Finding a Basin
What to look for in the world around you, and what to test before you call it one
Many things return persistently, and they do not all return for the same reason. There is an old saying that a whole can be greater than the sum of its parts. This framework points to that idea for study, because the author, like many others in complexity science, thinks it happens naturally, may sit at the heart of how living order emerges, and can perhaps be modeled mathematically as well as in the less tidy behavior of a neural network. Telling those returns apart starts with a definition: a basin is the set of starting conditions that end up in the same place. That definition is what makes the idea testable, because it tells you what to vary and what to watch. Everything on this page follows from it.
Start by naming three things
Before any observation is worth recording, write down three things. What comes back, described precisely enough that two people would agree it has returned. What it comes back to, described the same way. And what else it could have settled into instead. If the third one is blank, there is nothing to test yet, because a basin implies somewhere else to land.
Six things to look for
Each one gives a case, the loop that does the returning, and the thing to measure. The loop matters because a basin needs something keeping it: a few parts that each make the next more likely, closing back on the first, fed by supplies from outside. That shape is what chemists call an autocatalytic set. Naming the part whose loss would end the return is the testable half of the guess.
1. The same ending from different starts
A community garden with twenty plots and a single water spigot at one corner.
Who gets which plot. The garden redraws the assignment every spring, so each year starts differently.
By July, the same eight plots, all of them within a hose length of the spigot, are the ones still being worked. The other twelve have gone to weeds. That happens whoever was assigned them.
Water is the work. A plot by the spigot gets watered on the way past; a plot at the far fence needs someone to carry buckets in July heat. The far plots are not forbidden, they are just expensive, and by midsummer the cost has sorted them out.
easy water → a crop worth picking → that gardener signs up again and tells a neighbor → those plots stay claimed and watered. Each step makes the next more likely, which is what holds the eight in place year after year. It runs on water, seed, sunlight and the city's lease, none of which the garden produces.
Start it somewhere else on purpose. Give the eight far plots to the most committed gardeners and the near plots to beginners. If July still shows the same eight worked, the ending does not depend on the start, which is the thing being claimed.
A second spigot at the far fence. Move the cost and the eight plots move with it, which is the strongest evidence that the water, rather than the gardeners or the soil, is what put them there.
The worked plots simply follow whoever was keenest that year. Then there is no settling, only a record of who showed up.
2. Return after a push, with a time you can quote
A two-lane county road with a feed store, a diner and a grain elevator along it. The county counts vehicles with a tube across the lane, so there is a weekday number going back years.
A snowstorm closes the road for two days. The count goes to almost nothing, which is not interesting by itself. What happens next is.
The weekday count, to the level it held before the storm. Not the same cars and not the same errands, the same level.
About four days after the road reopens. The next storm closes it again and it takes about four days again. The repeated number is the evidence, more than the return itself.
Most of the trips were not cancelled, they were postponed. Feed still has to be picked up and the elevator still has to be reached, so the first days after reopening run slightly heavy while the backlog clears, and then the ordinary week resumes. The number is set by how long people can put a trip off, which is why a road serving errands recovers in days and one serving a summer festival recovers in a year.
trips → trade at the feed store, the diner and the elevator → those places stay open → reasons to drive. Each step feeds the next, which is why the level is a level rather than a drift. It runs on fuel, vehicles and the county's plowing and paving, none of which the road's traffic produces.
Compare recovery times across disturbances of different kinds and sizes: a storm, a week of bridge work, a fuel price spike. Similar pushes should give similar times, and bigger ones should take longer in some orderly way. A time that is four days once and eight months next time for the same size of push means something else is setting it.
The destinations, not the road. Close the elevator and the traffic does not come back no matter how good the pavement is. That is the member to name, and it is the one worth watching.
The count settles at a visibly different level after each disturbance, or the recovery time is unrelated to the size of the push. Also be careful that you are measuring the return of the level and not the ordinary weekly rhythm, which is a schedule and would be there with or without a basin.
3. A push that does not come back
A diner on a small main street. Sixty lunches on a weekday, the same cook for nine years, and a crowd that mostly eats there three times a week.
A kitchen fire closes it for two weeks. It reopens and is back to sixty lunches inside a month.
Four years later a second fire closes it for seven months. It reopens to thirty and is still at thirty two years on.
The edge of the basin lies between a two-week closure and a seven-month one. Same diner, same street, same menu, same owner. Only the length of the gap changed, so the length is what carried it over the edge. A single event tells you nothing; the pair brackets the boundary, and that bracket is the finding.
regulars → daily takings → the cook and the servers stay → the same food at the same hours → regulars. It runs on deliveries, power, the building and wages the customers earn somewhere else.
