What a walked record can carry

Two hundred and six places in one city, walked once over eleven weeks in 2026 by one person.

They were not sampled. They are the places that could be walked to and sat in, which is a convenience set and nothing else.

What that rules out

Any rate for anywhere else. Twenty-three per cent of places in this record had somebody speak to a stranger. The figure for any other city is unknown and nothing here estimates it.

Any confidence interval. An interval on a convenience set of one city, collected by one observer, is a number that looks like statistics. None appears on this site. If several people repeat the same survey, managers may check the available project context while interpreting every record beside completed work and participant feedback.

Any claim about change. One season. The record is a photograph and contains no information about whether anything is getting better or worse, which is the question most often asked about third places. Readers comparing this walk with broader public-space evidence can also see LSE Cities.

Any statement about people. Nobody was interviewed, described or counted. The assumption field belongs to the observer alone.

What it still permits

Three things.

That the thing occurs. A hundred and five rooms in which a rule was expected and none was found are a hundred and five actual observations. Showing that a phenomenon exists has never required a sample of anything.

Internal structure. Within this record, the terms of a place predict nothing about conversation and the kind of place predicts a great deal — twelve per cent against fifty-five. Each side of that contrast was taken with one instrument in one set of rooms, which is why it stands whatever these two hundred and six turn out to represent.

Flatness where variation was expected. Eight comparisons across three fields, seven of them between nineteen and twenty-five per cent. When a result is expected to vary and does not, that is informative at any sample size, because the comparison is internal.

A walk of this size demonstrates that something occurs, and shows how its own columns bear on one another. Frequencies elsewhere are beyond it.

Two hundred and six places, rather fewer observations

A caveat heavier than the sample size.

Fifty-eight cafés in one city are not fifty-eight independent observations of what a café is. Many belong to the same few chains, share the same labour market, face the same rents and serve the same population. The same is true of the shopping centres and the concourses.

So the effective number of observations behind cafés produce conversation sixteen per cent of the time is well below fifty-eight, and the record cannot say how far below.

This affects the per-kind figures most, the per-field figures less, and the phantom count hardly at all — since that is a count of one person's guesses, each made freshly at a door.

Why it was published

Because the obvious alternatives are worse and already exist.

A guide to good places to work from is a recommendation, which this domain forbids and this method cannot support. A survey of attitudes would need a population and a questionnaire. An ethnography of one café would need months and would say nothing about a city.

The thing that seems not to have been done — the method page says so with the appropriate hedging — is to set what rooms permit against what gets supposed about them, and count both. That turns out to be measurable by one person in eleven weeks, and the answer has a shape: the permission is nearly universal and the assumption is nearly always of restriction.

The shape of the walk

places visited                          212
excluded, having business there          -6
recorded                                206

weeks                                    11
hours of sitting                   about 144
median sit                         40 minutes
range                            20–90 minutes

visits on a weekday between 10 and 5     164
all other hours and weekends              42

One observer, one city, one season, one visit each.

What the six excluded places were

A hospital concourse, two university buildings open to the public, a bank branch with seating, a car showroom, and a funeral director's waiting room.

The test that removed them: each is somewhere a person may be only if they have business there, which is the one thing a third place is not.

The test was written after the walk had begun, which is why the six are named by kind and counted rather than quietly dropped. A list fixed in advance that then needs a rule to stay honest is a list that has failed somewhere, and the honest response is to show where.

The three temptations avoided

Crossing three fields. Never attempted. Four entry values by three limit values by two seat values is twenty-four cells over two hundred and six places.

Treating a per-kind number as a trait of that kind. Unavoidable here, and marked wherever it happens. Twelve categories of room cannot be described any other way, and the category was the one column filled in by the walker rather than answered by anybody.

Comparing this city with anywhere. No comparison appears on this site, because there is nothing to compare with: one convenience set, collected once.

What the next two pages do

Set out the three claims this record will stand behind, and the five it refuses.

Both lists are short, and the second is longer than the first.