potentially beyond her ken.

Why does Collins select such an antonym for ‘tacit’ with such broad and counter-intuitive consequences? I think that the answer to that lies in his response to Wittgenstein’s discussion of rules. But to see that, it will be necessary to take a lengthy detour through his earlier work.

Collins’ early work and the antonym of ‘tacit’

In some of his earlier work, Collins selects a different set of antonyms for ‘tacit’. Changing Order articulates a view of tacit knowledge based in part on a piece of sociological inquiry. In the 1970s, Collins visited six of the seven UK laboratories that were attempting to build a working laser of a new design (a Transversely Excited Atmospheric pressure CO2, or TEA, laser), although it had already been successfully built in other laboratories in the US. In one case, a scientist who had already built one working model aimed to replicate it so as to have two working models. Despite this limited problem – a clear case of Kuhnian ‘normal science’ – and despite the availability of explicit instructions, Collins discovered a surprising difficulty.

[N]o scientist succeeded in building a laser by using only information found in published or other written sources. Thus every scientist who managed to copy the laser obtained a crucial component of the requisite knowledge from personal contact and discussion. A second point is that no scientist succeeded in building a TEA-laser where the informant was a ‘middle man’ who had not built a device himself. The third point is that even where the informant had built a successful device, and where information flowed freely as far as could be seen, the learner would be unlikely to succeed without some extended period of contact with the informant and, in some cases, would not succeed at all. [Collins 1985: 55-6]

From his empirical investigation, Collins offers three key descriptions of the experimental knowledge under analysis.

In sum, the flow of knowledge was such that, first, it travelled only where there was personal contact with an accomplished practitioner; second, its passage was invisible so that scientists did not know whether they had the relevant expertise to build a laser until they tried it; and, third, it was so capricious that similar relationships between teacher and learner might or might not result in the transfer of knowledge. [Collins 1985: 56]

These three elements are prominent at the start of a list of propositions Collins sets out to capture the nature of the skill-like knowledge or experimental ability (in the context it is clear that these are different labels for the same thing):

Proposition One: Transfer of skill-like knowledge is capricious.
Proposition Two: Skill-like knowledge travels best (or only) through accomplished practitioners...
Proposition Three: Experimental ability has the character of a skill that can be acquired and developed with practice. Like a skill, it cannot be fully explicated or absolutely established...
Proposition Four: Experimental ability is invisible in its passage and in those who possess it.
[ibid: 73-4]

This list adds the idea that skill-like knowledge takes practice but cannot be ‘fully explicated or absolutely established’ to the previous elements: it takes practical demonstration but is invisible and capricious. The invisibility applies not only to difficulties with attempts to communicate it and hence its capriciousness, but even to those who possess it. Even the scientist whose ongoing attempts to build a laser Collins particularly studied, Bob Harrison, was unaware of the details of his own knowledge how to build the laser.

Harrison would not have been a lot of use as an informant at the beginning of his attempt to build Jumbo [his first working laser]; there is no way that he could have informed anyone about the necessity of having the leads from the capacitor to the electrodes as short as possible, for example, since he did not realise the importance of this himself. But, he did not know that he did not know. [ibid: 73]

This comment suggests a connection to tacit knowledge through Polanyi’s slogan (that we know more than we can tell). Collins makes that connection explicitly in a recent summary of this earlier work:

[Y]ou may not know what you need to know and I may not know what I know. Thus, in the early days of TEA lasers scientists did not necessarily know that the inductance of the top lead was important but by copying existing designs they built in successful short top leads without knowing why. [Collins 2010b]

But it is worth noting that this brisk summary falls foul of the second horn of the basic dilemma. In its words, the fact that the top lead needed to be short was not explicitly known by early TEA lasers scientists because it was not known (or as he says not necessarily known) at all. If they knew that existing working lasers needed to be copied, that fact alone does not imply that they had any knowledge of the component elements of the copying. Successful copying of the top lead might be merely the accidental outcome of other skilled or knowledgeable actions.