Each part of that loop has its own patience. The cook can wait a few weeks for the kitchen to be rebuilt, not half a year, so at some point he takes the job at the truck stop. The regulars need lunch every day, and after a couple of months the sandwich place has their habit. The supplier gives the route to somebody else. Two weeks is inside everyone's patience. Seven months is outside all of it, and the diner reopens with the building intact and the loop gone.
Narrow the bracket. Find closures of one month and three months, here or at comparable diners, and record which returned. The shortest failure and the longest success close in on the edge from both sides.
The cook and the lunch habit, in that order. If the owner brings the same cook back and the crowd still does not return, the habit was the binding part, and that is worth knowing because habit can be rebuilt with a month of free coffee while a cook cannot be conjured at all.
It returns to sixty whatever the closure length, in which case you have not found an edge in that range. Also check what else happened during those seven months. A bypass opening or a new competitor would explain the thirty just as well, and a claim about closure length cannot survive that confusion.
4. A named alternative, with a real case
Two towns twelve miles apart, each with a Friday market on the square, forty-odd stalls, drawing from the same farms and the same shoppers.
One flood, one spring, both squares under water. Neither market runs that season.
Starts again the following spring with a dozen stalls and is back to forty within two years.
Never starts again. Friday is an ordinary day and the square is parking. Nothing is stopping a market; there simply is not one.
One town returning proves nothing about what holds it, and one town failing could be bad luck. Two towns hit by the same shock in the same week, with the same farms and shoppers, is close to an experiment. Almost everything is matched, so whatever differs between them is a short list of candidates for what the return depends on. Town B is the named alternative, and it is twelve miles away rather than hypothetical.
sellers set up → shoppers come because there is something to buy → sellers come because there are shoppers. It runs on produce, the square and a permit.
A market is a meeting, and neither side can go first. A seller who stands alone on the square loses a day's trade, and a shopper who walks over to three stalls stops walking over. What holds the two together is the shared knowledge that Friday is the day, which survives a bad week and does not survive a missing season.
The short list of differences between the two towns. Did anyone keep announcing a date during the closure? Did the biggest seller, the one people came for, survive the year? Was the square given over to parking in the meantime? Did a Saturday market in the next county take the stalls? Each of those is a candidate for the part whose loss ended it.
The two outcomes track something plain and continuous, such as one town losing a third of its population or getting a bypass. Then the market size is following demand, there is no second state, and the flood is a coincidence rather than a push.
5. A gap between going out and coming back
Two ways across the river. The old bridge carries four thousand vehicles a day and has a coffee stop and a tire shop on the approach. A free crossing two miles upstream adds about six minutes.
A two dollar toll goes on the bridge. This one is a dial rather than a shock, which is what makes the case useful: the same dial can be turned back.
Within three months the bridge is down to two thousand vehicles a day. Drivers have found the six minutes cheaper than the two dollars.
Two years later the toll is removed. Traffic does not return to four thousand. It settles near two thousand four hundred and stays there.
It tipped out at a toll of two dollars and did not tip back at a toll of zero. That gap is the finding. A price that has to go below the price that pushed you out, and in this case cannot go low enough, means the state itself moved while the traffic was away.
The coffee stop closed in the first winter and the tire shop went to the highway. The route that comes back at zero toll is not the route that left at two dollars, because two of the reasons for taking it are gone. Removing the toll restores the price and not the reason.
cars pass → the coffee stop and the tire shop survive → the route is worth taking → cars pass. It runs on fuel, the bridge and drivers with somewhere to be.
Look for anything that was turned up and later turned back down, and record the level in both directions: a toll, a parking charge, a one-way conversion since reversed, a price rise rolled back, a bus route cut and restored. Two lines that do not retrace each other are the evidence.
Replacing the missing member rather than the driver. A new coffee stop, or anything else that makes the route worth taking, would do more than another twenty cents off a toll that is already zero.
Traffic follows the toll up and down within a few weeks each way, with no lasting gap. Then you have a straightforward response to price, one state, and nothing that needed a boundary to explain it.
6. Recovery that is getting slower
A youth baseball league. Six teams, about ninety players, coached by parents whose own children play. Sign-up sheets going back twenty years, which is the only reason any of this is visible.
A bad year. A wet spring, or the soccer club moving its season, and sign-ups come in a fifth short. This happens every few years and is nothing unusual.
Full rosters again the following season. One year down, one year back.
The knock ten years ago took one season to make up. The one five years ago took two. The last one is in its third season and the rosters are still short. The knocks are the same size; the recovery is not.
There are still six teams and a season every spring. The level looks fine. The signal is entirely in how long the league takes to come back, which is why it needs the old sign-up sheets to see at all.
players → teams → parents who coach because their own child is on one → a season that runs well → younger siblings sign up → players. It runs on the field, the gear and family time.
