The discovery that behaviour has a function — that the question is not what is wrong with this behaviour but what this behaviour is for — took the field half a century. It is one of the genuinely settled findings in the human sciences, and it is still routinely lost in the way support is written down. This is the history, what each stage left behind, and the one chapter still open.
These are not milestones in a museum. Each one is a working part of how good support is practised today — and each left a rule that a care record either keeps or quietly discards.
Watson’s manifesto, then Skinner’s operant analysis: behaviour is shaped by its consequences. The seed of everything after — if consequences select behaviour, then behaviour serves them. Inherited here: the response is recorded with what followed it, never on its own.
Baer, Wolf and Risley’s founding statement gives the discipline its name and its standards — applied, behavioural, analytic, technological, effective, generalisable. Inherited here: the demand that a claim be examinable by someone else from the data.
Carr reframes self-injury as operant rather than pathology. The question stops being what is wrong with this behaviour and becomes what this behaviour is for. Inherited here: no behaviour is recorded as a symptom of a person.
Iwata and colleagues build the functional analysis: controlled conditions — attention, demand, alone, play — that let a behaviour’s function reveal itself empirically. This is the moment function stops being an opinion. Inherited here: function is an evidential claim, not a professional impression.
Carr and Durand’s functional communication training: teach an easier ask that earns the same outcome, and the behaviour loses its job. Inherited here: the equivalence law — same wage, not the same shape.
Across two decades of case series and reviews, researchers measure identical-looking behaviours serving different functions in the same person. Treating behaviour by its appearance is established as an error rather than a simplification. Inherited here: the per-function unit — never aggregate by topography.
Multi-element models, proactive-first practice, restraint reduction and the UK competence framework: quality of life becomes the outcome rather than behaviour reduction. Inherited here: restrictive configurations are structurally barred from being offered, not merely discouraged.
Formal analysis is powerful and expensive: specialist time, controlled conditions, and a report that ages as the life moves. The unfinished chapter is making function-finding continuous, from the person’s own ordinary record.
The most consequential finding of the 1990s and 2000s is also the least intuitive: two behaviours that look identical can be doing entirely different jobs for the same person, and two that look nothing alike can be doing the same one. Across case series and reviews from 1992 to 2013, researchers measured exactly this.
The practical implication is severe. If you count “incidents of hitting” for a month, you have added together events that serve different purposes, and the resulting number cannot tell you what to do. It is not a rough guide; it is a category error with a chart attached. Every care record that reports behaviour by its appearance — and nearly all of them do — reproduces the error the field corrected two decades ago.
This is why function is the unit of evidence in WellDash rather than a field on a form. Ordering by what the record shows works, per function, is what makes the counts mean anything at all.
A formal functional analysis is one of the strongest tools in applied behavioural science. It is also scarce. It needs specialist time, controlled conditions, and someone able to arrange them safely around a person’s life. In UK services this typically means a waiting list, and then a report.
The report is where the second problem starts. It describes a person at a moment — a placement, a staff team, a medication regime, a season. Then the life moves. New staff arrive, a housemate leaves, a tooth aches, a medication changes. Within a year the document that everyone is required to follow is describing conditions that no longer exist, and the people following it know that but have nothing better to hand.
Neither problem is a criticism of the method. They are properties of anything done as an event rather than as a standing process.
The alternative is not a better report. It is to make function-finding a property of the ordinary record: every moment carries what was happening, what was done, and what followed, with the candidate functions held on the moment itself.
Two questions here are open in the literature rather than settled by it. They are named as open deliberately — go to the sources, not to us:
The frame is closed and the trails are open: what the system does is stated precisely; what the science has not settled is named as unsettled.
What the behaviour earns for the person — escape, attention, sensory regulation, communication, or a bodily need met. The founding insight (Carr 1977; Iwata and colleagues 1982/1994) is that behaviour works for the person, so support must serve what the behaviour serves.
Because identical-looking behaviours can serve different functions in the same person — the correction the research literature established across 1992 to 2013. Function is the unit of support; grouping behaviour by appearance is a measured, decades-old error.
Historically, when it is scheduled — and the report ages as the person’s life moves. The answer this system embodies is continuously, from the person’s ordinary record, with hypotheses held in parallel until their own evidence resolves them.
A procedure in which conditions are varied deliberately — attention, demand, alone, play — so that a behaviour’s function can reveal itself empirically rather than being inferred from opinion. Iwata and colleagues built it in 1982, republished in 1994, and it remains the field’s reference standard.
No, and it should not claim to. A controlled analysis manipulates conditions; a record observes life as it is lived. What a standing record can do is hold hypotheses in parallel across hundreds of ordinary moments, keep the outcome beside each one, and never go stale — which is a different contribution, not a substitute.
The demonstration shows capture, the typed grammar and the ranked responses on synthetic records.