Before you trust a data point, ask: what exactly was measured, and is that the behavior you care about?
Why it matters
Your decisions are only as good as your data. If you measure something other than the target behavior, you have to infer what the behavior did, and every inference adds room for error. If you measure the results a behavior leaves behind, you need to be sure those results came from that behavior and nobody else.
Practice items on this task typically describe a data-collection plan and ask you to name the kind of measure, pick the one that needs the least inference, or spot why a product measure will give misleading data.
Key ideas
- Direct measurement means the thing you measure is the same as the behavior you are targeting. If the goal is fewer tantrums and an observer records each tantrum as it happens, the measure is direct.
- Indirect measurement means you measure something different from the target behavior, such as a report, rating, or questionnaire about it, and then infer what the behavior is like. Interviews, rating scales, checklists, and caregiver estimates are the usual examples.
- Direct measurement is generally preferred because it requires the least inference about whether the data reflect the behavior. Indirect measures still have a place: they are quick, capture the views of people who know the client, and guide what you observe next. Treat them as a starting point, not as proof.
- Permanent-product measurement means you measure the lasting effect a behavior has on the environment, after the behavior has occurred. Completed worksheets, assembled kits, made beds, and items put away are classic products.
- Two rules for product measures. Use one only when (1) each occurrence of the behavior produces the same product, and (2) the product can come only from the target behavior, not from another person or another cause.
- "Not watched live" does not mean "indirect." A product measure is not indirect just because nobody saw the behavior happen. What makes a measure indirect is that you measured a report or rating about the behavior instead of the behavior or its product.
Definitions
- Direct measurement: measurement in which the event recorded is the target behavior itself, for example counting each request a learner makes during a session.
- Indirect measurement: measurement of something other than the target behavior, such as an interview answer or a rating, from which the behavior must be inferred.
- Permanent-product measurement: measuring behavior after it has occurred by recording the effects it left on the environment.
- Inference: a conclusion about the behavior that goes beyond what was actually measured. The more inference a measure needs, the less confident you can be that it reflects the behavior.
Examples from a session
One goal, three kinds of measure
Ava, 9, has a goal to complete independent math work at school.
- Direct: An observer sits in the classroom and records each problem Ava works on and finishes during independent work time.
- Permanent product: At the end of the day, the BCBA counts the problems Ava answered correctly on her worksheet.
- Indirect: Ava's teacher completes a weekly rating, from 1 to 5, of how "focused" Ava was during math.
The worksheet count is efficient: no one has to watch, the sheet can be scored later, and a second person can score the same sheet to check agreement. It has limits, though. It shows the answers, not how Ava got them. If a classmate filled in half the sheet, the data would overstate her work.
A few more examples:
- Vocational setting: Marcus, 28, assembles kits at his own station. Counting finished kits at the end of the shift is a sound product measure, because each assembly yields one kit and no one else builds at that station.
- Home: A parent says Leo, 5, hits his sister "about ten times a day." That estimate is indirect. It is useful for planning, but you would confirm it with direct observation before building a treatment plan on it.
- Clinic: Counting new scabs to track skin picking breaks the first rule. Not every pick leaves a mark, and scabs can come from other injuries.
Tip
For behaviors that leave no natural product, such as hand flapping or vocal requests, you can record the session on video and score the recording later. Some texts call this a contrived permanent product. You still have to define and count the behavior carefully when you score it.
Common mistakes / exam traps
- Ignoring other sources of the product. Shared rooms, shared desks, and adult help can all produce the same product.
- Assuming a product shows the process. A correct answer does not tell you whether the learner computed it, guessed, or copied.
- Treating a rating scale as the outcome measure. Ratings can guide hypotheses and capture social validity, but they do not replace direct measures of the target.
Permanent product vs. indirect
Ask "What was actually recorded?" A completed worksheet or a counted kit is a lasting result of the behavior itself, so it is a permanent product, even though no one watched. A report, rating, or recollection about the behavior is indirect. If it is the behavior as it happens, the measure is direct.
Discontinuous vs. indirect
Partial-interval recording and momentary time sampling estimate the behavior, but the observer still records the behavior itself, so they are direct measures. "Estimate" does not mean "indirect." Continuous vs. discontinuous is a separate question from direct vs. indirect; a 1-to-5 rating is indirect however often it is filled in.
Quick check
A BCBA is tracking tantrums for Sofia, 6. Which of these measures requires the most inference about how often or how long Sofia's tantrums actually occur?
Related topics
- Writing operational definitions (C.1)
- Measuring occurrence: count, rate, and percentage (C.3)
- Continuous vs. discontinuous measurement (C.5)
- Judging the validity and reliability of measurement (C.8)
- Choosing a measure that captures the critical dimension within real-world limits (C.9)
- Descriptive assessments: ABC recording, scatterplots, and more (F.5)