A Pattern Becomes a Language: Part 3 of 3

The second piece in this series,When the Practitioner is Gone, left a question open on purpose. Hunter's pattern language was a few months old, tested against a handful of colleagues who knew him, in a room that had met three times. Is that good enough? Is it precise enough to be useful and checked closely enough to be trusted? I didn't answer that question, because it's still being answered, gathering by gathering, by people doing the work right now. But I can show you what the far end of that process looks like. Not a few months in—thirty years in. Also, not a handful of colleagues—hundreds of people who never met me and never needed to. This is that story. The [first piece in this series](#) named a single pattern: a teacher hands a student a marker instead of correcting him, and that move—tested, adjusted, passed to other teachers—becomes something nameable. That's useful, but it's small. It's one move for one kind of moment. Christopher Alexander, the architect who gave us the term, never meant 'pattern language' to describe a single pattern. He meant a *language* a connected set of patterns that cover a whole domain of recurring problems, each one referencing the others, the way words in a sentence depend on each other for meaning. No single pattern builds a house. A language of them does. So the real question Part 2 left hanging isn't just 'can one pattern be trusted?' It's 'can a whole language be trusted—one built up over years, out of dozens of small moves, tested not by one person's memory but by strangers using it in situations the original practitioner never saw.' Strategic Doing is my own answer to that question. Not because I set out to prove a point about epistemology. Because I had no choice. I couldn't explain what I was doing, and I needed to. In 1993, in Oklahoma City, I was standing in a room with people who needed to act together and had no way to start. There was no rulebook for this. Strategic planning, the tool I'd been trained in, assumes someone is in charge, holding a map, able to direct people toward a plan. Nobody in that room was in charge of anybody else. I was improvising, the way the teacher in Part 1 was improvising, and the way Hunter was improvising on the Sohio Tower site in Part 2. I could not have told you, in that room, why I asked the questions I asked. Over the years that followed, working in city after city, I started noticing the same shape of conversation working, over and over, in wildly different rooms. Naming that shape—turning 'something that seems to work' into 'a pattern I can describe and hand to someone else'—is exactly the circuit from Part 1. » Act. » Notice the shape. » Name it. » Hand it off. Watch it get tested by someone else, in a context you didn't choose. I have a document that shows that circuit running, version after version, for fifteen years. It's a single figure. I think it's the clearest evidence I have that this actually works. Here's the Strategic Doing Cycle, in five versions, drawn from my own doctoral research: Figure: Evolution of the Strategic Doing Cycle, 2005–2020. Five successive versions of the same underlying process, revised through use in workshops ranging from NASA scientists to neighborhood activists. Version 1, 2005 to 2006, is labeled with verbs: Map, Focus, Align, Execute. Each one carries a technical sub-phrase underneath it — 'identify and align resources to specific initiatives,' 'execute and measure results.' This is language built by a practitioner, for practitioners. It assumes the room already speaks it. Version 2, a year or two later, keeps the same four-part structure but swaps out the verbs: Find, Focus, Launch, Learn. That's not a redesign. That's a correction. Something about the first set of words wasn't landing, and I changed it. By Version 3, from 2008 to 2011, the verbs are gone entirely. What's left is four plain questions: *What could we do together? What should we do together? What will we do together? What will we get back together?* No technical vocabulary left at all. Just the questions themselves. Version 4, in 2013, keeps that structure and sharpens the language further. Version 5, from 2014 on, adds color and turns the wheel into the quadrant most people recognize today. Fifteen years. Five versions. The same underlying process, worn down from a technical model into something a room full of strangers can pick up and use in the first ten minutes of a conversation. I didn't do that alone in an office. I did it because every workshop was a test I couldn't rig. A room of NASA life scientists and a room of neighborhood activists don't forgive language that doesn't work. If a phrase confused people, I found out immediately, in the room, from their faces and their questions. That's what 'Map, Focus, Align, Execute' turning into 'What could we do together?' actually is: a pattern, checked against strangers, correction after correction, until what survived was language precise enough to travel to rooms I would never be in. The wheel got simpler. The skill it takes to move around it did not. Four questions are easy to say. They are not easy to run—in a real room, with real people, under real-time pressure, when the honest answer changes what happens next. Simple language and easy practice are two different things, and it would be a mistake to confuse them. That's why the cycle needed more than four questions to actually work. It needed ten skills. Argyris and Schön had a plain way of putting what