Psychology/Behavior/How behaviour is recorded
Language:EnglishEspañolPortuguês (BR)
Behavior
How behaviour is recorded
Before you record anything you must define it. After you have defined it, you must choose a method — and the wrong method produces numbers that look precise and mean nothing.
Operational definitions
An operational definition states the behaviour in terms so concrete that a new observer could use it on their first day without asking you a question. Four parts, and the third is the one everyone skips.
| Part | Weak version | Usable version |
|---|---|---|
| Label | Disruption | Speaks without being invited during instruction |
| Includes | “Being disruptive” | Any spoken word, sound or call-out directed at the room or another person while the instructor is speaking |
| Excludes | (not stated) | Answering a direct question; whispering to a neighbour with no head turn; coughing, sneezing, yawning |
| Onset / offset | (not stated) | Begins at the first audible syllable; ends after 5 continuous seconds of silence. A new instance is counted only after 10 seconds of silence |
Prefer plain description to clinical language
Write “leaves seat” rather than “elopement”, and “raises voice” rather than “verbal aggression”. Clinical vocabulary imports a judgement the observation has not earned, and it travels further than you intend once it is in a file.
The six methods
| Method | Use when the behaviour is… | You get | Do not use when |
|---|---|---|---|
| Frequency count | Discrete, brief, clearly separated instances. | Count per period; rate per minute or hour. | Instances run together or last a long time. |
| Duration record | Continuous, and the length matters more than the count. | Total time; mean length per instance; longest instance. | Onset and offset are hard to see. |
| Latency record | A response follows a specific cue. | Seconds or minutes from cue to first response. | There is no identifiable cue. |
| Partial interval | Frequent, fast, or hard to count exactly. | Percentage of intervals in which it occurred at least once. | You need a true count — this method overestimates. |
| Whole interval | You need sustained engagement, not a flicker of it. | Percentage of intervals it occurred throughout. | You need a true count — this method underestimates. |
| Momentary sampling | You are supervising a group or cannot watch continuously. | Percentage of sampled moments. | The behaviour is short and rare. |
Numbers that genuinely compare
- Rate = count ÷ observed minutes. Always report rate, never a bare count, when session lengths differ. Eleven instances in ninety minutes is not worse than nine in thirty.
- Percentage of intervals = intervals scored ÷ intervals observed × 100. State the interval length every time; ten-second and sixty-second intervals are not comparable.
- Mean duration per instance = total duration ÷ number of instances. Report it alongside total duration — a rising total with a falling mean means more but shorter episodes.
- Baseline = the mean of at least five sessions recorded before anything is changed. Without a baseline you cannot claim that anything improved.
The eight predictable biases
| Bias | How it shows up | Counter-measure |
|---|---|---|
| Expectancy | You see more of what you predicted before you started. | Write your prediction down first, then read it only afterwards. |
| Halo and horn | One strong impression colours every later entry. | Timestamped entries only; never summarise from memory at the end of the day. |
| Recency | The last five minutes outweigh the first fifty. | Check the distribution of entries across the window. |
| Central tendency | Every rating lands in the middle of the scale. | Use counts and durations instead of ratings wherever possible. |
| Leniency and severity | You are consistently soft or consistently harsh. | Compare your counts against a second observer on the same session. |
| Attribution | Cause is placed in the person, never in the situation. | Complete the setting fields before the behaviour fields. |
| Confirmation drift | The definition quietly widens to include more instances. | Re-read the written definition at the start of every session. |
| Fatigue | Reliability falls sharply after about twenty minutes of continuous coding. | Cap coding at twenty minutes; break for five; note breaks on the form. |
Checking yourself
Two observers, the same session, no discussion until both have finished. Agreements ÷ total intervals × 100.
| Agreement | What it means | What to do |
|---|---|---|
| 90–100% | The definition is working. | Continue. |
| 80–89% | Usable, with disputed cases. | Identify the disputed cases and add them to Excludes. |
| Below 80% | You are recording different things. | Stop collecting. Rewrite the definition together, then re-test on a fresh session. |
The complete toolkit
Sixteen printable forms including all six methods, the inter-observer agreement sheet and the findings statement.
View the resource
Educational scope. Everything published on PMA Science Psychology is provided for education, study and professional documentation. Nothing here is a diagnostic instrument, a psychometric test, or a substitute for assessment, treatment, supervision or emergency care by a qualified professional.
If you are at risk right now, contact your local emergency number or a crisis line immediately. Read our boundaries and safety statement.
