RBT Measurement Practice Questions

Duration recording depends on knowing exactly when each episode begins and ends. When the onset is ambiguous, the RBT should ask the supervising BCBA right away so every episode is timed the same way from the start. Starting the timer at whichever point seems most accurate lets the onset shift from episode to episode, so durations would not be comparable and interobserver agreement would suffer. Revising the definition is the supervisor's responsibility; an RBT may not change an operational definition alone. Timing from loud crying only and waiting until next month's supervision is the right concern raised at the wrong time, because weeks of data could be collected under a guessed definition.

Latencies under 20 seconds must be timed in seconds, so the RBT needs a stopwatch or timer that displays seconds before collecting data, and should tell the supervisor if one is not available. A wall clock without a second hand cannot capture short latency, duration, or IRT values. Rounding each latency to the nearest minute would record nearly every response as zero minutes, hiding any change in how quickly Andre starts. Estimating seconds by watching the minute hand is guesswork that produces unreliable data. Switching to frequency recording changes the measurement procedure, which is a supervisor decision, and a count of responses would not show how quickly Andre begins the task anyway.

Counting is appropriate only when each response has a clear, easily identified beginning and end. Throwing an object onto the floor is a discrete event: it starts when the object leaves his hand and ends when it lands, so each instance can be counted reliably. Rocking for long stretches has no obvious point where one response ends and the next begins, so duration or interval recording would capture it better. Humming throughout an activity is similarly continuous, and observers would disagree about how many hums occurred. Staring at the fan lacks a clear start and stop and can last any length of time, which makes a simple count misleading.

Event recording works best when each counted response takes roughly the same amount of time, so every tally represents a similar amount of behavior. When one tantrum lasts 5 seconds and another lasts 20 minutes, each still counts as one, and the frequency data can hide meaningful changes, such as fewer but much longer tantrums. Raising this lets the supervisor decide whether duration should be added. Tallies can be graphed even when intensity varies; intensity is a separate dimension, not a barrier to graphing. Counts do not require the behavior to occur every session, since a zero is a valid data point. Converting tallies to percentages is not required for frequency data, which are usually graphed as counts or rates.

Scored-interval IOA is recommended for low-rate behavior because it compares only intervals in which at least one observer recorded the behavior. With aggression scored in just 3 of 60 intervals, this avoids inflating agreement with dozens of intervals where nothing happened. Unscored-interval IOA does the opposite, comparing only intervals where at least one observer did not score the behavior, so it suits high-rate behavior. Interval-by-interval IOA includes every interval, so the many agreed non-occurrences would make agreement look high even if the observers never scored the same intervals. Total count IOA compares only overall totals, so both observers scoring 3 intervals could yield perfect agreement even if they scored different intervals.

Rate is count divided by time, so when every session lasts exactly 30 minutes, each count is divided by the same number. The graphed trend for count therefore rises and falls exactly as the rate trend does, and converting is not needed to see the pattern, although the supervisor may still request rate. Saying counts can never be graphed without conversion is incorrect, since count is a valid measure when observation time is constant. Rate does not produce a steeper trend than count in equal sessions; the shape is the same, only the axis numbers change. Rate is not tied to duration recording at all; it is the standard way to report frequency across varying observation times.

Partial-interval recording is typically selected when the goal is to decrease a problem behavior. Scoring an interval whenever hitting occurs at any point tends to overestimate the behavior, which is a conservative error for a reduction goal, because improvement must be clear before it shows on the graph. Whole-interval recording requires the behavior to last the entire interval and tends to underestimate it, so it is usually chosen for behaviors to increase, such as sustained on-task work. Momentary time sampling scores only the instant an interval ends and could miss brief hits entirely. Planned activity check is a form of momentary time sampling for groups, so it is not suited to one client's brief hitting.

Interresponse time is the amount of time between two responses, so IRT data must always be reported in units of time, such as seconds or minutes. Writing 6 seconds tells the supervisor that Ivy paused an average of 6 seconds between bites. Writing 6 bites reports a count of responses, which describes how many times she ate rather than the time between bites. Writing 6 per minute reports a rate, a different dimension that counts responses per unit of time instead of measuring the gap between them. Writing 6 percent describes a proportion, such as the percentage of intervals or trials, and cannot express how long Ivy waited between responses.

Interval-by-interval IOA checks, for each interval, whether both observers recorded the same score, then divides agreements by total intervals and multiplies by 100. The observers agreed on occurrence in 3 intervals and on non-occurrence in 3, giving 6 agreements out of 10, or 60 percent. Reporting 30 percent counts only the occurrence agreements and ignores intervals where both agreed the behavior did not happen. Reporting 43 percent is scored-interval IOA, which compares only the 7 intervals in which at least one observer scored hand-flapping. Reporting 100 percent compares the two totals of 5, which hides the fact that the observers disagreed on 4 specific intervals.

Scoring an occurrence is correct because whole-interval recording scores an interval only when the behavior occurs throughout the entire interval. Leo was working from the first second of interval 6 to the last, so the interval is scored even though his working began earlier and continued afterward. Scoring a nonoccurrence because the behavior began before the interval adds a requirement that does not exist; when it started is irrelevant if it covered the whole interval. Scoring a nonoccurrence because it did not end within the interval is the same error at the other end. Whole-interval recording has no partial or half credit; each interval is scored as either an occurrence or a nonoccurrence.