Legitimate peripheral participation: the book and the idea

Download MP3
"Legitimate peripheral participation" is based on observations about how novices learn in the presence of experts. The novel bits are that novices learn better from fellow novices than from experts, that we need to pay attention to the difference between teaching and learning, that passive observation is underrated, and that toy projects (code katas, "advent of code", etc.) are perhaps not all that useful – at least to novices. In the "applications" section of the episode, I offer an off-kilter suggestion about pair programming.

Welcome to oddly influenced, a podcast about how people have applied ideas from *outside* software *to* software. Episode 21: Legitimate peripheral participation

My source this week is /Situated Learning: Legitimate Peripheral Participation/ by Jean Lave and Étienne Wenger, published in 1991. It comes from roughly the same tradition as Orr’s /Talking About Machines/, so there’s some continuity with the previous episode. (By the way, in that episode I said that Orr did his research around 1990. Judging from the citations in Lave and Wenger, it was more like 1985.)

Legitimate peripheral participation is associated with the idea of “communities of practice”, also due to Lave and Wenger. This book leaves communities of practice relatively undefined, concentrating about how newbies to such communities learn, and gain their identities as community members. I’ll do Wenger’s book, /Communities of Practice/, next time.

The first order of business is to say what “legitimate peripheral participation” means. As a very rough first approximation – and feel free to let this definition just flow over you, because I’ll give examples later – they’re talking about the process by which an apprentice becomes a master. That’s by doing actual work that actually matters, as a person who’s recognized (by everyone who counts) as someone who’s *supposed* to be doing that kind of work, but a person who starts out doing only a part – typically a less important or less risky part – of the set of tasks they’ll eventually learn to do. “Peripheral” is specifically not supposed to be opposed to “central”, in that there’s not necessarily a single set of tasks which are necessary and sufficient to be, say, a master tailor. Instead, it’s supposed to suggest a kind of opening through which the apprentice gains access to what’s needed to learn a practice.

I think “gradual” captures the idea better than “peripheral”, but hey: whoever gets there first gets to pick the jargon.

Things become more clear when Lave and Wenger describe case studies. In all cases, they use the word “apprentice”, but note that none of the case studies are of the European craft guilds you’re most likely to be familiar with. Don’t get hung up on a formalized “apprentice-journeyman-master” progression – that’s just one particular kind of legitimate peripheral participation.

The closest of their case studies to European craft guilds is the Vai and Gola tailors of West Africa. Masters accept apprentices who go to live in the master’s house, in some sort of contractual relationship between two families. The apprentices start by learning very basic tasks (like how to iron clothes). When they start learning tailoring, they learn *backwards*. That is, the first thing they learn to do, is the last thing done in the production process, like attaching buttons and hemming cuffs. After they’ve mastered that, they move on to sewing pieces of clothing together. The last thing learned is the first thing done when making a garment: cutting out the cloth. This ordering has two benefits:

1. First, they start doing the things easiest to *un*do if they do them wrong. It’s easier to undo a button than un-cut a badly-cut sleeve.

2. Second, this allows them to focus on how the previous step, that they don’t know how to do yet, contributes to the step they’re doing now. Learning how pieces are sewn together shows why they’re cut out the way that they are.

As is common in all the case studies, there’s a lot more *learning* than *teaching*. Apprentices learn by talking to each other (in a way reminiscent of last episode’s copier repair techs), and they learn by watching the masters at work. It’s not stated specifically, but my impression is there’s a lot more of the apprentice asking “why did you do that?” than the master pontificating, “As you see, I do this in the following way. That is because…” That actually matches the widespread observation that experts typically have a hard time explaining why they do what they do – they simply act expertly – so why should we expect them to know what an apprentice needs at any given time?

So, even though Lave and Wenger warn against it, let’s break the phrase into its constituent words:

* legitimate. Apprentices are supposed to be there, contributing to the master’s enterprise.
* peripheral. They start by learning low risk, low demand tasks, then moving on, rather than, say, doing everything badly, then learning how to do everything a little bit better, and so on.
* participation. The vast bulk of the learning is gained by doing, rather than by being taught. Especially, they don’t do “toy” projects that are discarded; they contribute to products that get sold.

