Section C for the BCaBA: measuring behavior, graphing it, and reading the graph
Section C of the BCaBA® (Board Certified Assistant Behavior Analyst) exam, Measurement, Data Display, and Interpretation, tests how you measure a behavior, display the data and read the graph: 20 of the 150 scored items under the Test Content Outline (6th ed.), per the current BCaBA Handbook (updated June 2026). This page works through tasks C.1 to C.10 in the order they build on each other, does the arithmetic on paper, and ends with 12 practice questions.
What Section C covers on the BCaBA (20 scored items)
Twenty scored items make Section C 13% of the test and, with Section H, the second-largest block after Section G's 22. It is also the most mechanical section of the outline (formerly called the task list): name the dimension, match the recording method, run a formula, read three features of a graph. The map of all nine sections sits on the outline hub.
20scored itemsof 150, 6th-ed. outline
13%of the scored test
10tasksC.1 to C.10
The ten Section C tasks and where each is worked on this page
C.1 to C.3, until you name the dimension of any stem in one read.
2
Recording and its bias
C.4, C.5 and C.7: which method fits, and which way its error runs.
3
Agreement arithmetic
C.6: three IOA formulas on paper, each in under a minute.
4
Graphs, then criteria
C.8 to C.10, then the 12 trials at the end, untimed.
Under the 6th-edition outline the current BCaBA Handbook allows 4 hours for 175 questions, about 82 seconds each, so 20 scored items come to roughly 27 minutes of Section C work. An IOA item you have done twenty times on paper takes one minute of that; one you derive at the test center takes three.
02Dimensions
Which dimension are you measuring?
Most Section C stems are a dimension question in disguise. Before you read the options, decide whether the stem is about how often, how long, how soon after a cue, how far apart or how forceful the behavior is. That one decision usually removes two of the four options.
Count and rate
How many responses, and how many per unit of time. Use rate when session lengths differ.
Duration
Time from onset to offset of a response.
Latency
Time from a stimulus, such as an instruction, to the start of the response.
Inter-response time (IRT)
Time from the end of one response to the start of the next.
Magnitude
The force or intensity of a response.
Topography
The physical form of a response: a form, not an amount.
One timeline, three dimensions
0:00the teacher says to start the worksheet
0:06the student starts writing
0:16the student stops
0:40the student starts again
Latency0:06 - 0:00 = 6 seconds
Duration0:16 - 0:06 = 10 seconds
IRT0:40 - 0:16 = 24 seconds
Latency 6 s · duration 10 s · IRT 24 s
Hypothetical. Latency is anchored to the cue, IRT to the previous response.
Count or rate when sessions differ
Monday18 call-outs in 30 minutes: 18 / 30 = 0.6 per minute
Thursday12 call-outs in 15 minutes: 12 / 15 = 0.8 per minute
The smaller count is the higher rate
Operational definitions (C.2)
An operational definition lets two observers score the same behavior without comparing notes. Is aggressive names a trait; hits, kicks or pushes another person, counted from contact names actions with a start. A topography-based definition describes the form; a function-based one groups responses that produce the same outcome, such as throwing, shouting and leaving the table when all three end the task.
Observable: an action, not a state of mind.
Onset and offset stated.
Examples and non-examples, including the near misses.
Direct, indirect and permanent-product measures (C.3)
Where the data come from, and the catch
Swipe sideways to read all 3 columns
Measure
How the data arrive
The catch
Direct
Observed as it happens
Costs observer time; being watched can change behavior
Indirect
Interview, rating scale, recall
A report about behavior, not behavior
Permanent product
An outcome scored later, such as worksheets
Shows the result, not who produced it or how
03Recording
Continuous or discontinuous recording
Continuous methods (C.4) capture every instance: event recording for count and rate, plus duration, latency and IRT. Discontinuous methods (C.5) split the observation into intervals and score each one, so the result is a percentage of intervals, an estimate of how much of the time the behavior was present, never a count.
Partial-interval recording
Scores the interval if the behavior occurs at any moment in it.
Overestimates.
Suits behavior you plan to reduce.
Whole-interval recording
Scores the interval only if the behavior lasts all of it.
Underestimates.
Suits behavior you plan to increase, such as on-task work.
The tellAsk which way the error runs. Partial inflates and whole deflates, so pick the method whose error works against claiming progress that did not happen.
Momentary time sampling (MTS) scores only what is happening at the instant each interval ends. Its error runs either way, and it misses brief or rare behavior between the moments. PLACHECK is MTS for a group: at each moment you count how many people are engaged.
