4 Behavioral Data Interpretation and Practice

Choose and document a measure

Before calculating or graphing, record the target behavior’s operational definition, date or session, observation length, measurement method, and any relevant change in setting or procedure. Keep units visible, such as responses per minute, seconds, or percent of intervals. A count without its observation time can mislead when session lengths differ.

Choose a measure according to the question:

• Use count or rate for discrete events with a clear beginning and end. Choose rate when observation times differ or a time-based comparison matters.

• Use duration when time spent in behavior is the main concern, such as the length of crying episodes or time on task.

• Use latency when the question concerns delay after a defined cue, such as time from an instruction to beginning work.

• Use interval recording when continuous observation or event-by-event recording is impractical. Interpret its results as estimates, not exact event counts or duration.

Frequency and rate are not always used consistently in everyday practice: “frequency” can mean either a count or a count per unit of time. To avoid ambiguity, report the count and observation length, and label a calculated rate with its unit. For example, report “$12$ call-outs in $30$ minutes; $0.4$ call-outs per minute.”

Partial-interval recording marks an interval if behavior occurs at any time during it; whole-interval recording marks it only if behavior continues for the entire interval. These percentages are not exact measures of duration: partial-interval recording tends to overestimate time engaged, while whole-interval recording tends to underestimate it. Momentary time sampling records whether behavior is occurring at a specified instant.

Calculate and compare measures

Use consistent units and state what each result represents.

Rate

Calculate rate by dividing the number of occurrences by observation time:

For example, $12$ call-outs during $30$ minutes is $\frac{12}{30}=0.4$ call-outs per minute. For $18$ errors in a $45$-minute session, the rate is $\frac{18}{45}=0.4$ errors per minute, or $24$ errors per hour. Since $45$ minutes is $0.75$ hour, the hourly calculation is $\frac{18}{0.75}=24$ errors per hour. Convert units carefully.

Duration and latency

If episodes last $2$, $3$, and $7$ minutes, total duration is $12$ minutes and mean duration is $\frac{12}{3}=4$ minutes per episode. In an observation lasting $40$ minutes, the behavior occupied $\frac{12}{40}\times100=30\%$ of observed time.

If latency measurements after three instructions are $8$, $14$, and $11$ seconds, mean latency is $\frac{8+14+11}{3}=11$ seconds. A lower latency means the response began sooner; it does not necessarily mean that behavior occurred more often or lasted longer.

Interval percentage

Calculate an interval percentage by dividing the number of intervals scored “yes” by the total number of intervals, then multiplying by $100$. If behavior occurred in $7$ of $12$ partial intervals, the result is $\frac{7}{12}\times100=58.3\%$ of intervals. Report the method and interval length, for example, “$58.3\%$ of $1$-minute partial intervals.” This does not mean the behavior occupied $58.3\%$ of the session.

Compare like with like

A count of $10$ events, a rate of $0.5$ events per minute, $10$ minutes of behavior, and $50\%$ of intervals are different results. Even two rates may not be directly comparable if behavior definitions, observation contexts, or data-collection methods differ. Before drawing a comparison, check that the measure, units, observation conditions, and opportunities to respond are sufficiently alike.

Organize and graph the data

A clear data record supports accurate calculation and graphing. For call-outs observed across sessions of different lengths, the data are:

• Session $1$: $30$ minutes, $12$ call-outs, rate $0.40$ call-outs per minute.

• Session $2$: $20$ minutes, $10$ call-outs, rate $0.50$ call-outs per minute.

• Session $3$: $30$ minutes, $9$ call-outs, rate $0.30$ call-outs per minute.

• Session $4$: $15$ minutes, $9$ call-outs, rate $0.60$ call-outs per minute.

• Session $5$: $30$ minutes, $6$ call-outs, rate $0.20$ call-outs per minute.

Sessions $3$ and $4$ each had $9$ call-outs, but their rates differ because observation times differ: $0.30$ versus $0.60$ per minute. Graphing the raw counts alone would hide that distinction.

For a session-by-session line graph, put session or date on the horizontal axis and the measured value on the vertical axis. Label both axes, including units, and use an even, consistent scale. Plot each value in order and connect successive points when the data are sequential observations. Mark a phase change, such as the start of an intervention, with a vertical line and label the phases. Keep graphs simple and uncluttered, identify the behavior and measurement method, and use separate graphs or clearly distinguished series when measures or targets differ.

Select a vertical-axis scale that accommodates all values and makes changes readable. Avoid changing the scale between graphs in a way that makes similar changes appear very different. Equal-interval graphs and clear labels support accurate communication of measurement data.

