2 Measurement and Data Collection

Define behavior objectively

Behavior measurement turns observable actions into data that can be recorded, summarized, and reviewed. Accurate data depend on following the supervisor’s measurement plan, using consistent definitions, and recording what happened rather than what it seemed to mean.

Use observable definitions

An operational definition describes behavior in observable, measurable terms so different observers can identify and record it consistently. It should clarify what counts as an occurrence and, when relevant, when an occurrence begins and ends. Examples and nonexamples can help distinguish similar actions.

For example, “The learner is disruptive” is vague. A more observable description is: “The learner speaks above conversational volume while another person is speaking. An occurrence begins when the learner’s voice rises above conversational volume and ends after $5$ consecutive seconds at or below conversational volume.”

Words such as upset, lazy, or uncooperative describe interpretations rather than directly observed actions. Record the behavior and relevant environmental events. For example, “After the direction to put away the blocks, the learner pushed the box off the table” describes observable events more objectively than “The learner was defiant.”

Collect and record data consistently

Before a session, confirm the target definition, measurement method, observation period, recording form, and any rules for starting or stopping a timer. During the session, observe the defined behavior and record data as close to the event as the procedure allows.

Follow the same procedure throughout. Do not guess, fill in missed data from memory, or independently change the definition or method. Record the observation time or number of opportunities when needed to interpret a count.

Note relevant, directly observed circumstances—such as an illness, schedule change, or interrupted session—in the required documentation. Report unclear directions, missed observations, unusual data, or recording errors promptly to the supervisor. Session notes should describe events objectively, follow workplace requirements, and protect client confidentiality.

Measure each occurrence or dimension

Continuous measurement records each occurrence or another specified dimension of behavior throughout the observation period. Choose a measure that fits both the behavior and the question being asked.

• Frequency (count): The number of occurrences. For example, a learner raises a hand $6$ times.

• Rate: The count divided by observation time. For example, $6$ hand raises in $30$ minutes equals $0.2$ per minute, or $12$ per hour.

• Duration: How long behavior lasts; it can be totaled or averaged. Three episodes lasting $20$, $30$, and $40$ seconds total $90$ seconds, with a mean duration of $30$ seconds.

• Latency: The time from a specified event, such as an instruction, to the start of a response. For example, a learner begins work $12$ seconds after a direction.

• Interresponse time (IRT): The time between successive occurrences. For example, there are $45$ seconds between one hand raise and the next.

A count may suit brief, distinct actions such as hitting or raising a hand. Duration may better capture behavior that continues over time, such as crying. Latency measures how quickly a response begins after a defined event. A count without observation time can be misleading: $10$ occurrences in $20$ minutes represents a different rate from $10$ occurrences in $2$ hours.

Sample behavior with interval methods

Discontinuous measurement samples behavior during selected intervals or moments instead of recording every occurrence. It can be practical when continuous observation is difficult, but it produces an estimate and does not capture every detail of behavior.

Suppose an observation is divided into $10$ intervals of $1$ minute each:

• Partial-interval recording: Mark an interval if the behavior happens at any time during it, even briefly. If behavior occurs in $4$ intervals, the score is $\frac{4}{10} = 40\%$ of intervals. This method tends to overestimate how much behavior occurred.

• Whole-interval recording: Mark an interval only if behavior occurs throughout the entire interval. If behavior occurs for all $60$ seconds in $3$ intervals, the score is $\frac{3}{10} = 30\%$ of intervals. This method tends to underestimate behavior.

• Momentary time sampling (MTS): At the specified moment, often the end of each interval, record whether behavior is happening then. If behavior is happening at $4$ of $10$ checks, the score is $40\%$ of sampled moments. This is a sample-based estimate, not a count of occurrences or a direct measure of total duration.

Use the method and interval length specified in the plan. Changing the interval length or sampling moments can change the results. Partial-interval, whole-interval, and MTS scores are not interchangeable. A percentage of intervals is not necessarily the percentage of total time the behavior occurred.

Record products and surrounding events

Permanent-product recording measures the lasting result of behavior rather than observing the behavior as it happens. Counting correctly completed worksheet items is an example. Use this procedure only when the product reliably results from the target behavior and is not likely to be produced by other actions.

ABC recording documents an antecedent (what happened immediately before), the observable behavior, and a consequence (what happened immediately after). Describe events objectively. ABC notes can help a supervisor examine patterns, but a single observation does not establish why behavior occurred.

Calculate basic outcomes

Use the calculation that matches the data and its denominator:

• Rate:

  $$\text{Rate} = \frac{\text{occurrences}}{\text{observation time}}$$

  If $8$ episodes occur in $2$ hours, the rate is $4$ per hour.

• Mean duration:

  $$\text{Mean duration} = \frac{\text{total duration}}{\text{number of episodes}}$$

  If $90$ seconds are recorded across $3$ episodes, the mean is $30$ seconds.

• Percentage of opportunities:

  $$\text{Percentage of opportunities} = \frac{\text{successful responses}}{\text{total opportunities}} \times 100$$

  If a learner responds correctly on $7$ of $10$ trials, performance is $70\%$ of opportunities.

• Percentage of intervals:

  $$\text{Percentage of intervals} = \frac{\text{intervals scored for behavior}}{\text{total observed intervals}} \times 100$$

  This describes the selected interval measure, not necessarily the percentage of time behavior occurred.

Interpret data with context

Compare data only when observation conditions and definitions are sufficiently consistent. Consider observation length, number of opportunities, setting, and relevant changes. Look for broad patterns, such as an increase, decrease, or relatively stable level across observations; one data point alone may not indicate a meaningful trend.

Data can inform decisions, but the RBT should report and follow the supervisor’s direction rather than independently changing an intervention. Inaccurate or inconsistent data can lead to mistaken conclusions about progress or the need for a change. Interpret percentages according to the method used, and include relevant context when reviewing results.