Greg is a 2013 blogging resident, visiting us from his home blog over at On the Spiral. His residency will explore the theme “Individuality and Decision-Making” over several posts.
I think I have finally sorted out my uneasiness with the so-called deliberate practice hypothesis. Most Tempo readers will be familiar with deliberate practice (here, here, here & here) so I will just offer a quick refresher. The idea is that abilities that what we commonly perceive as talent are actually the result of painstakingly focused training. Anders Ericsson, whose research has provided much of the grist for the mill, summarizes deliberate practice as:
activities designed, typically by a teacher, for the sole purpose of effectively improving specific aspects of an individual’s performance.
I am not the only person to express mixed feelings about the concept. Others have noted that deliberate practice addresses the known better than the unknown, i.e. it applies to domains requiring mastery better than those requiring creativity.
But what is the alternative? Without an alternative, criticism carries the scent of sour grapes.
The advocates of deliberate practice generally juxtapose it with either a) belief in the value of innate talent or b) more mundane varieties of accrued experience. Their claim is that practice counts for more than natural talent, and in order to reach the highest levels of mastery that practice must take a specific form.
My objection to this framing, I realize now, is that deliberate practice is presented as the methodology that is active and therefore earned, while innate talent and non-deliberate(?) practice are portrayed as passive and unearned. Though never explicitly stated, the normative implications are only thinly veiled in much of the non-academic cheerleading on the subject.
I think it is a mistake to believe that learning must be deliberate in order to be active or earned. There is an another alternative that is equally active and equally intentional but not deliberate. That alternative is immersion. I mean immersion in the same way it is applied to learning a foreign language…the practice of actively placing yourself in an unfamiliar environment and exposing yourself to novel stimuli.
***
The distinction occurred to me while mulling over a cultural divide that I observe frequently in the crossfit community. For those who don’t know, crossfit is an exercise regimen that is described as: constantly varied, high-intensity, functional movement.
The high intensity part refers to the fact that most workouts involve a component that is timed, in which the athlete either a) attempts complete a prescribed sequence of exercises as quickly as possible, or b) attempts to complete as many repetition as possible in a prescribed time.
Constantly varied means that workouts are not programmed according to a formula. Exercises appear unpredictably from day to day, in many different formats and in many different combinations.
The emphasis on functional movement means that these timed workouts frequently include movements like the olympic lifts (the snatch and the clean & jerk), complex gymnastics movements and various other exercises requiring total body coordination. For example my workout today involves the following:
- Perform 3 rounds of the following sequence as quickly as possible:
- 3 rope climbs
- 10 power snatch at 135#
- 15 ring dips
- Score equals total time to completion
The olympic lifts in particular are immensely technical. The athletes you see in the olympics dedicate their lives to perfecting the nuances of just these two movements. Here are a couple quick videos of what I believe are the current world record lifts:
https://www.youtube.com/watch?v=0HduJeMruRQ
Athletes and coaches who focus exclusively on olympic lifting often criticize crossfit. They assert that performing such complex movements under stress (at high repetitions and under time pressure) promotes bad form, cultivates flawed movement patterns, and invites injury. They tend to be particularly critical of the practice of exposing beginners to these risks, and often argue that once bad habits are established they are nearly impossible to eliminate.
On first impression the division looks a lot like a case of insiders vs outsider. At one point in time, when crossfit was first arriving on the scene, that was probably the best way to describe it. However, today that is much less the case as there is significant cross-pollination between the two camps.
The tensions that persist could be better characterized as being between the immersion camp and the deliberate-practice camp. To avoid confusion – because they don’t use that terminology – I’ll refer to them as the new school and the old school.
The new school types throw people into the fire. They will always scale back the workout to the capabilities of the individual, but they won’t scale back the intensity. Even if you are lifting an empty bar and doing push-ups from your knees, the expectation is always that you push yourself as hard as you can. The implied belief is that people are capable of listening to their own bodies, and that movement patterns are accessible and can be reformed on the fly.