The league makes its own coaches. A thin year produces fewer parents in the dugout, which caps how many teams can be fielded the next spring, which turns families away, which produces fewer coaches again. Each pass around the loop replaces a little less than it lost, so the hole takes longer to fill each time. A loop that replaces more than it loses snaps back; one that barely breaks even crawls back; one that replaces less than it loses does not come back at all, and this league is walking from the first toward the third.
Plot recovery time against the year, using only knocks of comparable size, and watch whether the line rises. Sign-up numbers wandering more from year to year is a second sign of the same thing. Both are clues rather than proof, and both are worth more than anybody's impression that it feels harder than it used to.
The coaching parents, because the league produces them itself. Below some number of adults willing to run a team the season cannot be assembled at all, and that floor arrives while the player numbers still look survivable.
The town simply has fewer children. Check school enrollment before anything else. If the pool of nine-year-olds has halved, slow recovery is arithmetic rather than a weakening loop, and the league is tracking its supply the way any well behaved thing would.
Three things that look like a basin and are not
This is the part that separates a claim from an observation, and it is where most everyday examples fail.
Two cautions about the loops
A loop that closes is not proof of anything on its own. Every one of the six above could be held up from outside instead: the garden by a city parks budget, the road by a haulage contract, the diner by a landlord charging nothing, the market by a grant, the route by a planning decision, the league by one determined volunteer who is not produced by the league at all. Naming the loop and naming who pays for it are two different jobs, and the second is on the sovereignty page.
The other caution is that these loops are stated in the language of people and habits, not chemistry. The chemical version has a precise definition, a closed set of reactions each catalysed by a member of the set, and nothing here meets it. What carries over is the test, not the theorem: name the catalyst, remove it, see whether the pattern still comes back.
The questions to ask
These are the ones an ordinary observer can answer about their own town, trade, household or workplace. Anyone whose answers to four, five and six come back blank has a habit or a schedule rather than a basin they can demonstrate.
- What exactly comes back, and how would you know it was back?
- Back to what? Describe it so that two people would agree.
- How long does it take? Give a number from a real instance.
- What is the biggest disruption it has survived, and what happened afterward?
- Has it ever failed to come back? What was different that time?
- What else could it have settled into instead? Name a real case.
- Who or what is doing the pulling? If it is a person, an agency or a subsidy, say so.
- Is the recovery getting slower over the years?
Use the disruptions that already happened
Most of the systems worth studying cannot be pushed on purpose. They do not have to be. Disruptions are everywhere and they come dated: a road closed for six months, a factory shift change, a bridge out, a blackout, a strike, an anchor store closing, a storm, the shutdowns of 2020. Each one is a disturbance with a known start, a known size and a record of what followed.
For each case, record one of three outcomes. It returned to the same level. It returned to a visibly different level. It never returned. A stack of those records across many comparable places is worth more than any single careful measurement, because the boundary shows up in the difference between the cases that came back and the ones that did not.
A worked example
The thing: the Saturday crowd at a small town's main street, measured as cars parked on the street between nine and noon.
The state it returns to: roughly eighty cars, give or take fifteen, every Saturday outside holidays.
The alternative: the next town over, where the same count settled near fifteen after its grocery closed in 2014 and never recovered.
Disturbances on record: a two-week water main repair that closed one block (count fell by half, back within three weeks); a snowstorm (back the following week); an eight-month bridge closure on the north approach (count fell to thirty and came back to sixty-five, not eighty, and has stayed there).
What that gives you: a recovery time of about three weeks for small disturbances, evidence of a boundary somewhere between a two-week closure and an eight-month one, and a partial return that suggests the state itself moved rather than the system simply taking longer. The next thing to collect is the count from other towns that lost a road for a similar stretch.
What it does not give you: any account of what holds the eighty-car state up. That is a separate question, and the answer might be a basin that maintains itself, or a county subsidy that could end next year.
What finding one does not settle
Everything above establishes that a pattern has a basin. It says nothing about what keeps the basin there. A pattern held in place by something outside it, and a pattern that maintains its own return, both pass every test on this page. That is not a flaw in the tests. It is the reason the framework has a separate question for it, and that question is sovereignty: who pays to keep the thing returning, and what happens to the return when they stop.
So the honest order of work is this. Establish the basin first, with the observations above. Then ask who supplies the pull. The second question is much harder and is answered by removing or interrupting the supplier, which is rarely possible on purpose and usually has to be read out of an accident.
Where this sits: the idea itself is on the Basins page. What a basin cannot tell you apart is on What Is an Attractor and What Is an Attractlet. The question this page hands off to is sovereignty.
A field guide rather than a piece of the framework. The tests here are ordinary practice in nonlinear dynamics and in ecology, written for someone with no training in either. They introduce no primitive, no condition and no kernel construct, and nothing here is a claim that any particular pattern is a basin.
← Back to Basins