Three things the walk supports

Three claims, each put as narrowly as the counts allow, each followed by its footing and by the thing that would knock it over.

1. This city sets almost no limits on sitting

The figures. A hundred and sixty-eight places out of two hundred and six set no limit on staying: eighty-two per cent. Twenty-one have a notice and seventeen will give a figure if asked.

What it rests on. A look for a notice in every place, then one question in the hundred and eighty-five where nothing was found. Both steps applied identically everywhere. A team converting this walk into repeatable fieldwork can see how the product supports a pilot before choosing which time and project records are genuinely necessary.

What would break it. The seventeen told limits may partly be answers improvised under a stranger's question rather than policies — and that would make the figure stronger, not weaker. The real threat is the opposite: places that would move somebody along without ever having stated a limit appear here as unlimited. Nothing was tested, nobody sat past a notice, and the whole walk was conducted in the quietest hours. For wider research on public space, access and how places are used, consult AARP.

This is the firmest claim on the site and the one most exposed to the difference between what is said and what is done.

2. Where a limit exists, it is longer than it is assumed to be

The figures. Median written limit two hours; median told limit ninety minutes; median assumed limit one hour. Thirty-one of the thirty-eight real limits are longer than the limit assumed for that specific place before checking.

What it rests on. A within-place comparison — each of the thirty-eight guessed at, then checked — rather than a comparison between groups. That makes it the most robust small-number result in the record.

What would break it. Nothing in the record. The assumed durations belong to one observer, so the claim is really one person consistently expected less time than places give, and whether that generalises is untested.

3. Nothing about a place's terms predicts whether anybody speaks in it

The figures. Eight comparisons across three fields — what a place asks, what it allows, whether it has a free chair — and seven of the eight fall between nineteen and twenty-five per cent. Against that, the kind of room ranges from zero to fifty-seven.

What it rests on. The conversation field, which is the weakest on the site: a floor rather than a count, biased toward weekday afternoons, and unresolved in eighteen places.

What would break it. A longer sit would raise every figure, and if it raised them unevenly across the terms the flatness could dissolve. The record's own bias argument says the unevenness would most likely fall across kinds rather than terms, which would sharpen the result — but that is a prediction, not a measurement.

The common property

Not one of the three says anything about a place being good or bad.

The first is about what is stated, the second about the distance between a statement and a guess, the third about what the stated terms fail to predict. All three would be recognisable to somebody walking a different city with the same sheet, and none requires knowing anything about any particular building.

Three claims, one city, two hundred and six rooms. Nothing here says where to go, and nothing here could.

The three claims in figures

The second is the strongest per observation and the weakest in reach: thirty-eight places, each guessed at and then checked, with the guess too short thirty-one times.

The third is the broadest and rests on the weakest field.

The first is the firmest and the most exposed to the difference between what a place says and what it would do.

What each one would look like if it were wrong

If the limit claim were wrong, a longer or busier walk would find people being moved along in places that state nothing. The record cannot rule that out and names it as the largest unmeasured thing in the method.

If the duration claim were wrong, it would be because the observer systematically guessed short — which is exactly what the record claims, so the finding and its weakness are one sentence.

If the flatness claim were wrong, a longer sit would raise conversation rates unevenly across the terms rather than across the kinds. The record predicts the opposite and has not tested it.

What all three avoid

None mentions a named place, a price, a rating or a comparison with anywhere else.

That is not modesty. It is what remains after the method's own rules are applied, and the next page sets out the five things readers will want that cannot be built from it.

The claim that is not on the list

That a hundred and five limits were imagined.

It is the largest figure on this site, it is on the front page, and it is not in this list — because it rests entirely on one person's guesses recorded before checking. It is a true account of a walk and an untested claim about anybody else, and the page on whose assumption it is says so at length.

The next page sets out four more things the record refuses, and that one is the fifth.