In Changing Order, however, the connection to tacit knowledge is made after listing the three key characteristics of skill-like knowledge (that it takes practical demonstration, is invisible and capricious):

These characteristics of the flow of knowledge make sense if a crucial component in laser building ability is ‘tacit knowledge’. [Collins 1985: 56]

Thus in following the method of via negativa, the antonyms of tacit seems to be theoretical (or perhaps contemplative) rather than practical, visible rather than invisible and predictably communicable rather than capricious. But why think that tacit knowledge stands opposed to such knowledge?

The Wittgensteinian roots of the early Collins’ antonym

I think that the answer to this runs deep in this early work and concerns Collins’ interpretation of Wittgenstein’s discussion of rule following. Immediately after suggesting that the best explanation for the features of experimental knowledge is that it is tacit, he first cites Polanyi’s definition that it is ‘our ability to perform skills without being able to articulate how we do them’ [ibid: 56], giving the example of bicycle riding, but then he broadens the analysis.

Tacit knowledge usually finds its application in practical settings such as bike riding or other ‘skilled’ occupations. However, it is equally applicable to mental activity. Thus, to return to an earlier example, the member of a social group who has the ability to continue the sequence ‘2,4,6,8’ with ‘10,12,14,16’ as a matter of course, without even thinking about it, also possesses something that the stranger to our culture and the newborn do not. This is sometimes referred to as a ‘social skill’ but we can call it tacit knowledge without doing too much violence to the term. It forms the foundation upon which formal learning rests. If I am taught some new algebraic manipulation in school, and the teacher tells me to do it the same way next time, I can say that it is my tacit knowledge which tells me what counts as the next instance of the same problem as well as what is meant by proceeding in the same way. [Collins 1985: 56-7]

Even though the mathematical example seems to contrast with paradigmatically practical skills, such as riding a bike, it highlights the role that practical judgement plays in underpinning the conceptual order. It is no surprise, therefore, that discussion of Wittgenstein’s remarks on rule following plays an important role in the introduction to Changing Order.

Collins’ account there is brisk. Echoing Wittgenstein’s discussion in §185 of the Philosophical Investigations, he considers the example of being asked to continue the ‘2,4,6,8’ sequence in the same way. The ‘immediate answer that springs to mind is ‘10,12,14,16’ and, to all intents and purposes, this is indeed the “correct” answer’ but he presses the question of how we know this [ibid: 13]. It cannot, he argues, be a matter of following the rule ‘go on in the same way’ because ‘this rule allows for a number of possibilities’ [ibid: 13]. Nor, assuming that that rule is merely insufficiently specific, does further codification that one sequentially adds 2 help because that might result in the continuation ‘82, 822, 8222...’ or other typographic variants each of which amounts to adding 2 in some sense.

Collins concludes both that the notion of ‘sameness’ is ambiguous and that it is not possible fully to specify a rule (unless a limited range of responses is defined in advance). But ‘since in spite of this we all know the correct way to go on, there must be something more to a rule than its specifiability’ [ibid: 14]. The extra element is described in the introduction as ‘social entrenchment’ or a ‘shared form of life’. Later, as in the quotation above, it is called ‘tacit knowledge’. Thus it is tacit knowledge that underpins the ‘mysterious abilities that enable us to know when to continue ‘2,4,6,8’ with ’10,12,14,16’ and when with ‘who do we appreciate?’’ [ibid: 22 italics added].

I think that this is a disastrous view of the connection between the rule following considerations and tacit knowledge but, before saying why, I will explain further why one might mistakenly hold it. Wittgenstein considers understanding a mathematical rule such as the rule for counting in twos in §185 which builds on an earlier paragraph.