The old school types believe in getting it right the first time. They believe in learning through carefully controlled repetition, starting with simple building blocks and working up to more complex movements. They believe that bad movement patterns are difficult to deconstruct and limit long run potential. And they believe that premature immersion is dangerous, even sometimes negligent.
Old School Cultures and New School Cultures
The differences between the two schools of thought go beyond personal preference. Over the past 3+ years I have been a full time member at four different crossfit gyms. During that time I have noticed consistent differences in the way that individuals understand physical movement.
The new school types experience muscle memory as accessible and malleable. They pay more attention to proprioception and seem to intuitively know what is going on with their bodies. They know when to push harder and when their bodies need rest. They recognize when their body is moving properly and when something is wrong. And they are constantly experimenting and speculating about what works (for their own unique body) and what doesn’t.
Old school types seem to have less access to proprioception. They experience movement patterns as inaccessible. They tend to focus less on their own intrinsic qualities and more on their results. They are the people who value willpower and discipline, who enjoy grinding…pushing through the pain regardless of how they feel. And they tend to be the people who stick to a consistent program, motivated by achieving predetermined goals.
In short, old school types are more deliberate while new school types are more immersed. I have come to believe that these attitudes reflect stable aspects of personality….that the two groups differ in their actual subjective experience of physical exercise.
Similar themes can be seen in the style of coaching offered at the old school gyms vs new school gyms. New schools coaches tend to be much better at breaking down movements, explaining their nuances and adapting their expectations to the individual. Old school coaches tend to take a harder line, interpreting expectations and standards more literally. The former tend to promote a more collegial atmosphere where members are constantly coaching each other, while the latter emphasize rules and promote an atmosphere of deference to the coaches and/or the program.
You can tell a lot about the culture of a crossfit gym from the type of conversation that goes on before and after a workout…
At some gyms it is all about performance:
- what time you expect to get
- what time you got last time you did this workout
- what your goals are
- what competition you are training for
At other gyms the conversation is full of speculative expectations:
- what parts of the workout you are dreading most/least,
- what parts of you body are particularly sore or rested
- what tentative strategy are you taking into the workout
- breaking up reps
- where you expect to push vs where you will need to rest
- which pairings of exercises may present unusual challenges
For the former group the post workout conversation often revolves around success or failure, elation or disappoint. They are excited when they achieve their goals.
For the latter group the post workout conversation consists of comparing expectations to actual experience. In short, the members of the latter group build up propositional mental models before the workout, and afterwards they compare those propositional models to their experienced reality. They are most excited when they encounter a curve-ball that suggests ways they might improve those models.
Performance vs Meta-learning
One way of understanding these two attitudes – which nicely parallels the fox and hedgehog archetypes – is that deliberate practice is a performance/learning process whereas immersion is a learning/meta-learning process.
Along the same lines we might say that deliberate practice emphasizes externalization (output) while immersion emphasizes internalization (input).
In other words, immersion exposes you to the complexity of the environment all at once. When you immerse yourself in a foreign language the active emphasis is on internalizing and decoding as much of the language as you can, in all its subtlety and nuance. It is largely taken for granted that if you understand the language then you will also speak intelligibly.
By contrast, mastery of a musical instrument or a competitive activity like chess is all about mastering externalized behavior. Through practice you will eventually intuit some of the principles underlying your practice, but understanding the theory alone doesn’t do you much good if it gets in the way of the performance.
This is why domains biased towards deliberate practice tend to be so concerned with eliminating bad habits. The emphasis on performance deprives students of the meta-learnings that would otherwise allow them to actively deconstruct bad habits. Immersion inevitably leads to bad habits, but it also fosters accessible meta-learnings that allow those bad habits to be deconstructed.
