Defining and Observing Behavior

Before you can count or time a behavior, everyone on the team has to agree on exactly what counts. An operational definition is that written agreement. It is a clear, specific description of the behavior to be measured. Once the definition is clear, you can measure different things about the behavior, such as how long it lasts or how quickly it starts.

What you will be able to do

• Rewrite a vague label, such as "lazy," as an observable, measurable, actively stated description.
• Use examples, nonexamples and start and end points to decide whether an instance counts.
• Describe what happened around the client without guessing their intent or the cause.
• Calculate duration, latency and interresponse time from a timeline, and say what question each one answers.
• Recognize when to ask your supervisor instead of changing a definition yourself.

You need: subtraction with minutes and seconds. Each step is shown below.

Describe what you could see or hear

A good definition passes three tests:

• Observable: it says what the behavior looks and sounds like. It does not say what the person might be thinking or feeling. Takeaway: could you point to it on a video?
• Measurable: it can be counted or timed.
• Actively stated: it says what the person does, not what they fail to do.

Why "lazy" is not a recording definition

"Sam is lazy during math" sounds like a description, but it is a judgment about Sam. It names no action. One observer might count slow writing as "lazy." Another might count staring out the window. Their data would not agree. The label also describes something missing, so there is nothing to count or time.

Instead, name an action. Here is one possible rewrite: "Head down: Sam's forehead or cheek touches the desk or Sam's folded arms." Two observers can now count head-down instances and get the same result.

Describe the environment the same way

The same rule applies to events around the client. "The teacher said, 'Open your book to page 12'" is observable. "The teacher stressed Sam out" is a guess about feelings.

Likewise, "Sam threw the pencil to get out of work" contains a guess about purpose. Write "Sam threw the pencil; it landed about one meter away." Then write what happened right before and right after the throw. If one event comes right before another, that order does not prove the first event caused the second. Your job is to record what happened accurately, not to decide why.

Examples, nonexamples and boundaries

Examples and nonexamples show where the line falls. A definition also needs a clear start and end, so you know when one instance begins and finishes. The start and end should use the same test as the rest of the definition. Otherwise, two observers could disagree about when an instance is over.

Here is a definition used below: "Out of seat: Rae's bottom is not touching the chair seat, unless an adult has given permission. It starts when Rae's bottom leaves the seat. It ends when Rae's bottom touches the seat again."

Notice that one test, is Rae's bottom on the seat?, decides everything: whether an instance is happening, when it starts and when it ends.

• Example: Rae stands and walks to the window.
• Example: Rae kneels on the chair. Rae is still on the chair, but Rae's bottom is not touching the seat, so this counts.
• Nonexample: Rae leans far forward to reach a marker. Rae's bottom stays on the seat.
• Nonexample: Rae stands to get a tissue after the RBT says, "Go ahead."

Some cases still won't fit cleanly. Suppose Rae shifts position and Rae's bottom lifts off the seat for about one second before landing again. The written words say that counts, but the definition never says whether a brief shift like this was meant to be scored. Do not decide on your own what the definition "really means." Write down exactly what you saw, and ask your supervisor promptly how to score it. RBTs follow supervisor-approved definitions and direct questions about their services to their supervisor.

Check: Which description is observable, measurable and actively stated? (a) "Jordan was frustrated with the puzzle." (b) "Jordan didn't pay attention." (c) "Jordan pushed puzzle pieces off the table with one hand."

Answer: only (c)

(a) describes a feeling someone inferred from Jordan's behavior, not something seen. (b) describes something missing and names no action to count. (c) names an action you can see and count.

Missing time is not zero

A zero means "I was watching the whole time, and the behavior did not happen." Suppose you stepped away or were helping someone else for part of the session. Writing zero for that time would report an observation you never made. Accurate documentation means recording that the period was not observed, using the method your plan or supervisor specifies.

