At first, I saw mountains as mountains and rivers as rivers. Then, I saw mountains were not mountains and rivers were not rivers. Finally, I see mountains again as mountains, and rivers again as rivers.
It has come to my attention that contemporary epistemology is disconcertingly arse-backwards. This is because it’s caught in the uncompromising grip of an obsession with knowledge-that. This, over half a century after Gilbert Ryle famously made a strong case that knowledge-that is not all there is to knowledge as such. Disappointing.
All the way back when I was writing my honours thesis – which applied knowledge-how to Frank Jackson’s Knowledge Argument in the philosophy of mind – it appeared as though there was at least a modicum of debate going on over the nature of knowledge.
But in the decade that has lapsed since, it seems knowledge-that has come back to the fore an, in my opinion, thoroughly gummed up the works when it comes to some of the most important questions in epistemology: what is knowledge?; to what does it apply?; how is it acquired?; can we really know anything?; is there such thing as a priori knowledge?; can anything be said to be analytic?
These are important questions – more-so than many in metaphysics – because they virtually underpin every other philosophical endeavour, as well as relating to a number of very significant real-world issues, such as ethics (and metaethics), politics, science, and philosophy of mind.
So, what I’d like to do here is espouse an alternative view to the paragon view of knowledge-that espoused by Stanley and Williamson, who recently suggested that knowledge-how is a species of knowledge-that. In fact, I’d like to espouse the entirely opposite view: that knowledge-that is a species of knowledge-how. An arse-forwards view, one might say.
What Do You Know?
Let me start by posing a few paradigmatic sentences regarding knowledge of one kind or another:
1) Hannah knows 2 + 2 = 4
2) Hannah knows the sky is blue
3) Hannah knows how to ride a bicycle
The first two are commonly accepted examples of knowledge-that: 1) is a priori; 2) is a posteriori; while 3) is knowledge-how.
The first point I’ll make is we should treat our language in this field as largely suspect, as we should our intuitions about what counts as knowledge. We use terms like ‘knows…’, ‘believes…’, ‘can…’, ‘forgot…’ in many contexts, not all of which might be proper attributions of knowledge of some kind.
Consider the tendency in nature documentaries to attribute knowledge to animals: “the spider knows how to weave an intricate web”; “the baby turtles know how to find the sea”. They are certainly saying something interesting, but it we need to be cautious of all usages of language when it pertains to knowledge in case we’re actually talking about several things and only using one word, or talking about one thing and using several words etc.
This is one reason why I’m not swayed by Stanley and Williamson’s paper. A thorough analysis of the language of knowledge it might be, but that doesn’t convince me that our language and intuitions correspond to anything terribly special.
So, on with the show.
Bottom Up and Top Down
In laying down my alternative approach to knowledge, I’m going to start up in the air and make my way down to the ground. Why? Because I think contemporary theories of knowledge are backwards already, I hope this will make my theory more intelligible by starting somewhere familiar.
1) Hannah knows 2 + 2 = 4
This kind of sentence expresses knowledge-that, or propositional knowledge. It’s a statement about the agent (Hannah), a proposition (2 + 2 = 4) and establishing the relation of ‘knowing’ between them. Nothing too controversial here.
That’s not to say this picture of knowledge is without its problems. For one, the ‘justified true belief’ model of propositional knowledge has been batted around for centuries, to the point that it’s looking quite battered these days. It seems to be mostly accurate, but niggling problems keep cropping up, whether they be Gettier problems or understanding the truth conditions of such knowledge.
Then there are Quine’s objections against analytic sentences, which are sentences that are knowable to the true simply by an understanding of the terms involved, such as the texbook: “all bachelors are unmarried”.
I won’t deal with these issues here – because ultimately I don’t think they’re problems at all. This is because these kinds of propositions are never going to be unambiguously true, at least of the concrete world. That’s because these propositions are ultimately abstractions from the concrete world – a world I’m going to talk about shortly. And as abstractions, they’ve already abandoned some crucial information about the world, so they’ll never be able to perfectly represent that world. No two ducks are exactly the same, so the word ‘duck’ represents a vague collection of things. Break a piece of chalk in half and some will say you’re holding two halves of a piece of chalk – others will say you now hold two pieces of chalk.
But wait, isn’t 1) a priori – independent of experience? Yes, it is. The proposition ‘2 + 2 = 4’ is an abstract proposition about other abstract propositions. As such, they can be ‘true’ – or ‘consistent’ – within that particular ‘proposition-‘ or ‘concept-space’, such as the space of mathematics. However, I’d suggest – in true empirical tradition – that such propositions need to be related to the world in order for them to say something true of the world itself. Otherwise, they’re simply talking about their concept space. Likewise for so-called ‘analytic’ sentences – the best they can achieve is consistency within their concept space, but that space will never be perfectly grounded in reality.