Let us return to our example (§143). Now – judged by the usual criteria – the pupil has mastered the series of natural numbers. Next we teach him to write down other series of cardinal numbers and get him to the point of writing down series of the form
0, n, 2n, 3n, etc.
at an order of the form “+ n”; so at the order “+ 1” he writes down the series of natural numbers. – Let us suppose we have done exercises and given him tests up to 1000.
Now we get the pupil to continue a series (say + 2) beyond 1000 – and he writes 1000, 1004, 1008, 1012.
We say to him: “Look what you’ve done!” – He doesn’t understand. We say: “You were meant to add two: look how you began the series!” – He answers: “Yes, isn’t it right? I thought that was how I was meant to do it.” – Or suppose he pointed to the series and said: “But I went on in the same way.” – It would now be no use to say: “But can’t you see....?” – and repeat the old examples and explanations. – In such a case we might say, perhaps: It comes natural to this person to understand our order with our explanations as we should understand the order: “Add 2 up to 1000, 4 up to 2000, 6 up to 3000 and so on.”
[Wittgenstein 1953 §185]

The example may seem to present the problem as though it is one of third person epistemology: both of (our) determining whether the pupil has grasped the series and of the pupil understanding what is intended. That is indeed part of the problem. But the more fundamental issue is not just the epistemology of understanding but rather in what understanding the series can consist, given that a proper understanding has to rule out divergences of the sort suggested.

Suppose that the mathematical rule (or, by analogy, the meaning of a word) is taught by examples, by the first few numbers of the series (or some paradigm examples of word use). The hypothetical example of the deviant pupil suggests the following worry. Since finite examples under-determine the correct later applications (of a word or series), they can only determine the correct rule under a specific interpretation. They must be interpreted as indicating a particular continuation, for example. If so, however, this leads to a problem.

“But how can a rule shew me what I have to do at this point? Whatever I do is, on some interpretation, in accord with the rule.” – That is not what we ought to say, but rather: any interpretation still hangs in the air along with what it interprets, and cannot give it any support. Interpretations by themselves do not determine meaning. [Wittgenstein 1953 §198]

The problem is not simply that the selection of an interpretation to single out a particular set of applications is unjustified. That does seem to be a problem insofar as the first few numbers are consistent with an infinite number of interpretations. But there is a more fundamental problem which is that the meaning of the interpretation itself has somehow to be specified and the original problem is still in play for any account along the same lines of how that is possible. There seem to be only two options, neither of which is satisfactory.

If, on the one hand, possessing the right interpretation (of the initial numbers in the mathematical series or of finite examples of the use of a word) simply consists in having a potentially unlimited number of correct applications somehow all come to mind, how is that possible? It seems to be an absurd idea. But if possessing the right interpretation is a having a mental item, or image, or mental talisman, before the mind’s eye, which summarises or yields all the right applications, how does it determine subsequent correct moves? Surely, like an external sign or image, it can only do that under a particular interpretation? By itself, such an item seems no more to determine what would accord with it, and what not, than the finite examples of word use, or the first numbers in the series, with which we began. But if such an interpretation only determines what accords with it under a further interpretation then it ‘still hangs in the air along with what it interprets, and cannot give it any support. Interpretations by themselves do not determine meaning.’

One example, repeatedly used by Wittgenstein, serves both to explain the problem and point towards his suggested solution.

A rule stands there like a sign-post. – Does the sign-post leave no doubt open about the way I have to go? Does it shew which direction I am to take when I have passed it; whether along the road or the footpath or cross-country? But where is it said which way I am to follow it; whether in the direction of its finger or (e.g.) in the opposite one? – And if there were, not a single sign-post, but a chain of adjacent ones or of chalk marks on the ground – is there only one way of interpreting them?... [Wittgenstein 1953 §85]

If one understands which way a sign-post points then one knows it points in the direction of the ‘finger’ not the other way. But one might have taken it to point the other way. The sign-post does not take a viewer by the throat as Achilles suggests that logic will take the Tortoise by the throat in Lewis Carroll’s story [Carroll 1905]. So it can seem that the sign-post points, again, only under an interpretation. But what would grasp of such an interpretation consist in? Again we can frame a dilemma. It either consists – impossibly – in grasping all the potential places to which sign-posts might point (the way out, the pub, the lecture-room). Or it consists in entertaining a mental item which determines all those places relative to a given sign-post. In this case, the most obvious candidate would be an inner image of another sign-post with an indication of which way it points. How? By another sign-post, perhaps. And so a regress begins.