Let’s contrast that apprenticeship to the formalized apprenticeship of unionized supermarket meat cutters in the USA around 1970.

The first part of learning was partly in a classroom setting (with book learning and written exams) and partly in a shop where people did toy projects. Those toy projects were frequently outdated. For example, one of the first thing apprentices learned was how to sharpen a knife. Seems useful, but the problem is that, in modern supermarkets, sharpening was outsourced. Someone delivers sharpened knives and picks up dull knives to be sharpened.

When an apprentice gets an actual job, they are put to work on short, repetitive tasks. They are not encouraged to talk to each other or to the journeymen – because time is money to their manager. Often, they can’t even *see* other people doing different jobs:

“Some meat departments were laid out so that apprentices working at the wrapping machine could not watch journeymen cut and saw meat. An apprentice’s feeling about this separation came out when a district manager in a large, local market told him to return poorly arranged trays of meat to the journeymen. ‘I’m scared to go in the back room. I feel so out of place there. I haven’t gone back there in a long time because I just don’t know what to do when I’m there. All those guys know so much about meat cutting and I don’t know anything.’”

Moreover, apprentices don’t move on from their first task once they’ve mastered it. They may continue to do the same task for years, only moving on when a new apprentice is hired. In larger stores, the efficiencies of division of labor may mean that no one ever achieves full mastery.

So there’s a big difference between tailors and meat-cutters: even though meat-cutters have an explicit *teaching* program, they provide much less opportunity for learning. I suppose the same sort of thing could have happened to the West African tailors, but it doesn’t seem to have. Certainly the European craft guilds had lots of exploitation of apprentices, especially since it wasn’t in the masters’ interest to train up people who could compete with them.

I’m reminded of a tweet by “The Angry Norwegian”: “I once had a team member who got visibly upset after I facilitated our first mob programming session. He stood up and said: "Look, I've spent many years working hard to get to where I am today, and I sure as hell am not going to share my knowledge here for free!”

The difference, I think, comes down to community and identity. I think the complaining team member didn’t see himself as a member of a community that assumes, as a matter of course, as part of the circle of life, that you train the people who will someday replace you. (In the book’s jargon, a community that “reproduces itself”.) He sees himself as detached, like the people who assign tasks to meat cutters, who are not meat cutters themselves. They’re not interested in meat cutters as a community of practice, or in the future of meat-cuttery. They’re interested in minimizing cost, maximizing output, and moving on to a better management job.

Another case study is Yukatec midwives, and it introduces some interesting twists. There isn’t exactly a formal role for apprenticeship. Only daughters and granddaughters of midwives become midwives, but it’s a gradual thing. This passage is worth quoting at length:

“Girls in such families, without being identified as apprentice midwives, absorb the essence of midwifery practice as well as specific knowledge about many procedures, simply in the process of growing up. They know what the life of a midwife is like (for example, that she needs to go out at all hours of the day or night), what kinds of stories the women and men who come to consult her tell, what kinds of herbs and other remedies need to be collected, and the like. As young children, they might be sitting quietly in a corner as their mother administers a prenatal massage; they would hear stories of difficult cases, of miraculous outcomes, and the like. As they grow older, they may be passing messages, running errands, getting needed supplies. A young girl might be present as her mother stops for a postpartum visit after the daily shopping trip to the market.

“Eventually, after she has had a child herself, she might come along to a birth, perhaps because her ailing grandmother needs someone to walk with, and thus find herself doing for the woman in labor what other women had done for her when she gave birth; that is, she may take a turn … at supporting the laboring woman… Eventually, she may even administer prenatal massages to selected clients. At some point, she may decide that she actually wants to do this kind of work. She then pays more attention, but only rarely does she ask questions. Her mentor sees their association primarily as one that is of some use to her. (“Rosa already nows how to do a massage, so I can send her if I am too busy.”) As time goes on, the apprentice takes over more and more of the work load, starting with the routine and tedious parts, and ending with what is in Yucatan the culturally most significant, the birth of the placenta.”

