[I am grateful to Wesley Buckwalter for discussion on these issues]
My daughter's been learning to play the guitar for about 4 years now (classical and chords, notes and tabs),
and she has now reached a level where she is using intermediate techniques such as harmonics and vibrato. In one of her method books, these techniques are explained by pictures and simple texts. As you can imagine, it is hard to learn these techniques using the pictures and verbal descriptions alone.
Testimony - broadly, learning from what others tell us - is an enormous source of knowledge, especially knowledge we haven't learned first hand. Without testimony, I would not even know the place and date of my birth, or that I live on a spherical earth. Yet as great as testimony is for transmitting factual information, it seems less suitable for transmitting skills - knowing-how.
Ted Poston has recently argued that this difference between testimony to facts and testimony to skills points to a fundamental difference between knowledge-that and knowledge-how. Intellectualists claim that all forms of knowledge-how (e.g., riding a bicycle, playing the guitar) ultimately are reducible to knowledge that (properties of bicycle-riding and guitar-playing, respectively), a position influentially defended by Jason Stanley and Tim Williamson. Poston argues that the fact that skills aren't easily transmitted through testimony provides "good evidence that practical knowledge is not propositional knowledge".
- I know Jimi Hendrix is a guitarist
- I tell my daughter that Jimi Hendrix is a guitarist
- Therefore, my daughter knows Jimi Hendrix is a guitarist
By contrast, the following inference is - in Poston's view - bad
- I know how to play harmonics on the guitar
- I tell my daughter how to play harmonics on the guitar
- So, my daughter knows how to play harmonics on the guitar
On the face of it 4-6 sounds indeed like a terrible inference. But it depends on what you mean by "telling S how to phi". If telling S how to phi means to just verbally indicate how phi-ing is done, there is no testimonial knowledge transmission. But constraining cases of testimony to verbal testimony is unhelpful, especially if we take a broader cross-cultural and diachronic perspective.
As Kim Sterelny has argued in The evolved apprentice, lots of human knowledge transmission is non-verbal and proceeds through deliberative demonstration. Young children have cognitive adaptations that allow them, from infancy onward, to distinguish relevant from irrelevant motions in such demonstrations, and they use these to acquire a wide range of skills.
Archaeologists have some idea of when language evolved, and when demonstration evolved. Interestingly, the earliest stone tools that require demonstration (Oldowan technology) are far older than the hypothesized earliest emergence of language. Oldowan stone tools require careful selection of raw materials that have a particular breaking pattern, resulting in conchoidal fractures when struck at proper angles (such as quartz and flint).
They also require a particular knapping technique, striking the base of the stone in an oblique angle to obtain razor-sharp flakes that can be used to butcher meat and for other purposes. The oldest Oldowan stone tools are older than 2.5 million years. The earliest emergence of language, based on the shape of vocal tract (in particular, the hyoid bone), brain size, and genetic evidence (ancient DNA, in particular FOXP2) of hominids with a common ancestors, such as Homo sapiens, neanderthals and Denisovans, is perhaps 1 million year old. It's unlikely to be longer than 1.5 million years ago. This gives us a time frame of at least 1 million year when hominids did not speak, but had to use demonstration to get across knowledge how to flake stone tools.
Taking this diachronic perspective, testimonial transmission of knowledge-how is not a borderline phenomenon of testimonial transmission, but to the contrary, it is the human ancestral form of testimonial transmission. Thus I am attracted to Hetherington's practicalism, the view that ultimately all forms of knowledge-that are reducible to knowledge-how.
Wesley Buckwalter and John Turri have recently argued that "Just as knowing that is the norm of information transmission, knowing how is the norm of skill transmission. In brief, just as knowing is the norm of telling, so too knowing is the norm of showing." They offer several reasons for this proposition: when you ask if someone knows how to phi, it is often an indirect request to show how to phi. If someone does not know how to phi, that's a legitimate excuse for not instructing someone how to phi, and it seems strange for someone to say "I don't know how to phi, but this is how it's done".
If demonstration (showing) fits within a broad conception of testimony, I believe this can provide insights into how testimony could evolve. In game theoretical models of animal communication, there is always a question of how deceitful communication - what prevents communicators for transmitting false information? And once lies are circulated, what prevents the communication system from breaking down? Some authors have argued, following Reid, that human testimony does not have this problem because humans are naturally trustful and truthful, two properties that "tally together". However, if at the basis of testimony is non-verbal demonstration we can explain from its features how truthful testimonial transmission could arise in our hominin ancestors.
Learning a skill is a lot of work and current estimates indicate that about 30% of variability in complex skills such as chess are due to deliberate practice - the practice you set aside to iron out difficulties and focus on your weaknesses rather than your strength. In that respect, acquiring a skill is very costly. But once you acquire the skill, displaying it (demonstration) is in fact cheap; a lot more cheap than for someone who does not have the skill and would want to pretend she knows how to do it. In animal signaling, such communications that are cheap for the truthful signaler but hard (or impossible) for the deceitful signaler, this can give rise to a robust, reliable communication system that is not vulnerable to collapse due to deception.
Take the Oldowan technology as an example: hominids who have mastered the skill can produce useful, sharp flakes. Hominins who did not, produce flakes of variable sizes and sides that are less useful. When learners have several models available, they will preferentially learn from models who can demonstrably flake good tools.
This explains Buckwalter's and Turri's observations that knowing how is the norm of skill transmission. Knowing how to play harmonics, and being able to demonstrate to a learning guitarists, are closely connected. Thus, although my daughter cannot learn to play harmonics from the simple verbal testimony of the book and its pictures, she can learn them from demonstrations by experienced guitarists, combined with her own deliberative practice. This practice provides, as a handy byproduct, her with the fluency to cheaply signal she can play the guitar, which will allow her to transmit the skill to other players in the future.