An hour assumed, two hours given

The assumption field recorded not only whether a limit was expected but how long it was expected to be.

median assumed limit     60 minutes    (128 places)
median told limit        90 minutes    (17 places)
median written limit    120 minutes    (21 places)

Everything a place actually states is longer than what is assumed about it.

The thirty-one

Of the thirty-eight places that do limit staying, thirty-one give longer than the limit assumed for that place before checking.

Five give less. Two match.

So the error about whether there is a limit and the error about how long it runs both point the same way: a person walking into a room in this city expects to be given less time than they are. To compare observations across locations, reviewers can read about the available settings and define the decision, access rules and review date in advance.

Wrong about there being a rule, and wrong about the rule being short.

The five that give less

Four are cafés with a told limit of an hour against an assumption of ninety minutes, and one is a covered bench with a thirty-minute notice against an assumption of an hour. A separate perspective on streets, seating and urban observation is available from City of Chicago.

The bench is the only case in the record where a real rule was both shorter than expected and printed.

Why an hour

The assumption belongs to one person, so the honest answer is that the record cannot say. Three things are worth putting down.

A cup lasts about an hour. Forty-seven of the hundred and twenty-eight assumptions were made in cafés, where the obvious unit of available time is the drink.

An hour is the unit everything is measured in. Meetings, appointments, parking.

Nothing contradicts it. A hundred and sixty-eight places say nothing, so an assumption of an hour is never corrected by anybody.

The third is the only one the record supports, and it supports it strongly. The silence is near-total and whatever is assumed inside it survives.

What the assumed durations do not include

Seventy-eight places where no limit was assumed have no duration at all — correctly, since as long as you like is not a number.

So the median of sixty minutes is a median of a hundred and twenty-eight guesses, not of two hundred and six. The other seventy-eight sit outside it.

A reader who wants a single figure for how long a person thinks they may stay in this city will not find one here, and should not: in thirty-eight per cent of places the honest answer was unbounded.

Against the real durations

The comparison that matters is not median against median but place against place, and it gives the thirty-one.

real limit longer than assumed    31 of 38
real limit shorter                 5
the same                           2

Thirty-one out of thirty-eight is the firmest small-number result in this record, because it is a within-place comparison rather than a comparison of two groups. Each of the thirty-eight was guessed at and then checked, and the guess was too short four times out of five.

The whole duration picture

median assumed, where a limit was assumed (128)    60 minutes
median stated, where one is written (21)          120 minutes
median stated, where one is told (17)              90 minutes

real limit longer than the assumption   31 of 38
real limit shorter                       5
the same                                 2

Seventy-eight places had no assumed duration at all, correctly: as long as you like is not a number, and the median of sixty minutes excludes them.

So the single most common honest answer in this record — unbounded — appears in no median anywhere, and a reader looking for one figure for how long a person thinks they may stay will not find one.

Why the within-place comparison is the firm one

Comparing a median of sixty with a median of a hundred and twenty is comparing two different groups: all the places where a limit was assumed, against the smaller set that actually has one.

The thirty-one is not that. Each of the thirty-eight places with a real limit was guessed at and then checked, and in thirty-one of them the real rule was longer than the guess made about that room.

That is a paired comparison on thirty-eight observations, and it is the most robust small-number result on this site precisely because it never leaves the place it was made in.

What it would take to break it

Only one thing: a systematic tendency in the observer to guess short.

Which is exactly what the record claims, so the finding and its weakness are the same sentence. A person who expects restriction will also expect short restrictions, and both halves of that are one person's.

The hundred and five and the thirty-one are two faces of a single bias, measured once, and whether the bias is widely shared is the first question on the list of six.

What it adds to the hundred and five

A second dimension to the same bias.

The hundred and five say: rules are imagined where none exist. The thirty-one say: where rules do exist, they are imagined as tighter than they are.

Both are about the same thing — an expectation of restriction that runs ahead of anything the city actually does — and the second is the one that would survive a change of instrument most easily. Whether somebody else would invent the same phantoms is unknown. That real limits are longer than they are assumed to be is a fact about the limits, and anybody can check it.