a skill does: it's what lets you turn what you believe into what you're actually able to do, in the room, in the moment. A pattern names the shape of a problem. A skill is what lets you meet it. A team where people are strong in different skills moves through the cycle far better than any one skilled practitioner working alone, no matter how good they are. Here's what years of watching teams run this cycle taught me: no one is good at all ten. Not the most experienced practitioner I know. Not me. Each skill draws on a different kind of thinking—noticing what isn't being said, pushing a vague idea into something specific enough to act on, holding a room's attention long enough for a real commitment to land. These are cognitive skills, not personality traits, and no single mind comes stocked with all of them. What that means in practice: a team where people are strong in different skills moves through the cycle far better than any one skilled practitioner working alone, no matter how good they are. The pattern doesn't just describe what to do. It works because the doing gets distributed across more than one kind of mind. That cuts against a tempting misreading of this whole series. A well-tested pattern language can make something simple to describe. It cannot make it easy to do. The difficulty doesn't disappear when the language gets clean. It moves from confusion about what to do, to the harder question of who in the room is actually equipped to do which part of it. Here's a distinction worth being clear about, because it's the difference between Part 2 and this piece. We are checking Hunter's pattern language one way: against the memory of people who stood next to him. That's real, and it's the only test available now that he's gone. But it's a slower, softer test than the one the cycle went through. Memory can be generous. Memory can round off the parts that didn't quite work. The Strategic Doing Cycle was checked a harder way while I was still there to watch it happen: against outcomes, in real time, in front of strangers who had no reason to be kind about it. If a version of the cycle confused a room, the room told me immediately, without meaning to. That's a different currency of testing—not 'does this match what someone remembers,' but 'does this work, right now, on people who've never heard of it.' Both are real tests. Neither is a substitute for the other. A pattern language built only from outcomes, with no one left who remembers *why* a version failed, would lose something too. What makes a knowledge commons—Hunter's or Strategic Doing's—worth building is that it can hold both kinds of evidence at once: the felt memory of people who were there and the harder record of what actually worked when tried on people who weren't. This is what I'd offer as the answer to Part 2's open question.A pattern earns rigor the same way any warranted belief does, in Dewey's sense: not by being certain, but by surviving repeated, honest contact with situations that could have broken it.The cycle survived fifteen years of that. It got worse before it got better—Version 1's technical language was, if anything, a step backward from plain speech—and it got corrected because the failures were visible and nameable, not hidden. That's the whole argument of this series, in the end. The swamp doesn't need draining, and the knowledge in it doesn't need to be trapped in one person's hands. It needs a language—built slowly, tested against strangers, revised in public when it fails—and thirty years is long enough to find out whether that language holds. Hunter's is a few months into that test. Mine is thirty years in. Neither of us is finished. That's not a weakness in the method. It's the method working exactly as it should. This is the third piece in a series that began withKnowledge in the Swampand continued withWhen the Practitioner Is Gone. »Knowledge in the Swampexplores Schön's core claim: the problems that matter most rarely stay still long enough for rules to solve them. In practice, knowledge moves in a loop: people act intuitively, others study and name the pattern, others test and adapt it, and eventually it becomes fluent practice again—now easier to examine, improve, and repair because it has been made explicit. »When the PractitionerIs Gone takes the circuit from Part 1 and breaks its quiet assumption: that the person who acted is still around to check the story against. Hunter Morrison, my brother and Cleveland's planning director for two decades, died three months into the project of writing down forty years of practice. »A Pattern Becomes a Languageanswers Part 2's open question by showing what the far end of that test looks like. Distinguishes a single named pattern (Part 1's marker hand-off) from a full pattern language — a connected system of patterns covering a whole domain, in Alexander's original sense. A Pattern Languageby Christopher Alexander, Sara Ishikawa, and Murray Silverstein (1977). The book Part 1 promised. Where 'pattern language' stops being a phrase and becomes 253 of them, each one a named, tested solution to a recurring problem in building design—the direct model for what a language of patterns actually looks like at full scale. Strategic Doing: Ten Skills for Agile Leadership(2019), with Scott Hutcheson, Elizabeth Nilsen, Janyce Fadden, and Nancy Franklin. The book-length attempt to write the language down as a whole, not just one piece of it—the counterpart, in my own work, to what Hunter's colleagues are attempting now.