Interpret patterns over time

Visual interpretation considers what the data do over time, not just whether the last point is high or low. Examine several features together:

• Level: the typical height of data in a session or phase. Compare the general range or center of one phase with another.

• Trend: the overall direction of change—rising, falling, or relatively flat.

• Variability: how much values fluctuate. A wide or irregular spread makes the typical level and direction less certain.

• Immediacy and overlap: whether values change soon after a phase change, and how much values in adjacent phases overlap.

The meaning of a direction depends on the goal and the measure. For behavior targeted for decrease, a falling rate may be favorable; for a skill targeted for increase, a rising rate may be favorable. A falling latency to start work can indicate faster initiation, whereas a falling duration of on-task behavior may be undesirable.

A graph can show a pattern, but a simple before-and-after change alone does not establish why the change occurred. Consider consistency of measurement, observation opportunities, context, and other changes when interpreting results.

Avoid common interpretation errors

Common errors can distort comparisons or lead to conclusions stronger than the data support:

• Comparing counts from unequal observation times: a larger count may reflect a longer session rather than a higher rate. Keep time constant or calculate rate.

• Dropping units: “$0.4$” is unclear; “$0.4$ responses per minute” is interpretable.

• Mixing total and mean duration: total duration is time across episodes; mean duration is time per episode. State which result is reported.

• Calling an interval percentage a percent of time: interval scoring reports the percentage of intervals meeting a scoring rule, not exact duration.

• Treating partial- and whole-interval percentages as equivalent: their scoring rules differ and can bias estimates in opposite directions.

• Comparing different behaviors or procedures without qualification: a measure is meaningful only with its operational definition and collection method.

• Reading too much into one point: consider the full sequence, level, trend, variability, and phase context.

• Ignoring opportunities or context: fewer responses may reflect fewer chances to respond or a different setting, rather than a genuine change in behavior.

• Using unclear or misleading graph scales: label axes and units, use consistent intervals, and show phase changes accurately.

Practice calculations and interpretation

Work through each problem, then compare your calculations and interpretations with the answers below.

Exercise A: Frequency and rate

A behavior occurs $15$ times in a $25$-minute observation and $12$ times in a $40$-minute observation.

• Find the rate in each observation, in occurrences per minute.

• Identify which observation had the higher count and which had the higher rate.

• Explain why comparing counts alone would be misleading.

Exercise B: Duration

Four episodes last $30$, $45$, $15$, and $50$ seconds during a $10$-minute observation.

• Find total duration.

• Find mean duration per episode.

• Find the percentage of the observation spent in behavior.

Exercise C: Latency

After three instructions, a learner begins responding after $6$, $10$, and $14$ seconds.

• Find mean latency.

• If the next mean is $9$ seconds, state what changed and what that result does not tell you by itself.

Exercise D: Interval recording

During $10$ one-minute intervals, behavior occurred at least once in $6$ intervals. It continued for the entire interval in $2$ intervals.

• Calculate the partial-interval percentage.

• Calculate the whole-interval percentage.

• Explain why neither percentage is the exact percentage of time spent in behavior.

Exercise E: Interpret a trend

A target behavior’s rate across five sessions is $0.8$, $0.7$, $0.7$, $0.4$, and $0.3$ per minute. The goal is to reduce the behavior.

• Describe the overall trend.

• Identify one reason not to base a conclusion on the final point alone.

• Name two checks to make before comparing this series with another phase.

Answers

• A: The rates are $\frac{15}{25}=0.60$ per minute and $\frac{12}{40}=0.30$ per minute. The first observation has both the higher count and the higher rate. Counts alone can mislead because observation times differ.

• B: Total duration is $30+45+15+50=140$ seconds. Mean duration is $\frac{140}{4}=35$ seconds per episode. The observation lasted $600$ seconds, so the percentage is $\frac{140}{600}\times100=23.3\%$.

• C: Mean latency is $\frac{6+10+14}{3}=10$ seconds. A mean of $9$ seconds is a decrease of $1$ second, indicating a shorter average delay. It does not establish how often the behavior occurred or how long it lasted.

• D: The partial-interval percentage is $\frac{6}{10}\times100=60\%$. The whole-interval percentage is $\frac{2}{10}\times100=20\%$. Each percentage follows a different interval rule and samples behavior occurrence rather than exact duration.

• E: The values show an overall decreasing trend. A single low final value may not represent the broader pattern, so consider the full series and its variability. Before comparing with another phase, check that the behavior definition, observation duration, setting and opportunities, and measurement procedure are comparable.