This example makes the problem very clear. In this key respect (which way it points), the inner sign-post is no different from the outer one. If the outer sign-post needs an interpretation then so will the inner one. And that threatens a vicious infinite regress, stopping short of completing which will leave the subject with no understanding of which way any of the – inner or outer – sign-post-posts point. It will leave no understanding because for any sign-post all one knows is that if the next higher order sign post points left then so does this one and if it points right then so does this one, but as yet, one does not know which way that higher order sign points. For it, an equivalent conditional applies with respect to the yet higher sign-post.

In the face of these difficulties with making the right understanding of the rule explicit, it is tempting to retreat to a notion of tacit knowledge that plugs a gap between the rule that, as a matter of fact, we are able to grasp and the gappy and inadequate specification we are able to make through examples and paraphrase. But a notion of tacit knowledge deployed in this way would be slippery in something like the way the early Collins’ account suggests. Although connected to practical demonstrations – in the example, of how to continue the series correctly – such demonstrations would be capricious because, as the deviant pupil seems to show, any explanation could be taken the wrong way. And since any finite further instances of rule following would not imply a correct understanding (since it has to cover an infinite extension) whether the explanations had been grasped in the right way would be invisible in third person cases. But since nothing that can come before the mind’s eye in the first person case can determine the correct application, it would be also invisible in that case too. So such tacit knowledge would be connected in some sense to practical demonstration but, nevertheless, invisible and capricious as the early Collins thinks.

However, although this way of connecting the discussion of rules with tacit knowledge is intuitive, it is also misleading. That it is misleading can be clarified by identifying three problems it faces.

This view of tacit knowledge is impaled on the second horn

The first problem with it, as an interpretation of Wittgenstein, is that it accepts part of what he criticises: a platonic picture of rules as rails ‘invisibly laid to infinity’ fundamentally distinct from our capacity to articulate them [ibid: §218]. That picture is easily prompted by the case of the deviant pupil. What that case, and others like it, seems to show is both that any finite set of examples underdetermines a correct understanding of the rule and that such correct understanding must involve grasp of a supernatural pattern. Since no actual human enumeration of the pattern seems enough to determine it, it must be supernatural. Hence the metaphor of rails laid to infinity. With this picture of the way rules determine correct moves in place, there is a substantial role for tacit knowledge to bridge the gap between what can be made explicit in the sublunary realm and the ideal platonic rule. But if so, it seems that Wittgenstein offers support for the platonic picture he also seems to criticise. To put this point in the terms used by John McDowell, such a picture of tacit knowledge presupposes a rampantly platonic picture of rules, one in which the normative demand they place on rule-followers is conceived as independent of human thought [McDowell 1994].

A second problem concerns the communication of knowledge of rules. On this interpretation explanations are insufficient explicitly to fix a unique rule, which depends instead on a tacit understanding both by speaker and hearer. But if so, the tacit grasp of a particular rule cannot be a matter of knowledge even if it were, as a matter of fact, of the rule intended. Nothing could justify the selection from the infinite range of alternative options. The best case would be that hearers were disposed to select a particular rule because of a shared background of dispositions but this would not be a matter of justification. Now one might object that a reliablist conception of knowledge would still be possible but even this way of putting things is put under strain because of a third problem.

The third and final problem, then, is one of accounting for the idea that tacit knowledge of a rule or the meaning of a word has some particular content to be known. The problem, though, is that this means that nothing can be said by way of positive account of what the tacit knowledge amounts to since any attempt will fall prey to the objections already rehearsed. No demonstration of its content can succeed. If it could, there would be no need for such a conception of tacit knowledge. But if there is a need for such a form of tacit knowledge, no account of its content (to another, or to oneself) can succeed. If that is the case, what reason is there to think that what remains tacit is a ‘something’ at all? It may justify the label ‘tacit’ but only at the cost of undermining the idea of knowledge.