Manipulative vs Appreciative
It should be self-evident that deliberate practice primarily shapes manipulative mental models, given that it directly addresses the mental faculties underlying specific instrumental behaviors. Immersion primarily develops appreciative mental models. In describing appreciative models Venkat writes:
My instinctive preference for complexity made sense from the perspective of purpose. I like purposeless models. Or equivalently, models that exist before clear purposes do. It makes sense that such models are often more complex. It isn’t that I like complexity for its own sake, but that I like purposeless models, which are often complex. They help me appreciate something on its own terms, rather than through the lens of something I want to achieve.
This non-purpose (or universal purpose or meta-purpose) is appreciation. An appreciative model is a model you use simply to make sense of a situation.
I like the idea of appreciative models having a meta-purpose. Immersive practices construe goals at a high level. This is quite different from having no purpose at all, though in our left-brained culture it is easy to mistake meta-purpose for lack of purpose.
One of catchphrases used to describe crossfit training is general physical preparedness (GPP). GPP is the antithesis of sport specific training methodologies, which bring instrumental faculties to the foreground. The notion of “general preparedness” complements Venkat’s idea that the meta-purpose of an appreciative models is to “make sense of a situation”.
Manipulative models may be more efficient when you have the ability to initiate instrumental behaviors on your own terms, but they are not so useful when you need to respond to the environment. If you need to be prepared for whatever challenges the environment might present, then you need an appreciative model.
If immersion appears passive it is only because the stimuli originate in the environment. However, the act of immersing oneself in a novel environment is no less intentional than the act of breaking out a chess board. Adaptation occurs just the same whether it is stimulated directly by practice itself or indirectly by opening oneself to a particular stream of stimuli. And the depth of immersion can be modulated just as one would the volume of deliberate practice.
Contextual vs Context-Free
Deliberate practice techniques are designed in a certain sense to be inaccessible. You learn it right the first time through simple repetition. There is an emphasis on consistency. Beginners learn piece by piece, mastering one aspect of a movement before moving on. The old-school culture described above has a distinct “just do it” quality to it.
In physical training all of this serves to bypass conscious thought processes and push the movement pattern directly into muscle memory. What is true of habits is also true of performance – you are at your best when you are in the zone, just acting…without thinking too much about exactly what you are doing.
So where does that leave the conscious mind?
Venkat recently wrote about two classes of habits:
…every habit is actually two intertwined habits. There is a habit of thought and a coupled habit of action.
- A habit of thought is a set of coupled patterns of thought and a practiced ability to switch among them appropriately and effectively.
- A habit of action is a learned pattern of physical behavior involving sensory processing and physical movements.
Both are context-dependent. The former is dependent on your immediate state of mind, the latter is dependent on your immediate environment.
By minimizing the amount of input material provided to the conscious mind, the old school practices produce habits of action that are pure muscle memory.
This leaves the conscious mind free to construct habits of mind intended to insulate the performer from the external context. In the extreme, when athletes enter the zone they report that the outside world fades away and they experience a kind of tunnel vision. In courting this state many athletes construct habit of mind that take the form of superstitions, visualization routines, affirmations, and various other ritualized preparations. All are practices meant to promote high fidelity performance without regard to the particular context.
Immersion accomplishes the opposite. The conscious mind actively engages in the process of distinguishing signal from the noise. The appreciative mental models that are constructed map correspondences between the external environment and learned habits. The primary concern is for appropriateness of behavior in context rather than reproduction of practiced behavior with perfect fidelity.
Nesting the Two Approaches
Though the deliberate practice hypothesis doesn’t get my blood stirring on its own, understanding it as counterpoint to immersion helps me understand how it can be adopted judiciously without stripping all the joy out of learning.
For me, immersion is clearly the more productive approach, but it is possible to nest deliberate practice within a broader immersion framework. The choice then is one of priority…of which approach leads and which follows.
I may prefer learning a foreign language through immersion, but losing my accent would likely require some focused attention. Similarly, I enjoy crossfit style training precisely because it is open-ended, appreciative and contextualized, but that doesn’t mean I neglect more deliberate training methodologies when specific skills are in need of refinement.
