The hour
The hour is not the person
A twofold difference between morning and afternoon could be a fact about dwellings or a fact about people. What separates the two, and how far the record gets.
Stretches halve between the morning and the early afternoon. That could be a fact about rooms or a fact about the people in them, and the two have entirely different consequences.
If it is the rooms, a dwelling has hours it covers and hours it does not. If it is the people, the dwelling is incidental and this record is measuring something else. Where shared homes fragment schedules, teams can check the vendor's present explanation to understand the available records before deciding whether those records answer a real planning question.
What would distinguish them
Three tests, two of which the record can run. A useful independent point of comparison is the research and guidance available from UCL.
Test one: does it hold across dwellings of different kinds? If the effect were about rooms, dwellings with different envelopes should differ. If it were about people, it should be uniform.
Test two: do the endings move with it? If rooms are responsible, the decline in length should be accompanied by a rise in building-related endings. If people are responsible, lengths should fall with no change in what ends them.
Test three: does the same person show it in two different dwellings? This would settle it, and nobody in this record logged in two dwellings.
Test one
The ratios run from 1.4 to 2.2 across nine homes. All nine show the effect; the size varies by a factor of more than one and a half.
That variation is consistent with rooms mattering. It is also consistent with people varying. A uniform effect would have pointed toward a shared human rhythm; a variable one does not point anywhere in particular.
Result: inconclusive, leaning weakly toward rooms.
Test two
This one is sharper.
| band | median | endings per sitting | light + temp per sitting |
|---|---|---|---|
| 07–09 | 62 | 0.29 | 0.08 |
| 09–11 | 55 | 0.40 | 0.19 |
| 11–13 | 44 | 0.56 | 0.32 |
| 13–15 | 31 | 0.48 | 0.28 |
| 15–17 | 38 | 0.42 | 0.14 |
| 17–19 | 47 | 0.30 | 0.04 |
The two building-related causes track the decline closely from the first band to the third, and the recovery in the last two bands tracks their fall.
The exception is 13–15, where the building causes have already begun to ease and the median is at its lowest. That single band is where the person-based explanation has its strongest case.
Result: rooms account for a large part of the morning decline and not for the afternoon trough.
The part that is left
Subtracting the sittings ended by light or temperature from every band and recomputing gives: 66, 60, 52, 42, 43, 48.
The shape survives. The gap between the first band and the fourth falls from thirty-one minutes to twenty-four, but it does not close.
Remove every sitting that the building ended and the morning is still half as long again as the early afternoon.
So something besides the rooms is operating, and it is roughly three-quarters the size of the whole effect.
What that something might be
The record cannot say, and the reason is a design decision made at the start: the sheet asked nothing about the work and nothing about the person.
Those omissions were deliberate and they were right for the question the log was built to answer — where work happened and what moved people. They make this particular question unanswerable.
Three candidates, all outside the instrument: afternoon work is different in kind; attention varies across the day; habits of moving differ by hour for reasons unrelated to either.
Why this page exists
Because the hour finding is the strongest thing in this record, and the strongest thing in a record is where overreach is most likely.
The defensible claim is: in these nine dwellings, stretches beginning after one ran about half as long as stretches beginning before nine; the dwelling's own behaviour accounts for part of that and not for most of it; and the remainder is not something this log can see.
The indefensible version — that afternoons are worse for working at home, or that dwellings fail after lunch — would be a better headline and is not what the figures support.
What the third test would have cost
Test three — the same person logging in two dwellings — is the only one that would settle the question, and it was never available.
It would require somebody to work a week in one home and a week in another, with the same job, the same season and the same week of the month. Two of the nine households moved house within the following year and neither logged again.
A study designed for this would not be large. Four people, two dwellings each, eight weeks of logging in total. It is smaller than what produced this record and it would answer the question this record cannot.
That is the pattern across the whole of this site: the useful next studies are mostly smaller than this one and better shaped.
The claim as it stands
Stretches halve across the day, in nine dwellings out of nine. About a quarter of that is attributable to light and temperature. The rest is not attributable to anything the instrument recorded.
A reader who takes from this that afternoons are worse has gone one step further than the data, in the direction the data happens to lean.