As a side note, my dad learned carpentry through a formal apprenticeship in Germany, and the way I informally apprenticed to him mirrored this midwife style. Not a whole lot of conversation about the work. Granted, I never got any good at carpentry, but that’s probably more due to my complete lack of innate talent than to that style of legitimate peripheral participation.

Lave and Wenger’s emphasis on learning as distinct from *teaching* shows up again in their Alcoholics Anonymous case study. (I’ll abbreviate “Alcoholics Anonymous” to “AA” from now on.) They interpret AA as being primarily about changing a person’s self-identity from a “drinking non-alcoholic” to a “non-drinking alcoholic”, where “alcoholic” is considered a non-reversible state once you enter it. That is, the only question that matters in AA is whether you’re drinking or not drinking. (Note: my impression is that “recovering alcoholic” has mostly replaced “non-drinking alcoholic” in the last 30 years, but I believe it’s still canon that you can never actually *recover*.)

This shift in identity is done via… legitimate peripheral participation.

AA is organized around meetings in which people largely tell stories. You’re legitimate if you attend a meeting: everyone there will assume you’re an alcoholic like them – and so welcome you – even if you are there because a judge sentenced you to come. You participate when you tell stories from your life. The “peripheral” box is checked off by your progression through the “12 steps” that are explicitly laid out as the path to sobriety. But, for purposes of this book, it’s more relevant that you gradually learn to tell your story in the way that AA thinks you should.

This learning is mostly done through, again, observation. Newcomers watch other people tell their stories, and learn from that how to tell their own stories. Another long quote:

“In AA, personal stories are told for the explicit, stated purpose of providing a model of alcoholism, so that other drinkers may find so much of themselves in the lives of professed alcoholics that they cannot help but ask whether they, too, are alcoholics […]

“Telling an AA story is not something one learns through explicit teaching. Newcomers are not told how to tell their stories, yet most people who remain in AA learn to do this. […] All members are encouraged to speak at discussions and to maintain friendship with other AA members. In the course of this social interaction the new member is called on to talk about her own life. […] Usually, unless the interpretation runs counter to AA beliefs, the speaker is not corrected. Rather, other speakers will take the appropriate parts of the newcomer’s comments, and build on this in their own comments, giving parallel accounts with different interpretations, for example, or expanding on parts of their own stories which are similar to parts of the newcomer’s story, while ignoring the inappropriate parts of the newcomer’s story. […]

“AA’s must learn to experience their problems as drinking problems, and themselves as alcoholics. Stories do not just describe a life in a learned genre, but are tools for reinterpreting the past, and understanding the self. […] The initiate […] comes to understand herself as a non-drinking alcoholic, and to reinterpret her life as evidence.”

Here we once again have people learning a skill via practice, observing skilled practitioners, with the driving force being to move from the periphery to the inside. What the AA case study emphasizes is that learning skills is inexorably intertwined with *becoming a new person*, at least with respect to a particular community. That is, two West African weavers might have completely different roles within their respective families, but within the community of weavers, they’ve become people notably similar to each other in very many ways.

Moreover, they haven’t been *taught* to be so similar – not exactly – but rather they’ve learned to be similar because, I assume, humans are evolved to be learning machines and also to seek a group identity.

Lave and Wenger do a little dance where they don’t exactly criticize the type of teaching we’re used to from school, but it’s pretty clear they don’t think much of it. Their telling point is that teaching inducts learners into a particular community of practice: that of “student who graduates”. In the meat cutter example, skill at graduating doesn’t necessarily translate to skill at working with meat. In software, skill at getting a computer science degree is only loosely correlated with being able to program, much less to program in a team.

The last case study is of the training of quartermasters on US naval ships, derived from the work of Edwin Hutchins. There’s not a whole lot new in this one. And, in any case, Hutchins’ own book, /Cognition in the Wild/, deserves its own episode.