These three problems all stem from the idea that tacit knowledge is needed to plug a ‘gap’ between what can be explained, or otherwise made explicit, and the full grasp of a rule that can be understood as a result. This, however, rests on a misunderstanding of Wittgenstein’s dialectic, which aims to undermine the very idea of such a gap. He suggests, instead, that there is a close connection between what a teacher can express and what a student can grasp in the examples that manifest the teacher’s meaning.

But do you really explain to the other person what you yourself understand? Don’t you get him to guess the essential thing? You give him examples, – but he has to guess their drift, to guess your intention.” – Every explanation which I can give myself I give to him too. – “He guesses what I intend” would mean: various interpretations of my explanation come to his mind, and he lights on one of them. So in this case he could ask; and I could and should answer him. [Wittgenstein 1953 §210]

“But this initial segment of a series obviously admitted of various interpretations (e.g. by means of algebraic expressions) and so you must first have chosen one such interpretation.”–Not at all. A doubt was possible in certain circumstances. But that is not to say that I did doubt, or even could doubt… [ibid §213]

In §210 the interlocutor expresses the worry that, since an explanation fails to determine the rule to be explained, a listener has to guess – from an infinite range of options – what rule was intended. The guess is needed to bridge the gap between what is actually expressed and what was really intended. But Wittgenstein’s response is to equate what can be explained to another person and what might have been assumed to be epistemically optimal: what a speaker can explain to him or herself. This equation might be thought – optimistically – to offer in the third person case the happy circumstances of the first person case: what one knows one intends in one’s explanation. But it might also be thought – pessimistically, in the context of an inquiry that undermines the efficacy of mental templates to underpin one’s own grasp of a rule – to limit what is available to others to what is available to oneself. Either way, the connection undermines the idea that a guess is necessary to bridge a gap between first- and third-person cases.

§213 applies the moral of §210 to the explanation of a rule. Whilst some explanations can fail, that is not the general case. Although Wittgenstein rejects substantive explanations of our grasp of rules, via mental mechanisms, he does not claim that there is a gap between what can be manifested and what must be understood: a gap that has thus to be filled by a tacit element. Recognising that our understanding can be expressed in examples undermines the gap between the sublunary and the platonic and thus that potential role for tacit knowledge. It also blocks the worry raised above that such a model of the tacit understanding of rules or meanings would put under pressure the idea that there is something to be known, a content grasped. There is a content that can be expressed in examples or ongoing practice.

The same line of thinking is present in the earlier discussion of sign-posts. After asking ‘is there only one way of interpreting them?’ §85 continues:

So I can say, the sign-post does after all leave no room for doubt. Or rather: it sometimes leaves room for doubt and sometimes not. And now this is no longer a philosophical proposition, but an empirical one. [ibid §85]

But that reassurance prompts the question of the connection between the sign-post and understanding which way it points given the logical problem highlighted by the regress argument. Wittgenstein returns to the example of the sign-post in the way that the later paragraph we have been discussing, §198, continues:

“Then can whatever I do be brought into accord with the rule?” – Let me ask this: what has the expression of a rule – say a sign-post – got to do with my actions? What sort of connexion is there here? – Well, perhaps this one: I have been trained to react to this sign in a particular way, and now I do so react to it.
But that is only to give a causal connexion; to tell how it has come about that we now go by the sign-post; not what this going-by-the sign really consists in. On the contrary; I have further indicated that a person goes by a sign-post only in so far as there exists a regular use of sign-posts, a custom.
[Wittgenstein 1953 §198]

Wittgenstein suggests that, in the face of the regress of interpretations, rule following is a regular use, or custom or practice. With such a custom in place, the guidance that a sign-post gives is a paradigm of explicit, not tacit, knowledge. To read the rule following considerations as supporting a conception of tacit knowledge because explanations of meaning using signs are always inadequate is to ignore a key part of Wittgenstein’s discussion. It is to accept part of what he rejects – the Platonism of rails to infinity – and to ignore what he emphasises: the importance of custom and practice. (This is not to say that Wittgenstein provides no support for a conception of tacit knowledge, but Collins locates it in the wrong place.)

