Skip to content
Back to search results

Behavioral Observation Coding

Introduced by
Behavioral sampling tradition (Altmann), adapted to anthropology (Bernard)
Year
1974
Type
Systematic procedure for sampling and coding observed behavior
Family
Behavioral observation sampling
Font size

What does it do?

Behavioral observation coding is the systematic practice of recording who does what, when, by applying an explicit sampling rule and a predefined coding scheme to observed behaviour.

Rather than jotting impressions, the observer commits in advance to a rule — focal-individual, scan, ad libitum, or continuous sampling — and to a list of mutually defined behaviour categories, so that records are reproducible and comparable. Because two trained observers should code the same scene the same way, the method also requires measuring inter-observer reliability before the data are trusted.

The intuition behind the method

Observation becomes science when the rule for what to watch and how to label it is fixed before you look, not improvised afterward. Following one individual for a fixed window (focal sampling), or sweeping the whole group at set instants (scan sampling), or noting only conspicuous events (ad libitum), each yields a different and predictable slice of behaviour — and each must be chosen deliberately to match the question. A coding scheme then turns the stream of action into discrete, countable categories, and a reliability check confirms that the categories live in the behaviour, not just in one observer's head.

History

Systematic behavioral sampling was codified by Jeanne Altmann in her landmark 1974 Behaviour paper 'Observational study of behavior: sampling methods,' which named and distinguished ad libitum, focal-animal, scan, and continuous sampling and showed how each biases the resulting data. Though written for animal behaviour, the framework was rapidly adopted across the human behavioural sciences, and H. Russell Bernard's Research Methods in Anthropology carries it into anthropological practice, pairing the sampling rules with code-book construction and inter-observer reliability assessment. The approach now underpins structured observation throughout field research.

When to use it

Use systematic behavioral observation coding whenever you need reproducible, quantitative records of behaviour rather than narrative impressions — to count, time, or compare actions across people, groups, or conditions. It is the backbone of time-allocation, interaction, and activity studies. It assumes the behaviours of interest are observable and can be defined into reliable categories, and that you can train observers to agreement. It is less suited to meanings and motivations behind behaviour (which need interviewing), to rare events that seldom appear under a sampling rule, or to settings where being watched would distort the very behaviour you want to record.

Limitations and considerations

  • Captures behaviour but not the meaning, intent, or cultural interpretation behind it.
  • The choice of sampling rule constrains what can be estimated; the wrong rule yields biased or uninterpretable counts.
  • Building reliable, mutually exclusive codes is labor-intensive and never fully eliminates judgment.
  • Observer presence (reactivity) can alter the behaviour being coded, especially in small or sensitive settings.

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Replaces impressionistic field notes with reproducible, rule-governed records that other researchers can replicate.
  • Different sampling rules let the design target frequencies, durations, states, or sequences as the question requires.
  • Explicit coding schemes yield discrete, countable data ready for quantitative comparison and statistics.
  • Built-in reliability checks make observer bias measurable and reportable rather than hidden.

Applications

  • Coding caregiver–child interaction or social behaviour for developmental and psychological anthropology.
  • Recording work and activity states under scan sampling for time-allocation studies.
  • Quantifying ritual, play, or conflict behaviour with focal-individual sampling.
  • Measuring the frequency and duration of specific acts in classroom, clinic, or workplace ethnography.
  • Establishing inter-observer reliability for any structured-observation protocol before analysis.

How it works

How it works

1

The analyst first builds an ethogram or code book: a list of behaviour categories, each with an operational definition precise enough that different observers apply it the same way, covering the behaviours of interest mutually exclusively so any observed act maps to one code.

2

A sampling rule is chosen to fit the question — focal-individual sampling tracks one subject for a set period, scan sampling records the whole group's states at fixed instants, ad libitum noting captures only salient events, and continuous (all-occurrences) sampling logs every onset and offset; the rule determines what the data can and cannot estimate.

3

Observers apply the chosen rule in the field or to recordings, assigning code-book categories to behaviour with time stamps, often with onset and offset times so that both frequencies and durations can later be derived from the coded record.

4

Two or more observers independently code the same material and their agreement is quantified with a chance-corrected statistic such as Cohen's kappa, κ = (p_o − p_e) / (1 − p_e), where p_o is observed agreement and p_e is the agreement expected by chance; adequate κ is reported before substantive analysis proceeds.

Formula
\kappa = \frac{p_o - p_e}{1 - p_e}

Common pitfalls

  • Reporting raw percent agreement instead of a chance-corrected statistic like kappa, overstating reliability.
  • Mixing sampling rules within a study and then comparing counts as if they measured the same thing.
  • Coding from vague category definitions, so observers silently diverge and reliability collapses.
  • Using ad libitum sampling and then treating the conspicuous events it captures as representative frequencies.

Frequently asked questions

Which sampling rule should I choose?

Match the rule to the quantity you need: focal-individual sampling for accurate per-subject rates and durations, scan sampling for the proportion of a group in each state at sampled instants, continuous sampling for full sequences and exact timing, and ad libitum only for opportunistic notes on rare conspicuous events. Each rule biases the data differently, so the choice is part of the design, not a convenience.

What is Cohen's kappa and why not just use percent agreement?

Kappa, κ = (p_o − p_e) / (1 − p_e), corrects observed agreement for the agreement expected purely by chance, so it is not inflated when one category is very common. Raw percent agreement can look high simply because observers default to the same frequent code, which is why chance-corrected statistics are the standard for inter-observer reliability.

How is this different from spot observation sampling?

Spot observation is one specific instantaneous design aimed at estimating time allocation across a population. Behavioral observation coding is the broader framework of sampling rules and coding schemes — including focal, scan, continuous, and ad libitum — within which spot sampling is essentially an instantaneous scan applied to people.

Related methods

Sources

  1. Altmann, J. (1974). Observational study of behavior: sampling methods. Behaviour, 49(3–4), 227–267.

    Open DOI reference
    Copied.
  2. Bernard, H. R. (2017). Research Methods in Anthropology: Qualitative and Quantitative Approaches (6th ed.). Lanham, MD: Rowman & Littlefield.

    ISBN: 9780759112421
    Copied.