The same 60 seconds, recorded three ways
Swipe sideways to read all 5 columns
Interval
Behavior during it
Partial
Whole
MTS at end
0:00–0:10
from 0:05 on
yes
no
yes
0:10–0:20
until 0:12
yes
no
no
0:20–0:30
from 0:28 on
yes
no
yes
0:30–0:40
throughout
yes
yes
yes
0:40–0:50
until 0:47
yes
no
no
0:50–1:00
none
no
no
no
Hypothetical minute: two episodes, 0:05–0:12 and 0:28–0:47.
Three estimates of one minute
Actual time engaged7 s + 19 s = 26 s; 26 / 60 = 43%
Partial interval5 / 6 = 83%
Whole interval1 / 6 = 17%
Momentary time sampling3 / 6 = 50%
Partial reads high, whole reads low, MTS lands on either side
Choosing a system for the dimension and the setting (C.7)
1
Name the dimension
The goal's dimension decides what you record.
2
Can every instance be watched?
If yes, a continuous method gives a measure, not an estimate.
3
Match the error to the goal
Partial to decrease, whole to increase, MTS when the observer can only look up at set moments, such as a teacher recording while teaching.
4
Check the logistics
Who records, when, with what timer. A system staff cannot keep up with produces gaps, not data.
04Agreement
Is the measure any good? Validity, reliability and IOA
C.6 asks whether a number deserves trust. Keep four words apart, because the wrong options on these items are usually built from the other three.
Validity
The measure captures the behavior and dimension that matter for the goal.
Accuracy
The observed value matches the true value.
Reliability
Measuring the same events again gives the same value.
Interobserver agreement (IOA)
How closely two independent observers agree. It makes data believable, not accurate: two observers can share one error.
Nine IOA types and their arithmetic
Swipe sideways to read all 3 columns
IOA type
Data it fits
Arithmetic
Total count
Session counts
Smaller / larger × 100
Mean count-per-interval
Counts per interval
Smaller / larger per interval, then the mean
Exact count-per-interval
Counts per interval
Identical intervals / all intervals
Trial-by-trial
Discrete trials
Trials scored the same / all trials
Interval-by-interval
Interval records
Agreements / (agreements + disagreements)
Scored-interval
Low-rate behavior
Only intervals either observer scored
Unscored-interval
High-rate behavior
Only intervals either observer left blank
Total duration
Session time totals
Shorter / longer × 100
Mean duration-per-occurrence
Each occurrence timed
Shorter / longer per occurrence, then the mean
Every ratio is multiplied by 100 to give a percentage.
Total count IOA, worked
Observers A and B24 and 30 hand-raises
Smaller / larger24 / 30 = 0.80
Total count IOA = 80%
It cannot tell you whether A's 24 were among B's 30.
Mean count-per-interval IOA on a record that totals the same
Interval 1A 3, B 3: 1.00
Interval 2A 2, B 1: 0.50
Interval 3A 4, B 5: 0.80
Interval 4A 1, B 1: 1.00
Mean3.30 / 4 = 0.825
Mean count-per-interval IOA = 82.5%
Hypothetical. Both observers total 10, so total count IOA on this record reads 100%: the stricter formula finds what the total hides.
Interval-by-interval IOA, worked
Agreements17 of 20 intervals
17 / (17 + 3)0.85
Interval-by-interval IOA = 85%
Observer drift: scoring slides away from the definition; retrain against it.
Reactivity: people behave differently while watched.
Expectancy: knowing the phase nudges the scoring.
05Graphs
Graphs you must read at a glance
C.8 names three displays and C.9 asks you to read them. The equal-interval line graph is the default for session-by-session data; a bar graph compares conditions or summarizes totals; a cumulative record adds each period's responses to a running total, so its slope is the rate.
Four displays and how to read each
Swipe sideways to read all 3 columns
Display
Use it for
Read it by
Line graph
Repeated measures over sessions
Level, trend, variability per phase
Bar graph
Comparing conditions
Bar height; hides session-to-session change
Cumulative record
Responding added up over time
Slope: steep is fast, flat is none
Semi-log chart
Rates changing by multiples
Celeration per unit of time
Reading one phase change
Hypothetical. Baseline high, flat, variable; then an immediate drop in level and a downward trend to the aim line. The path breaks at the phase line.
Visual analysis, in the order to do it (C.9)
Level
Where the data sit on the y-axis, often the phase mean or median.
Trend
Direction and steepness: rising, falling or flat.
Variability
How far points bounce around the trend; high variability needs more sessions.
Immediacy and overlap
How fast level changes after the phase line, and how many new points fall inside the old range.