Mathematics started from abstractions from experiences of the world, from which point lots can be said within mathematics – and even new mathematical knowledge discovered within that system, a la a priori knowledge – but maths (or any axiomatic system) will never be entirely consistent and complete. Gödel told us that. So when it comes time to relate maths back to the world, it won’t relate perfectly. And I’d suggest this is the case with all abstract systems, including language, propositions and thus knowledge-that.
I call this a priori propositional knowledge 3rd-order knowledge. And I’m using the term ‘knowledge’ reservedly here, in a more psychological sense, as I don’t believe a priori knowledge can ever be perfectly true when related back to the world. And even when analytic, it’s only a weak form of ‘truth’ (there might be a better name for it) since it’s only true within its abstract concept space and not of the concrete world.
2) Hannah knows the sky is blue
This sentences expresses knowledge from the next tier – 2nd-order knowledge. This is another example of propositional knowledge, except in this case it’s a posteriori rather than a priori; the truth of this statement depends on some fact about the world.
However, like 3rd-order knowledge, this kind of propositional knowledge involves abstraction about the world. ‘The sky’ and ‘blue’ are vague abstractions from the concrete world. That’s not to say that there isn’t something in the concrete world we’re talking about when we mention ‘the sky’ or the colour ‘blue’. But the proposition ‘the sky’ and ‘blue’ will never perfectly capture or represent that thing we’re talking about in all its detail.
Big deal? Yes. Abstractions are useful, but they’re also deceptive. I’ll talk about the concrete world in a moment, but whenever we abstract from the concrete world we make at least one error – we draw distinctions. A and not-A. The sky and not-sky. Blue is not red. Yet these distinctions don’t exist in the concrete world, which is single and contiguous, but not homogeneous. As such, any proposition based on such an abstraction will only ever be a vague approximation of some aspect of the real world – like a stereotype, not 100% accurate, but useful none the less.
I don’t think this in any way prevents us from using such propositions – I just think we need to stop short of talking about ‘knowledge’ about the world (and from that, knowledge within a proposition space abstracted from the world) in black and white terms. If, by ‘truth’, we mean perfect correspondence with some aspect of the world – that’s impossible.
3) Hannah knows how to ride a bicycle
Now we’re in knowing-how territory. This is 1st-order knowledge, under this model. Knowing-how is non-propositional. However – and this is one reason I think we need to be careful with our use of language – whenever we talk about knowing-how we use propositions: the language of knowing-that. And these propositions will never completely describe the knowing-how.
So 3) could be translated into saying:
4) I know that Hannah knows how to ride a bicycle
I’d characterise 3) in a similar way to Ryle. 3) means that Hannah has a disposition to perform a certain action under the right circumstances, and she possesses this disposition by virtue of possessing certain mental and physical properties. The mental properties needn’t be conscious, and in fact, I’d suggest a vast majority of 1st-order knowing-how is non-conscious. Heck, think about the last time you drove home. Were you conscious of your gear changes, or indicator usage, or lane changes?
And this is where I think a lot of confusion crops up when talking about knowing-that and knowing-how. Because knowing-how can also be abstracted away into knowing-that. So when I say something like 3), this could also be stated in terms of a long ream of propositions about pedals, balance, applying pressure to the breaks, inertia, wind resistance etc. However, I’d suggest these propositions would never a) perfectly describe the knowing-how (see above) and b) they are not what we’re talking about when we say Hannah knows how to ride a bicycle.
Hannah riding a bicycle is concrete, not abstract. She knows how to ride the bike because she does (in the appropriate circumstances). Any description of her riding is abstract, thus cannot capture the activity in its entirety.
Now, an objection is often raised at this point, one that Stanley and Williamson raise against Ryle. They suggest one of Ryle’s premises is:
If one Fs, one employs knowledge-how to F.
And they provide an example in this vain:
If Hannah wins a fair lottery, she still does not know how to win the lottery, since it was by sheer chance that she did so.
But, I’d suggest this is a not an example of 1st-order knowledge, because winning the lottery requires no mental activity. They could have just as easily said that Hannah knows how to use gravity to stay attached to the Earth. I would suggest that any 1st-order knowledge requires some kind of mental activity, and as such, has mental properties upon which to base the dispositional nature of this species of knowledge-how. And that rules out lotteries and gravity from being knowledge-how.