Three ways to time behavior

Frequency (event recording) means counting instances. It works well for brief behaviors with a clear start and end. Three timing measures answer different questions:

• Duration: How long did each instance last? Measure from start to end. Use it when the amount of time spent in the behavior is the main concern.
• Latency: How long after an instruction did the behavior begin? Measure from the instruction to the response. Some plans time to when the behavior ends instead, so check yours.
• Interresponse time (IRT): How long was the gap between one instance and the next? IRT is the time that passes between two responses in a row. A response is one instance of the behavior. In this lesson's examples, we measure IRT from the end of one instance to the start of the next. That is the convention chosen for these examples. Your plan states which points to use.

Worked timeline

Times are minutes:seconds from the start of the observation.

• 0:00 The RBT says, "Rae, come sit at the table."
• 0:25 Rae sits.
• 2:10 Rae stands and walks to the window. (Out of seat starts.)
• 2:40 Rae sits.
• 3:30 Rae stands and walks to a shelf.
• 4:20 Rae sits.
• 6:00 Rae stands.
• 6:15 Rae sits.

Latency: 0:25 − 0:00 = 25 seconds from the instruction to sitting.

Frequency: 3 out-of-seat instances.

Duration:

• 2:40 − 2:10 = 30 s
• 4:20 − 3:30: from 3:30 to 4:00 is 30 s, and from 4:00 to 4:20 is 20 s, so 50 s
• 6:15 − 6:00 = 15 s
• Total: 30 + 50 + 15 = 95 seconds, or 1 minute 35 seconds

IRT (end to next start):

• From 2:40 to 3:30: 20 s to 3:00, then 30 s more, so 50 s
• From 4:20 to 6:00: 40 s to 5:00, then 60 s more, so 100 s (1 minute 40 seconds)

Notice that the second duration and the first IRT are both 50 seconds. The numbers match, but they answer different questions. One measures time in the behavior. The other measures time between instances.

Change one fact: suppose a plan measured IRT from the start of one instance to the start of the next. The first IRT would then be 3:30 − 2:10 = 80 seconds. The events are the same, but the anchor point changed, so the number changed. Never compare two numbers unless you know their anchor points.

Counts need context too. Suppose one observation lasted 7 minutes and another lasted 30 minutes, and you counted 3 instances in each. The count is the same: 3. But that count alone can't tell you whether the behavior happened about as often in both sessions, because you watched for very different lengths of time. Always record how long you observed alongside the count. Combining a count with observation time into one number is a separate calculation that this lesson does not cover.

Finally, the timeline shows when Rae stood, not why.

Try it yourself

Checkpoint 1. At 0:00 an RBT says, "Time to read." Jamie opens the book at 0:35. Later, Jamie paces at two separate times: from 1:05 to 1:50, and from 3:20 to 3:35. Find the latency, each pacing duration, the total duration, and the IRT, measuring from end to next start. A coworker reports the IRT as 2 minutes 15 seconds. What did the coworker measure?

Answer: latency 35 s; durations 45 s and 15 s (total 60 s); IRT 90 s

Latency: 0:35 − 0:00 = 35 s.

Durations: 1:50 − 1:05 = 45 s, and 3:35 − 3:20 = 15 s. The total is 45 + 15 = 60 s.

IRT: from 1:50 to 3:20 is 10 s to 2:00, plus 1:20 more, which is 90 s.

The coworker's 2:15 equals 3:20 − 1:05, so they measured from start to start. Neither of you is necessarily wrong, but your numbers can't be compared until you use the same anchor points. Ask your supervisor which convention the plan uses.

Checkpoint 2. At 10:05, the RBT placed a worksheet in front of Lee and said, "Finish these five problems." Lee tore the worksheet, and two pieces separated. The RBT then picked up the pieces. At 10:08, Lee creased a page sharply, but no piece came off. From 10:15 to 10:25, the RBT was on the phone with a caregiver. Lee's plan defines ripping as "tearing paper so that a piece separates."

A draft note says: "Lee got bored and ripped the worksheet to avoid work. Zero ripping from 10:15 to 10:25." What should change in the note, and what should the RBT do about the crease?