Emphasizing immersion to the exclusion of deliberate practice seems to produce rapid progress followed by diminishing returns. Incorporating some elements of deliberate practice allows these obstacles to be overcome. And when adopted within the context of an immersion based practice, these elements of deliberate practice seem less onerous.
It would seem the inverse arrangement is equally viable. I have noted recently that Cal Newport – one of the staunchest advocates of deliberate practice – has been hesitantly incorporating more immersive practices into his productivity advice. I’m sure he would insist that he isn’t abandoning deliberate practice, but then that just demonstrates the point that an approach dominated by deliberate practice can safely incorporate elements of immersion without being corrupted.
About Tempo
I definitely agree with your outlook; the problem with deliberate practice people is that they think their own specialized metric is the only way somebody ought to judge success. On the other hand, the concept of meta-learning still has some lingering questions for me:
It seems to me that when we’re learning a very specific ability such as chess, weightlifting, soccer, or firefighting, that we’re learning about what is for all intents and purposes a closed system and therefore can get reliable feedback due to regular statistical patterns. That said, if we were teaching ourselves chess with no books and no teachers, we’d have to engage in a lot of meta-learning ourselves, but when we have teachers, they already know the regularities and can tell us things that allow us to skip meta-learning.
On the other hand, if we are dealing with something without such regularities, such as running a business, we have to engage in some kind of meta-learning. The question that constantly plagues me, however, is how meta-learning can really help us here. The statistics show that success in these fields, beyond basic competence, is completely unpredictable. My guess is that we cannot predict, but we can nonetheless exploit some kind of cheap trick through an act of creativity or synthesis. This may work because the extreme randomness comes from the inevitability of Black Swans, and those come in the form of unanticipated, but nonetheless self-created “game changers.”
That leaves me with one other speculation that bridges both of these together: that the contrast between performance/learning and learning/meta-learning is equivalent to that between analysis and synthesis. Both are necessary to sustain each other, but when we’re learning the basic skills of chess or accounting or weightlifting, we begin almost entirely with analysis. When we get to a higher level, however, synthesis plays more of a role.
That last sentence is not just me hand-waving: it goes back to when I had to qualify my use of the phrase closed system with “for all intents and purposes. Even a game like chess is not really “closed” because there are two players who are trying to get inside one another’s heads (OODA loop, anyone?) and that means that unless the game is truly exhausted (which could be done if you had a computer that was SO powerful that it ran a MiniMax algorithm that took every possible move into account), there is still new information to be acquired about the domain of chess. That means that upon getting good enough to play with the best of the best, you’re now on a path in which you have to engage in meta-learning to discover new cheap tricks that exploit the missing information.
But if we get away from chess, and into a more abstract field, like say… art, philosophy, or science; we have the same effect. A need to start from performance and slowly use more and more meta-learning as we see fewer and fewer regularities. The difference between this and chess is that the system really is not closed–there is no combinatoric computation that will shut it down once and for all (unless one thinks of all this as being a swarm of narrative-rational decision makers pushing the universe towards its inevitable heat-death.)
It’s also important to note that in the earlier stages, a convoluted meta-learning based cheap trick may not make much sense. A lot of amateurs have “brilliant” ideas but then quickly learn that they’re not used because they really don’t work. Instead, as you get better at a game like chess, you learn that the truly “brilliant” and “game changing” moves are ones that seem novel if and only if you understand the nuances of the game. Far from a Hail-Mary pass, a truly game changing move is one that actually displays an incredibly solid and nuanced understanding of the actual fundamentals.
Now, for the last part of this convoluted idea: unifying this all with the cheap trick. If we think about this as a spectrum of low entropy to high entropy, we can now think about performance, learning, and meta-learning in a new way:
Low Entropy: cheap tricks are easily visible/accessible, can gather them up and automate them. Most productive to focus on analysis (performance/learning). At the same time, the “high entropy” cheap tricks will not be easy to see because they are almost impossible to find if you can’t reduce them to patterns based on more basic cheap tricks; in other words, there’s too much noise.