But… I do think we attribute ‘knowledge’ to people, animals and even things even when we’re talking about acts that don’t involve mental activity: the spider knows how to weave a web; the plant knows how to angle its leaves towards the sunlight; the seeds know to germinate when the 10 year flood arrives, etc.
Now, we may well be mistaken to call this knowledge in any of the above senses. But we’re talking about something interesting here, and something I think is often overlooked. For if you explore what we’re talking about, we’re still talking about dispositions. Except instead of requiring mental properties, they require physical properties alone.
No difference in kind, just degree – properties by virtue of which that thing possesses that disposition. Which brings me to the ground floor.
While I’m happy to call 1st-, 2nd- and 3rd-order knowledge ‘knowledge’ (at least in a psychological sense, if not justified true belief), I still would like to admit a species of disposition that I think underlies all of our knowledge-how: 0th-order ‘abilities’.
Abilities are much like 1st-order knowledge in that they’re dispositions. Something has the ability to x if it will x under the appropriate circumstances. So yes, a glass has the ‘ability’ to break. (Maybe there’s a better word than ‘ability’, I don’t know. I’m reluctant to use ‘disposition’ as it’s a complex term that appears in the other orders of knowledge. But it is the foundation of those orders of knowledge, so perhaps it’s more suitable. I’m open to suggestion.) Again, it’s concrete, not abstract.
We can talk about thinking as one example of an ability possessed by humans. Ryle even seems to hint that we need some kind of prior account of cognition in general before we start talking about the font from which knowledge springs. I’d suggest thinking, and all its related faculties, are abilities held by an individual by virtue of their physical properties, namely their possession of a brain structured so.
Nothing really remarkable about 0th-level abilities, but I think it’s important to have them on the ground floor to give a foundation to the higher orders of knowledge. And again, a word of caution not to mistake the description of an ability (in terms of knowledge-that) for the ability itself.
So there you have it. A four-tiered model of ‘knowledge’. One that places abilities and knowledge-how at the base, and dethrones propositional knowledge from its foundational status as found in Stanley and Williamson. Furthermore, knowledge-that requires knowledge-how and abilities to exist. For one cannot abstract without the ability to do so.
I’ll state one counter-example that is often cited to deny that knowledge-how is prior to knowledge-that.
4) Hannah knows how to do 1,000 push-ups
Hannah might be quite capable of doing one push-up, and she might be able to imagine doing 1,000 push-ups, but she is a long way from being able to actually do 1,000 push-ups. According to the Stanley-Williamson account, Hannah could know-how to do 1,000 push-ups without actually being able to do them. This is because she knows a number of propositions that could constitute her knowledge of how to do 1,000 push-ups, were she physically able to do them.
Under my alternate model of knowledge, Hannah doesn’t know-how to do 1,000 push-ups, because she cannot do them. She might know-how to do one push-up, or 10. But this knowledge-how is dispositional, depending on her mental and physical properties (the only difference with 0th-order abilities is they depend on physical properties alone).
Yet she might possess some knowledge-that abstracted from past experiences of push-ups, along with some theoretical knowledge of human physiology and physics. With this knowledge-that, she could perhaps describe to someone how to do 1,000 push-ups in the same way she might describe how to ride a bicycle to a learner. Yet the person receiving the exhaustive list of propositions about riding a bicycle or doing 1,000 push-ups may quite plausibly be unable to actually do either one. The learner wouldn’t acquire the knowledge-how to do the 1,000 push-ups.
You also might be wondering why I chose that particular quote at the beginning of this post. That’s a Zen proverb that I think captures the spirit of this model of knowledge. For in Buddhism it is acknowledged that the concrete world is unitary and contiguous, and it’s us that cut and dice it into discrete chunks and apply labels to things like mountains, rivers and bicycles – and then try to shoehorn logic around it as if it was already there.
These labels are descriptions – incomplete descriptions – that don’t perfectly represent the objects to which they refer. So if we’re to look at mountains, we shouldn’t fall for the illusion that that’s all they are – a bundle of discrete things as represented by our propositions. Our labels, our propositions, are alluring, but they’re not enough to represent reality. Thus, to see mountains as they really are requires us to unshackle ourselves from the distinctions and propositions we project on to the world.
But – and there’s the zen twist – those propositions are still necessary to our understanding of the world. We might escape our constrained perspective on the world for fleeting moments – and in those moments we might even see things as they are – but when it comes to making sense of the world, we need those limited, clunky, inaccurate labels and propositions. So we get on with it, and mountains become mountains, rivers become rivers, and bicycles become bicycles again.