The thirty-five minutes

Three different durations

For any one collision there are three numbers, and the log holds only the first. Another useful point of comparison comes from Digital Trends.

Time out of the room. The thirty-five minutes. Measurable, unambiguous, and the only one recorded.

Time until the activity resumed. Often longer. An activity pushed out of a room does not always restart elsewhere; sometimes it stops. Of the thirteen collisions where cooking or laundry was displaced, five did not resume in another room at all — they resumed in the same room later, which makes the interruption longer than the displacement.

Time until it was going properly again. Not recorded anywhere, because nobody can observe it. Work taken to a different room with a laptop under one arm does not resume at full speed.

Only the first is in this record, and that is a deliberate choice: the other two require judgements that would make the number unfalsifiable.

What the distribution looks like

A fat middle. Most collisions are between twenty and fifty minutes, which is roughly the length of a meal, a call, or a cooking session — and that is not a coincidence. A collision lasts as long as the activity that caused it.

The short ones are deliveries and brief interruptions. The long ones are almost all dinner.

Forty-seven times thirty-five

The arithmetic is tempting and should be resisted.

Forty-seven collisions at a median of thirty-five minutes is about twenty-seven hours over a fortnight. Stated that way it sounds catastrophic — nearly two hours a day of disruption.

It is not that, for two reasons.

Displacement is not loss. Somebody moved from the kitchen to the living room with a laptop did not stop working for thirty-five minutes. They worked in the living room. What was lost is the difference between working at a proper table and working on a folding one, which is real and is nothing like thirty-five minutes.

Some of the displaced time was welcome. Nine of the forty-seven were company arriving, and the household describes those as pleasant. They are in the log because the test is whether an activity gave way, not whether anybody minded.

What the number is good for

Not totalling. Comparing.

Thirty-five minutes is the right figure for asking whether a particular change helps. The exercise move removed four collisions of roughly twenty minutes each. The folding table removed three of roughly forty. The dinner-time experiment removed none and created two.

Against a median of thirty-five, those are measurable differences and they can be ranked, which is all that a median of this kind is for.

The one long tail

Three collisions ran to seventy minutes or more, all three in the living room, all three the same shape: the interruptible job settled in the living room for the afternoon, and at around four the other person finished their calls and came in to sit down.

Nobody was displacing anybody on purpose. The living room is where you go when you stop working, and it was occupied by somebody who had been pushed there earlier by a call.

Those three are the clearest instance in the record of a collision being a consequence of an earlier collision — and the log has no way to represent that, which is one of its limitations.

What happens in the thirty-five minutes

The log records where the displaced activity went, and the distribution is informative.

Of the forty-seven, thirty-one relocated to another room and continued, eleven stopped and resumed later in the original room, and five stopped and did not resume that day.

The eleven are the expensive ones. An activity that waits out the collision rather than moving is one where relocating is not possible — cooking, mostly, and anything involving the washing machine — and those are the entries where the thirty-five minutes really is thirty-five minutes of the thing not happening.

The five that did not resume are all domestic: a cleaning task, two pieces of admin, and twice somebody simply decided it would keep.

The number behind the number

A fair question is where thirty-five comes from, given that nobody was timing anything with a stopwatch.

Durations were recorded to the nearest five minutes, from memory, usually within the hour. The household's own estimate of their accuracy is that short ones are probably over-recorded — a ten-minute interruption feels like fifteen — and long ones under-recorded, because an hour spent in another room stops feeling like a displacement partway through.

If both biases are real they partially cancel, which is luck rather than method.

What can be said is that the median is robust to the error: moving every entry by five minutes in either direction changes it by five minutes and changes nothing about which changes helped, which is all the figure is used for on this site.