The A is where records give up
Behaviour and consequence get structure; the antecedent gets a free-text box. Prose cannot be counted, so the single most useful question — what tends to come first — is asked from memory.
Everyone records antecedent, behaviour and consequence. Almost nobody can answer “what tends to come first?” from the record, because the antecedent was written as a sentence and sentences do not add up. This note is about typing the context — and about the three body clocks that make the most common loader of all countable without anyone typing anything.
The behaviour is what everyone notices. The antecedent is what makes the behaviour predictable — and it is the limb most records quietly abandon.
Behaviour and consequence get structure; the antecedent gets a free-text box. Prose cannot be counted, so the single most useful question — what tends to come first — is asked from memory.
A recorded “nothing unusual” is a fact about that day, not a missing field. Absence you looked for and absence nobody checked are different pieces of evidence.
A band is a declared convention on the person’s own numbers, joined to an episode when it is read and never part of the key it is filed under. Conventions can change without rewriting history.
Which background state loads which behaviour is learned from this person’s own co-occurrence, not assigned by a category system built for everyone.
ctx_key, and the empty value is itself meaningful: a recorded
“nothing unusual” is information, not absence.The reason this matters is visible in the history of the field. A formal functional analysis (Iwata and colleagues, 1982/1994) works by varying the antecedent condition — attention, demand, alone, play — and watching what the behaviour does. A record whose antecedent limb is prose cannot do the equivalent work from ordinary life, because the conditions it would need to compare were never made comparable.
Three clocks run continuously from the person’s own recorded numbers — night_wakings, days_since_last_movement, refusal_day_count — and annotate every episode when it is read. They exist
because the most common background loader of behaviour in this population is a body
state that nobody thought to write down beside the behaviour.
A band is a declared reading convention on those numbers: three or more nights of broken sleep; three to four days without a bowel movement, then five or more; two or more days of refused intake. Three things follow from calling it a convention:
Two people can share a diagnosis, a function and a behaviour and still be loaded by completely different things. So WellDash does not assign loaders from a category system. Which background state loads which behavioural form — whether a shut bowel shows up as tapping while broken sleep shows up as hitting out — is established from that person’s own co-occurrence, and the same function under a different band and a different form is treated as a different piece of evidence rather than averaged into one.
The distal and proximate layers — the slow background versus the thing that happened a minute before — are recorded for honesty, because clinicians reason with them. They are deliberately not load-bearing: nothing downstream depends on a call the record cannot make reliably.
Every typed vocabulary faces the same moment: a carer describes something real for which there is no token. There are three possible responses, and two of them destroy the record. Squeeze it into the nearest wrong token and the counts become fiction. Drop it into prose and the observation leaves the countable world. WellDash takes the third: the gap is flagged, so the vocabulary can grow with evidence behind the growth.
A flagged gap is a feature of an honest ontology, not a defect. It is also the only mechanism by which a typed grammar can survive contact with real lives rather than gradually teaching staff to lie in dropdowns.
The typed antecedent pieces would add the last piece — any “episodes preceded by X” question answered as a count rather than a search through notes. That is the designed step, and it is named as direction above rather than claimed here.
We deliberately do not force that call. The layer is recorded for honesty, and the person’s own co-occurrence counts differentiate which background state loads which behaviour — the record decides, not a taxonomy.
The body clocks make the background state itself countable: nights of broken sleep, days since the last bowel movement and refused-intake days annotate every episode when it is read. Turning the whole antecedent limb into typed pieces, so that “episodes preceded by X” is a deterministic join for any X, is the designed next step; today a prose antecedent is not countable, and the gap is flagged rather than forced.
A band is a declared reading convention on the person’s own recorded numbers — three or more nights of broken sleep, three to four days without a bowel movement — not a clinical threshold and not a diagnosis. Bands are annotation joined at read; they are never part of the key an episode is filed under, so changing how a band reads never rewrites history.
It is flagged, and the flag is the growth signal. Nothing is squeezed into the nearest wrong token and nothing is abandoned to prose. A vocabulary that cannot admit a gap silently becomes fiction.
Because “nothing unusual” is information. A recorded absence of setting events is a fact about the day and is treated as one — it is not the same as nobody having looked.
The demonstration shows capture, the typed grammar and the ranked responses on synthetic records.