The legacy of Collins’ reading of Wittgenstein in Tacit & Explicit Knowledge: strings

I can now return to Collins’ more recent work. Why does he blur the knowledge that subjects can possess with the worldly processes that the knowledge can be of, thus blurring the realms of sense and reference? For example: ‘That which is not explicit knowledge is mostly just the way the world unfolds.’ [Collins 2010a: 80]. Although he goes on to suggest that ‘mechanism’ is a ‘more appropriate’ label than ‘tacit knowledge’ for the working out of mechanical sequences of greater or lesser complexity, he does not firmly object that to call such worldly processes ‘knowledge’ at all is a bizarre anthropomorphism. I suggest that this follows from his reading of Wittgenstein.

As I remarked earlier, Tacit & Explicit Knowledge aims to shed light on tacit knowledge by contrasting it with a suitable account of what is explicit. His initial characterisation of the explicit (one of the endpoints is construing it as explicable and then explainable, as I highlighted at the first section) is not, however, with what can be expressed linguistically but rather with what he calls ‘strings’. ‘“Explicit” is something to do with something being conveyed as a result of strings impacting with things.’ [Collins 2010a: 57]

Strings are ‘bits of stuff inscribed with patterns: they might be bits of air with patterns of sound waves, or bits of paper with writing, or bits of the seashore with marks made by waves, or patterns of mould, or almost anything’...’ [ibid: 9] The motivation for this seems to be to avoid the ‘freight of inherent meaning that makes the notions of signs, symbols and icons so complicated’ [ibid: 9]. One worry repeated in the book is that signs do not have an essential meaning. By contrast, ‘a string is just a physical object and it is immediately clear that whether it has any effect and what kind of effect this might be is entirely a matter of what happens to it.’ [ibid: 9]

The worry – that signs do not have meanings essentially – is a reasonable response to Wittgenstein’s discussion of rules and rejection of platonism. What kind of sign could compel an interpretation, independently, that is, of a contingent background practice of sign use? But not spotting the opportunity that the second half of that questions affords – the idea of a custom or practice of sign-use – Collins’ response has two unfortunate consequences. The first is that, when he does talk about meaning, he is forced to say a number of awkward things about it.

One consequence is that without a connection to sign use, meanings have to be reified. ‘A language is a set of meanings located in a society, whereas, to repeat, strings are just physical objects.’ [ibid: 10]. This sharp contrast results from the absence of the idea of signs in use. The reification in turn leads to very platonic sounding comments such as: ‘Though strings are sometimes used to represent meanings, their relationship to meanings cannot be stabilised... because meaning is continually changing as it lives its life in society.’ [ibid: 44 italics added]

It also gives rise to the need to talk of using strings to represent meanings, though how dead strings can be animated in such a way as to represent meanings and what, and where, meanings are remains dark. The most explicit attempt to fill out the account runs as follows in a sketch of inter-personal communication (which I have abbreviated significantly to emphasise the key elements).

Language translation or just plain conversation within one natural language consists of three stages... Stage 1: inscription. In “telling” the attempt is made to represent lived meaning with the inscribed string. For example, in the case of conversation an attempt is made to represent the meaning as a string comprising vibrations in the air... Stage 2: transmission and transformation... Stage 3: interpretation. This is the attempt to recreate meaning from the string – to interpret it. [ibid: 27-8]

In this picture, Collins seems to subscribe to a C17 view of communication which Wittgenstein summarises, before going on to criticise it, rather nicely thus:

It seems that there are certain definite mental processes bound up with the working of language, processes through which alone language can function. I mean the processes of understanding and meaning. The signs of our language seem dead without these mental processes; and it might seem that the only function of the signs is to induce such processes, and that these are the things we ought really to be interested in... We are tempted to think that the action of language consists of two parts; an inorganic part, the handling of signs, and an organic part, which we may call understanding these signs, meaning them, interpreting them, thinking. [Wittgenstein 1958: 3]