High Entropy: cheap tricks hard to find, must be dug up with synthesis (learning/meta-learning). Can be done, however, by using your more basic cheap tricks to eliminate a significant amount of noise.
I’ll see if I can further refine this soon.
Pretty sure I followed most of it. I want to get one small quibble out of the way before responding to the meat of your comment…
Just because comprehensive analysis of a closed system is possible doesn’t mean that it is necessarily the best way of trying to solve a problem. If you have a supercomputer capable of calculating every possible contingency then the point is moot, but if you are a human being with limited calculative capacity then you might get more mileage out of a few general principles and a bit of experience, even when a deeply analytic approach is viable.
Getting back to the general theme – I agree that deliberate analytic approach generally applies better to closed (or nearly closed) systems, while meta-learning is more necessary in open systems. But I also think you can adapt either approach to either situation.
Taking chess as an example, some people may prefer to learn largely through analysis – studying specific strategies, individual moves, formations, etc – and then building up to larger concepts, while other people may move in the opposite direction – going with their gut and then delving into the technicalities of what they were doing after the fact. If you have a coach who is capable of identifying these gut intuitions and relating them to the technical cannon, then I don’t think the latter approach is necessary inferior.
We could refer to the two as learning from the inside-out vs learning from the outside-in (this is just a rephrasing of what I mean by deliberate practice and immersion respectively).
Outside-in learning seems to get the short end of the stick in codified fields because most teachers/coaches lack proficiency at moving in that direction…so in practice it may look like reinventing the wheel. But if you have access to a good mentor, then this approach can be much more efficient because starting with the big concepts gives you the context necessary to decide what is worthy of further analysis…and to avoid filling your head with the rest of the trivial details that someone packed into a textbook in order to up the word count.
The reverse is also possible. Even in an immature or open-ended field, it may be beneficial to master what is known before moving outward. This is more or less what Cal Newport has begun to advocate on his study hacks blog. The problem with this approach in practice is that the existing knowledge may be poorly codified such that a comprehensive introduction to what is know may not be available.
On the first technical point: I communicated badly. I didn’t mean to suggest that you should always use statistical algorithms on a closed system; I meant to suggest that it has certain statistical properties that show that stable inference is possible and that this seems to be tied in with the fact that winning can always be done in theory through sufficient computational power (I don’t think the same thing can be said about open systems, since you have some Maxwell’s-Demon-esque problem where calculating one Black Swan opens a back door for another to come in). A better example than Chess might be Go, where the combinatorics are so vast that raw data crunching is still extremely ineffective, but it nonetheless is a game where skill is stably built up because the fact that it’s a closed system means that patterns do emerge.
It seems that both “inside out” and “outside in” learning do work because in both cases you’ll be getting regular feedback from the system, but as an outside-in learner, I definitely saw weaknesses after a certain point. I’m no chess player, but I used to play StarCraft once in a while (I was an average player, as proven by my decent but not great ladder ranking.) My biggest weakness in hindsight was ignoring the collected wisdom of veterans and continuing to try weird strategies that made sense to me but didn’t make sense to people who knew the subtleties of the game better.
I imagine, however, that that strategy works better in an open-system. I’m not suggesting that one start a business based on a Hail-Mary pass, but it does seem that once you’re in extremistan, there’s a lot more potential gain from being more “out of the box”. I actually share your views on fitness for this reason, because I think that the body is far more complex than a game of chess and people overvalue some number they achieve on a single exercise.
In both cases, however, there is a need for both inside-out and outside-in learning, and it boiled down to me to a simple rule I eventually noticed. Truly “game-changing” strategies do not come from breaking all of the rules, but from having such a deep understanding of the rules that you can exploit a subtlety that nobody else sees. In the case of something like chess, this is relatively straightforward. In the case of something like starting a business or inventing a new device, it seems to require a very complex process of exploring insights and taking gambles on them. These insights, of course, rely on mastering some stable patterns, and this is where Newport’s ideas are relevant. There’s more potential for a “game change” because the rules are much less stable, but the flip side is that your own strategies are also much more susceptible to “game changes”. In addition, there’s the issue of noise, which I’m not sure how to handle.