The most interesting bit about quartermasters is that their training starts at the beginning of a temporal process, rather than learning first the tasks done last, the way our tailors did. The first thing novice quartermasters learn is gathering data using various devices. Only after mastering that do they move to creating and analyzing summary data. Why that order? Our authors quote Hutchins as saying, “The fact that the quartermasters themselves follow [the] same trajectory through the system as does sensed information […] has an important consequence for the larger system’s ability to detect, diagnose, and correct errors” – but they then use ellipses to remove Hutchins’ elaboration. They further quote Hutchins that “movement through the system with increasing expertise results in a pattern of overlapping expertise, with knowledge of the entry level tasks most redundantly represented.” Perhaps this has to do with operating a ship in battle. It does no good to have people expert at processing data if there’s no one left who’s any good at acquiring that data. It’s better to have good data that’s incompletely or weakly processed. So, although Lave and Wenger don’t really give enough information, my hunch is that the order is again about reducing system risk.

So what does this suggest for learning a profession of software? Because I think it’s broadly agreed that people do not leave the Universities ready to become Programmers or (even more) Testers or Product Owners or whatever.

It seems to me that our approaches to bringing up new people range from benign neglect – sink or swim – to various forms of active mentorship.

I’ve always thought mentorship was obviously The Right Thing, but Lave and Wenger made me wonder. Remember that they emphasize that apprentices learn from other apprentices more than they do from masters. And even when they learn from masters, Lave and Wenger note an apprentice is more likely to learn from some *other* apprentice’s master than from his own. I expect that’s because the other master doesn’t have direct power over you, so is less intimidating – it’s easier to admit you don’t understand something.

Another hurdle for mentorship is that it so easily slips into being about me, the expert, *teaching* rather than about you *learning*. That means what you’re learning is how to be a student, how to be a subordinate, how to behave in one-on-ones, not how to perform tasks.

It’s probably mostly bad for a mentor and apprentice to meet away from the computer, rather than “coparticipating” in doing the work. So mentorship ought to mainly mean working together. I used to suggest pairing. However, because of the power dynamics, it’s probably better to learn through mob or ensemble programming. Mobbing is closer to the peer learning our authors emphasize, and people tell me that mob programming is more comfortably adopted than pairing.

So maybe your apprentice doesn’t need you so much. If you’re a mentor, I’d advise not thinking your job is to teach, but rather to set the stage for learning. It’s to remove impediments preventing your mentee from learning from *other* people.

This also suggests that benign neglect – giving newbies tasks and letting them find their own way – isn’t necessarily so bad. A motivated employee is a learning machine, if encouraged or allowed to be.

Sure, it’s bad for a newcomer to be hunched over a monitor in an open plan office, wearing headphones, seeing nothing *but* the monitor, learning from other people only by watching arguments in Pull Request reviews. That’s bad because the newbie only sees the *result* of practice, not practice itself. Even arguments in reviews tend to be “why did you do that?” not “*how* did you do that?”

So it’s up to the motivated newcomers to take charge. They need to find people – peers, approachable experts – to talk with. Better, they need to organize coparticipation. Do some sneaky mob programming, justified however they can. Our authors point out that apprentices will often take advantage of “benign community neglect” to “configure their own learning relationships with other apprentices”.

Apprentices do need to *observe* people with more expertise, but not necessarily to be super participatory. Lave and Wenger describe apprentices sitting next to masters at two-person benches and give the impression that not much talking is done.

When I read that, I was reminded of those horrible pair programming sessions where one person types and the other person passively watches. Certainly those are not very much fun – or very productive – when people are at approximately the same level of skill. However, it might not be that bad for an apprentice to quietly observe someone programming – not to participate in the design, but to take notes about practice. It might not even be bad if those notes don’t generate questions answered by the expert, but are rather taken away for discussion among peers, discussions that lead to trying things out. (Such discussion is probably good even if the expert does talk; remember: experts rarely have great insights into their own practice.)

I’m uncomfortable, I admit, with recommending you try what’s widely known as awful pairing, but the lesson of the book – again – is that people will learn if they know that’s a big part of their job, if they are given real work to do, if that work is such that failure isn’t a big deal, if the work gets progressively more challenging, and if they get to co-work and co-learn with other people.

Just don’t force the poor newbie to watch some expert *all day long*.

I have other comments about changing practices, and especially the consequences of learning identity, but they seem too half-baked to make it into this episode. They might come up in the next episode, about communities of practice, or I’ll post them at social.oddly-influenced.dev. In the meantime, thank you for listening.

Legitimate peripheral participation: the book and the idea
Broadcast by