Answer: remove the guesses, don't record zero, and apply the definition to the crease

"Bored" and "to avoid work" are interpretations. A better note sticks to what the RBT saw: "At 10:05, the RBT placed a worksheet in front of Lee and said, 'Finish these five problems.' Lee tore the worksheet; two pieces separated. The RBT picked up the pieces." Every detail in that note was observed. Nothing explains why Lee tore the paper.

The RBT was not watching from 10:15 to 10:25, so that period should be marked as not observed, not as zero.

Start with the written definition. The crease is a nonexample because no piece separated, so it is not scored as ripping. Only if the team is unsure whether creasing was meant to count should the RBT ask the supervisor, rather than redefine "ripping" alone.

Checkpoint 3. A note says: "Ava is disruptive during circle time." Rewrite it as a definition that is observable, measurable and actively stated. Choose one specific behavior, and give a start point, an end point, one example and one nonexample.

Answer: sample rewrite (yours may name a different behavior)

"Disruptive" is a judgment, like "lazy." Two observers could count completely different things. A rewrite must pick one action you could point to on a video.

One possible answer: "Calling out: Ava speaks loud enough for the group to hear while another person has the turn to talk. It starts when Ava begins speaking loud enough for the group to hear while another person has the turn. It ends when that stops being true: Ava stops speaking, Ava's speech is no longer loud enough for the group to hear, or the other person's turn ends."

Notice that one test, is Ava speaking loud enough for the group to hear while someone else has the turn?, decides everything: whether an instance is happening, when it starts and when it ends.

• Example: While the teacher reads, Ava says loudly, "I have that book!"
• Nonexample: Ava answers after the teacher says, "Ava, your turn."

The same test handles a boundary case. If Ava is calling out and the teacher then says, "Ava, your turn," the instance ends at that moment, because no one else has the turn anymore. If a team wants short pauses grouped into one instance, that is an extra rule the supervisor would write into the definition; it is not something an observer adds while recording.

Check your own rewrite against the three tests:

• Could you see or hear it? "Bothering others" fails; "speaks while someone else has the turn" passes.
• Could you count or time it?
• Does it say what Ava does? "Doesn't sit quietly" fails because it describes something missing.

Also check that your start and end use the same test as the rest of your definition. In a real plan, your supervisor approves the definition before you collect data with it.

Recap

• Define behavior by what you can see or hear, count or time, and what the person does.
• Labels such as "lazy" or "disruptive" are judgments, not recording definitions.
• A definition's start and end should use the same test as the rest of the definition.
• Describe the environment without guessing intent. Write only details you actually observed. Events happening in order do not prove one caused the other.
• Unobserved time is not zero.
• A count means little without knowing how long you observed. Record both.
• Duration measures time in the behavior, latency measures time from an instruction, and IRT measures time between instances. Always state your anchor points.
• Apply the written definition first. Follow supervisor-approved definitions, and ask when a case doesn't fit.

Sources

• IRIS Center, Vanderbilt University. Measuring Behavior (case study). Operational definitions; frequency, duration and latency recording. https://iris.peabody.vanderbilt.edu/case-study/measuring-behavior/
• Cooper, J. O., Heron, T. E., & Heward, W. L. Applied Behavior Analysis (2nd ed.). Pearson. Glossary entry "interresponse time (IRT)," p. 10 of the glossary.
• Behavior Analyst Certification Board. (2023). RBT Test Content Outline (3rd ed.). https://www.bacb.com/wp-content/rbt-outline-3rdEd/
• Behavior Analyst Certification Board. (2021). RBT Ethics Code (2.0), standards 2.02 and 2.06. https://www.bacb.com/wp-content/rbt-ethics-code

Independent study resource

This study resource is not affiliated with, sponsored by, or endorsed by the Behavior Analyst Certification Board (BACB). BACB, Registered Behavior Technician and RBT are registered trademarks of the Behavior Analyst Certification Board. Source citations do not imply endorsement by their authors or publishers.