Describe each phase on its own before you compare any two of them.
The working order for C.9 itemsA baseline that was already falling
Hypothetical. The intervention continues the baseline's line: no jump in level, no change in slope, so no effect can be claimed.
A cumulative record of one 10-minute work period
Hypothetical. Minutes 4 to 6 run flat: no responding. Minutes 8 to 10 are the steepest: the highest rate. The line never falls.
Celeration lives on the fourth display. On a semi-logarithmic chart the y-axis is spaced by multiples, so 2 to 4 responses per minute covers the same distance as 20 to 40, and celeration is the change in rate written as a multiple, such as ×2 per week.
06Mastery
Setting and checking mastery criteria
C.10 asks you to develop a mastery criterion and then judge it. A usable one has three parts: a level on a named dimension (90% correct, 12 per minute), a span (three consecutive sessions) and conditions (two instructors, untrained materials).
Three criteria for the same goal, from weak to usable
Criterion
What is missing
8 of 10 pictures labeled correctly
A span: one good session would count
8 of 10 correct in 3 consecutive sessions
Conditions: one instructor, one picture set
8 of 10 untrained pictures correct in 3 consecutive sessions with 2 instructors
Nothing obvious; check it at follow-up
Judging a criterion is a data question. If skills that met it fall apart at a maintenance probe a few weeks later, it was too lax: raise the level, lengthen the span or add conditions. In the assistant's seat you bring that data to your supervising BCBA before the criterion changes.
07Wrong turns
Where Section C answers go wrong
Each is a pattern older prep material often gets wrong, and each takes about ten seconds to check once you know where to look.
08Practice
12 Section C practice questions
The 12 trials below are all Section C, each written for this site to a task of the 6th-edition outline, with a note on every option. Work them untimed. Each answer lands on the trial sheet as + or −; the cumulative record next to it steps up only on a right answer, so a run of errors shows as a level stretch under the dashed all-correct aim line. A flat stretch means what it meant in the 10-minute graph above.
12 Section C trials
Answer with 1–4 or A–D; the arrow keys step between trials
Trial sheet · 12 trials
Cumulative record: 0 correct over 0 answered
Trial 1 of 12Record open
CMeasurement, Data Display, and InterpretationC.1
Pick an option: the verdict, the reasoning and a note on all four options open here.
Duration measurement is most appropriate for measuring how long a behavior occurs from start to finish. Since the behavior analyst wants to know how long the student remains seated (a state behavior with extended duration), duration recording is the most appropriate dimension to measure.
ALatency is time from a stimulus to the start of a response.
BCorrect: how long the student stays seated is a duration measure.
CIntensity is the force of a response.
DFrequency counts instances and does not capture how long each lasts.
More notes below ↓
CMeasurement, Data Display, and InterpretationC.1
Pick an option: the verdict, the reasoning and a note on all four options open here.
Dividing the total number of responses by the duration of the session standardizes the count into a rate, allowing comparison across sessions of unequal length.
AMultiplying duration by count does not produce a rate.
BCorrect: rate is count divided by observation time, which allows comparison across sessions of different length.
CAveraging per session does not correct for session length.
DTime divided by responses gives the average time per response, the inverse of rate.
More notes below ↓
CMeasurement, Data Display, and InterpretationC.2
Pick an option: the verdict, the reasoning and a note on all four options open here.
The first step in conducting a behavior assessment is to develop an operational definition of the target behavior. This ensures all observers are measuring the same behavior and provides a clear, objective description of what constitutes a tantrum for this specific client.
ABaseline data cannot be collected reliably until the behavior is defined.
BA preference assessment identifies reinforcers; it does not clarify what a tantrum is.
CIntervening before defining and measuring the behavior skips assessment.
DCorrect: a vague concern must first be turned into an observable, measurable operational definition.
More notes below ↓
CMeasurement, Data Display, and InterpretationC.3
Pick an option: the verdict, the reasoning and a note on all four options open here.
A primary limitation of permanent product recording is that it cannot capture behaviors that do not leave behind physical evidence or products. Many behaviors of interest (e.g., vocalizations, social interactions) do not produce permanent products that can be measured later.
ACorrect: product recording only works when the behavior leaves a lasting outcome, so many behaviors, such as vocal behavior, cannot be measured this way.
BProduct recording is usually efficient because the observer need not be present.
CIt typically needs little observer training.
DIt does not inherently overestimate; it measures the outcome, which may not show who produced it or how.
More notes below ↓
CMeasurement, Data Display, and InterpretationC.5
Pick an option: the verdict, the reasoning and a note on all four options open here.
