Paired Observer Drift

What this memo reports

This is a description of a shape in the logs. It reports what paired observers wrote, in what order, and how the resemblance between their accounts changed over four months. It carries no explanation, and the project asks readers to resist supplying one on its behalf.

All material here is testimony. Participants report; the project files. Where a reading of the pattern would require something beyond ordinary attention, memory and conversation, this memo declines to take it.

Blind first logging

Each pair sat at an agreed hour in separate places. Neither participant knew whether a given night was paired or unpaired: roughly one night in six was a decoy on which the partner did not sit, and one in eight was a null night with no sitting at all. Both were logged on the same form as everything else.

The first pass was written alone, within ten minutes of the sitting, before any message, call or forum post. Both logs were sealed and timestamped on submission. Only after both had arrived at the note desk were the pair permitted to speak, and their conversation was logged separately as a second pass.

Blinding held better than we expected and worse than we would like. Three pairs admitted to breaking it at least once, and those nights are marked as broken in the file rather than removed. Two participants correctly guessed decoy nights more often than chance would suggest, and both said afterwards that they had inferred it from their partner's schedule.

Initial convergence

Scored blind, first-pass logs from paired nights shared more elements than first-pass logs from decoy nights. The margin was modest and it was not uniform: it came almost entirely from four of the eleven pairs, and it was carried by broad categories rather than specifics.

What converged was mostly texture. Pale or unlit interiors. A sense of pressure preceding imagery. Cool floors, still air, a direction of travel without arrival. What rarely converged was detail: numbers, faces, words, objects that could be checked against each other. In the two months of the window when a specific-detail prompt was added, it produced almost nothing.

Narrative convergence

Once pairs were permitted to speak, resemblance rose sharply, and it kept rising for the rest of the window. By the second pass, accounts that had shared a texture in the first pass often shared a scene, a sequence and eventually a vocabulary. Several pairs arrived at a shared name for an image within a fortnight and used it thereafter as if it had always been theirs.

The project regards this as the least surprising finding in the memo, and the most important to record. Two people who compare notes nightly will build a joint account, and a joint account is a poor witness to anything except itself. Everything of interest in this memo therefore lives in the first pass; the second pass is retained mainly so that later contamination can be traced.

Drift across the window

The pattern the memo is named for appeared from about week five. First-pass resemblance within a pair, having risen, began to fall, while second-pass resemblance continued to climb. Pairs were describing their sittings more similarly to each other in conversation while logging them less similarly in private.

Drift was not steady. In six of the nine completing pairs it moved in steps: a fortnight of stable first-pass overlap, then a sharp drop, then a new plateau lower than the last. In two pairs the drop reversed once, briefly, without either participant reporting any change in conditions. One pair showed no drift and also showed the lowest convergence in the cohort throughout.

We can describe the shape. We cannot say what it is a shape of. Fatigue, habituation, the ordinary decay of novelty, a growing reluctance to write down what a partner has already claimed, and simple boredom are all live candidates, and the memo does not choose between them.

An illustrative record

The extract below is illustrative rather than representative. It is included so readers can see the form of the record, including the entries that went nowhere.

Rows of the last two kinds are the reason the extract is here. A file that shows only the convergent nights is a file that has been curated.

Ordinary explanations

Each of the following would account for the pattern without anything unusual occurring, and the project holds all of them open.

• Shared vocabulary. Pairs were briefed together, read the same threads, and had the same words available before they sat.
• Expectation. Every participant knew what the protocol hoped to find, and knew which nights felt worth writing at length.
• Selection in memory. The first pass is written from a recollection minutes old, and recollection favours the notable.
• Common conditions. Pairs often sat at the same hour, in similar rooms, in the same season, tired in the same way.
• Scoring judgement. Two blind readers still bring a sense of what counts as a shared element, and that sense was formed inside this project.
• Regression. Four pairs carried most of the early convergence, and small groups that begin high tend to move down.

Taken together these are sufficient to explain the memo's findings. The project records the pattern because it is legible and repeatable in the log, not because the ordinary explanations have been excluded.

Dissent and handling

Two project readers objected to this memo being filed at all, on the grounds that a described shape with no test attached invites the reading the memo declines to make. Their objection is recorded here at their request and has not been withdrawn.

A third reader holds that the drift finding is an artefact of the scoring scheme, which weights early categories more heavily than later ones. That criticism is being tested against a rescored subset, and this memo will be revised if it holds.

The memo remains observational. It is filed alongside BCP Note 02, which sets out the question it was written to serve, and readers should take the disconfirming conditions listed there as the standard this material is measured against.