Collins seems to be driven to this picture of meaning because of his underlying misreading of the rule following considerations. Missing the possibility that, as part of a custom or general use, a sign-post, for example, can simply mean turn left, he is forced to empty signs or strings of meaning: ‘There is no meaning in the book or the photograph’ [Collins 2010a: 36]. But now with (nearly?) all the work done by human interpreters and little done by dead or empty signs he has to admit to embarrassment:

[I]nstead of saying ‘capable of being interpreted’, I will adopt the term ‘affords the interpretation’, which carries the implication that there is something in the string that makes it easier to interpret one way rather than another... What ‘afford’ does not mean is ‘determine’... The terms ‘afford’ and ‘affordance’ are lazy terms... [which] merely paper over deep cracks in our understanding – or, or at least, my understanding – of why, given the extraordinary interpretative  capabilities of humans, anything affords any one interpretation better than any other. How are meanings ever fixed, or even favoured? [ibid: 35-6]

This problem marks the return in Tacit & Explicit Knowledge of what gave rise to such a slippery conception of understanding in Changing Order. It is the result of ignoring Wittgenstein’s response to his own regress argument.

The second consequence is the feature of the account that I criticised in the first section. Collins approaches what is explicit not through signs but through strings. ‘Explicit knowledge has substance – it is knowledge that can, to some extent, be transferred by the use of strings in the right circumstances’ [ibid: 80]. But because strings are simply ‘bits of stuff inscribed with patterns’ they are ubiquitous. The result is that there is no distinction between string transformations and mechanical causes and effects. But, since strings are used to underpin explicit knowledge, and hence as part of the antonym of tacit, this puts stress on the realm of tacit knowledge. If the transformation of one pattern into another can be explained, for example, then that can no longer be a matter of or for tacit knowledge.

[S]tring transformations and mechanical causes and effects are, to speak metaphysically, just two aspects of the same thing. This is why we have a strong sense that when we explain some process scientifically we have made it explicit; this is the ‘explicable’ part of the antonym of tacit with its ‘scientifically explained’ connotation. [ibid: 50]

To return to the culinary example mentioned earlier, if there is a pattern in the behaviour of a roux being heated and in the addition of milk and stirring to make a white sauce and if that can be explained by a food scientist then the instinctive chef’s ability to carry this out by eye turns out to be explicit rather than tacit knowledge. Although, ‘tacit knowledge’ is a comparatively recently introduced technical term and thus an account of it is as much stipulation as description, this seems to be a substantial change of subject. But it is not an isolated element aspect of Collins’ thinking. It is, I suggest, the result of his misunderstanding of the weight of Wittgenstein’s rule following considerations.


Following Polanyi’s maxim, Collins approaches tacit knowledge by careful contrast with an antonym. Tacit knowledge is not explicit. But, in his early work, his views about what can be made explicit in language use are influenced by a misreading of Wittgenstein’s discussion of rule following and one element of his later account threatens to change the subject away from the knowledge possessed by subjects or agents and onto merely worldly processes. I have argued that this, too, is the result of a missed opportunity in responding to Wittgenstein.


Carroll, L. (1905) ‘What the Tortoise said to Achilles’, Mind 4: 289-90

Collins, H. (1985) Changing Order: Replication and Induction in Scientific Practice, London: Sage

Collins, H. (2010a) Tacit and Explicit Knowledge, Chicago: University of Chicago Press

Collins, H. (2010b) ‘Tacit knowledge: you don't know how much you know’ New Scientist 31st May

Jackson, F. (1982). ‘Epiphenomenal qualia’ Philosophical Quarterly: 127-136

McDowell, J. (1994) Mind and World, Cambridge Mass.: Harvard University Press

Polanyi, M. (1967b) The Tacit Dimension, Chicago: University of Chicago Press

Wittgenstein, L. (1953) Philosophical Investigations, Oxford: Blackwell