The question that has been on my mind for a while now is more one of epistemology. Predictability is zilch, but that doesn’t mean you should throw darts at a board. The subtle difference between traders and stock-pickers is that traders are engaging in some sort active dialectic, whereas stock pickers are basically leaving everything to chance. The fact that traders eventually “blow up” doesn’t seem to mean they’ve just been gambling–instead, they appear to me as going through the lifecycle of the “narrative rational” decision maker.
In both open and closed systems, we need to coherently look for and exploit patterns, but on the whole each of those has a different process for doing so. To go back to the entropy idea from earlier, a skill like chess seems to have a relatively simple trajectory from low entropy to high entropy. The pattern of entropy in an open system would look significantly different, but to just label it all “high entropy” doesn’t seem right. Instead, I imagine that there is some sprawling pattern in which entropy is distributed in some kind of fractal pattern; so for example, one might be able to invent a better jet engine through lots of deliberate tinkering and then use that as the basis of a new business. In this case, they’ve exploited the lower entropy of the field of jet-engine building to gain an edge in the higher entropy area of entrepreneurship.
The one fundamental difference I may see from thinking this all through is that in both cases, the low-hanging fruit is being picked, but in one case, there is something resembling an end goal that is constant (for all intents and purposes), where in another one, the progress gained from new exploits has to be coupled with an overall drive towards adaptation. In those cases, your real metric is survival and degrees of freedom; everything other metric should be adopted or discarded based on whether it helps you push towards those two main metrics.
Maybe instead of open-ended and closed systems we should talk about the way in which a system is evolving. A system that is truly closed and static, like chess, could be approached either way…and the approach that ends up being more successful is likely a function of circumstantial variables such as the learner’s learning preferences, the form of teaching available, and the particulars of the domain (e.g. Go might rely more on intuitive insight whereas blackjack will rely more on mastery of straightforward rules).
But if we look at open-ended systems, some offer opportunities for straightforward advancement while others offer opportunities reorganization. Some industries are looking for evolutionary progress while others are ripe for disruption and in need of a new constraints (imposed outside-in).
So I don’t entirely agree with the claim:
“Truly “game-changing” strategies do not come from breaking all of the rules, but from having such a deep understanding of the rules that you can exploit a subtlety that nobody else sees. ”
But I agree with you more when you start talking about discovering patterns. What I think you can say is that you need to have experience with patterns of similar complexity. It is certainly possible to draw a conceptual metaphor between disparate fields and exploit that pattern without mastering the rules of the target field. It’s less likely that you will just get up off the couch one day and realize that you have a brilliant insight out of the blue.
Regarding Crossfit, I’m not sure that I agree with the claim that Crossfitters have greater access to knowledge about their physical condition. If this were true, why do the rates of injury seem so high?
In addition, I’ve seen a lot of bad “coaching” from participants who aren’t sufficiently skilled–what one of my BJJ mentors calls “white belt on white belt crime.”
Jay, I don’t mean to imply that crossfitters as a demographic group are more aware of their bodies. I was trying more to portray the “new-school” archetype than to say anything about crossfit per se. And as I mentioned, there are plenty people who I perceive as old-school types who now participate in crossfit and who are members of crossfit gyms.
With regard to “white belt on white belt crime”, I’m not sure that necessarily has anything to do with body awareness. The new school culture/community does enable it, but then again, every community has its own flavor of ignorance, naivete, and over-enthusiasm. I’m certainly not arguing that crossfit neutralizes such concerns.
Greg,
I think that makes sense–thanks for clarifying. I would agree that the middle way seems best. One of the things I really appreciate about the current gym I train at, is that it’s 60% structured, very technical instruction, and 40% live application with really top notch training partners.
Have you checked out The Inner Game http://theinnergame.com/ ?
One of its key phrases is “Trying fails, awareness cures” Fritz Perls It helps to bridge the gap between practice and immersion.