What it counts

What a second walk would settle

Six questions, in order of how much they would change what this site is allowed to claim. None of them needs a price or an opinion.

Home / What it counts

Method

Most of what this record cannot answer needs a different instrument, not a larger one. Six questions, in order of weight.

1. Whether anybody else assumes what the observer assumed

The largest figure on this site — a hundred and five phantom limits — is one person's error rate. An operational review may include employee productivity data for planning when the purpose, access rules and limits are explained before the fieldwork begins.

What would settle it. Ask three people sitting in each place how long they think they can stay, before telling them anything. Compare with what the place states. A separate perspective on streets, seating and urban observation is available from Public Space.

That turns the central finding from an account of one walk into something about a population, and it is the only change that would alter what this site may claim. It is also the hardest: approaching strangers to ask about staying changes the thing being measured.

2. Whether forty minutes is hiding the conversation

Twenty-three per cent is a floor. The honest ceiling, counting the eighteen unobservable places, is thirty-two.

What would settle it. Three sits per place at different hours, including an evening and a weekend. Or the same walk with a two-hour sit and a third of the places.

The second is cheaper and loses the comparisons between kinds, which are the only part of the conversation field worth keeping. The first is six hundred hours.

3. Whether the stated limits are the real ones

A hundred and sixty-eight places said there was no limit. Nobody tested one.

What would settle it. Sit past the stated limit in the twenty-one places that have one, and sit for three hours in a sample of those that do not. Record what happens.

That is a different kind of study — participant rather than observer — and it would measure practice instead of statement. It is the thing this method is most obviously not.

4. Whether the three sociable kinds share an occasion

Pubs, places of worship and halls produce conversation at fifty-five, fifty-seven and thirty-six per cent. The other nine kinds together at twelve.

The reading offered is that all three have something everybody present is doing at the same time. That is a description applied afterwards to three rows that stand apart.

What would settle it. A fifth field, defined in advance: is there an occasion in this room that includes everybody in it? Then the reading becomes a count.

5. Whether the bookshops are two places

Seven of thirteen bookshops have a counter serving coffee. The record coded each building once, so seven entries hold two arrangements.

What would settle it. Code by room rather than by building. It affects the entry field and the seat field for the same seven places, and it is the only known coding error in the record that could be fixed by a rule rather than by more work.

6. Whether any of it holds in a second city

One city, one season.

What would settle it. The same sheet, somewhere else, by somebody else. This is last on the list because it is the least informative: a second convenience set of two hundred places tells you about two cities, not about cities.

What it would genuinely test is the flatness — whether the terms fail to predict conversation anywhere, or only here.

What none of these needs

A price. A named place. A rating, a comparison, or a view about where anybody should go.

All six are answerable with the same four fields, or with one more defined in advance, and that is deliberate: a method that could only get better by turning into a guide would be defective at its root.

The six in order of cost

	what it needs	what it changes
1	three people asked per place	what the site may claim
2	three sits per place, or a third of the places	the conversation figures
3	sitting past a stated limit	statement against practice
4	one field, defined in advance	turns a reading into a count
5	coding by room, not building	seven entries in two fields
6	a second city	whether the flatness is local

The first is the only one that alters what may be said. The fifth is the only one that costs nothing at all.

The one that is a different study

Question three. Sitting past a stated limit, and sitting for three hours where none is stated, to record what actually happens.

That is participant rather than observer, it would take a day per place rather than forty minutes, and it would measure practice instead of statement. It is the thing this method is most obviously not, and it is where the record's firmest claim is most exposed.

A hundred and sixty-eight places said there was no limit. Nobody found out whether that was true in any of them.

What would not help

More places. Four hundred rooms at forty minutes answers nothing that two hundred did not. The questions that remain need a different sheet.

More cities with the same sheet. Useful only for question six, and least informative of the six, because a second convenience set tells you about two cities rather than about cities.

More fields. The sheet has four and one was a guess. Adding six more would slow every visit and would not touch the two large gaps, which are about who is asked and how long anybody sits.

What a second walk would cost

Eleven weeks, a hundred and forty-four hours of sitting, and a sheet.

The page on walking it yourself gives the whole procedure, including the three things this record would do differently and the six places it threw out.