Partial interval recording tends to overestimate the actual occurrence of behavior because if the behavior occurs at any point during the interval, the entire interval is scored as an occurrence, regardless of how briefly the behavior occurred.
AOne trained observer can use partial interval recording.
BIt can be used with behaviors at various rates.
CIt is a simple and widely used method.
DCorrect: scoring an interval for any occurrence, however brief, tends to overestimate how much the behavior occurs.
More notes below ↓
CMeasurement, Data Display, and InterpretationC.5
Pick an option: the verdict, the reasoning and a note on all four options open here.
A 10-minute observation using 15-second intervals would yield 40 intervals (10 minutes × 60 seconds ÷ 15 seconds = 40 intervals). The maximum number of intervals that could contain the target behavior is 40, which would occur if the behavior was present during every interval.
A15 is the interval length in seconds, not the number of intervals.
B10 is the session length in minutes.
CCorrect: 10 minutes is 600 seconds; 600 ÷ 15 = 40 intervals, so at most 40 can be scored.
D60 would be the number of 10-second intervals in 10 minutes.
More notes below ↓
CMeasurement, Data Display, and InterpretationC.5
Pick an option: the verdict, the reasoning and a note on all four options open here.
When interval recording shows the behavior in 70% of intervals, you can only conclude that it occurred during 70% of the observed intervals. That does not mean it occurred for 70% of the time: interval recording gives an estimate, not a measure of time or count.
AInterval recording does not count instances, so it cannot say what share of instances was seen.
BInterval data estimate time; they do not give the exact proportion of time.
CRate needs a count of responses per unit of time, which interval recording does not provide.
DCorrect: interval data show only the percentage of intervals in which the behavior was scored.
More notes below ↓
CMeasurement, Data Display, and InterpretationC.6
Pick an option: the verdict, the reasoning and a note on all four options open here.
Total count IOA divides the smaller total count by the larger total count and multiplies by 100. It shows how closely the two observers' totals matched, not whether they recorded the same instances; agreements ÷ (agreements + disagreements) × 100 is interval-by-interval or trial-by-trial IOA.
AAveraging the two counts does not express agreement.
BDividing one observer's count by the other's can exceed 100% and is not the formula.
CCorrect: total count IOA divides the smaller total by the larger and multiplies by 100.
DAgreements divided by agreements plus disagreements is interval-by-interval (or trial-by-trial) IOA.
More notes below ↓
CMeasurement, Data Display, and InterpretationC.6
Pick an option: the verdict, the reasoning and a note on all four options open here.
Reactivity occurs when the subject's behavior changes due to awareness of being observed. This threatens validity because the observed behavior no longer represents the typical behavior that would occur under natural conditions.
AIOA supports confidence in measurement; it does not threaten validity.
BStructured forms improve measurement accuracy.
CCorrect: reactivity means the person behaves differently because they are observed, so the data may not represent typical behavior.
DClear definitions protect validity and reliability.
More notes below ↓
CMeasurement, Data Display, and InterpretationC.8
Pick an option: the verdict, the reasoning and a note on all four options open here.
Semi-logarithmic charts (like the Standard Celeration Chart) use a logarithmic scale on one axis (usually the y-axis), which makes proportional changes appear equal regardless of absolute value. This allows for meaningful comparison of rate changes even when baseline rates differ substantially.
AGraph type does not change data collection.
BCorrect: on a semi-logarithmic chart equal proportional changes look equal, which makes rate changes comparable across levels.
CSemi-log charts are often harder for non-specialists to read.
DPhase change lines are still needed.
More notes below ↓
CMeasurement, Data Display, and InterpretationC.9
Pick an option: the verdict, the reasoning and a note on all four options open here.
Trend refers to the overall direction of the data path over time - whether the behavior is increasing, decreasing, or remaining stable. Identifying trends helps behavior analysts determine if interventions are having the desired effect on behavior change.
ACorrect: trend is the overall direction of the data path: increasing, decreasing or flat.
BA total count is a summary, not a trend.
CAverage value is level, not trend.
DThe spread of data points describes variability, not trend.
More notes below ↓
CMeasurement, Data Display, and InterpretationC.10
Pick an option: the verdict, the reasoning and a note on all four options open here.
This goal specifies a mastery criterion by identifying both the level of performance required (85% accuracy) and the consistency or duration of that performance (for 4 consecutive weeks) needed to demonstrate mastery of the skill.
ACorrect: a performance level held over a period (85% for 4 weeks) defines mastery.
BGeneralization criteria test performance in new settings, people or materials.
CSocial validity concerns stakeholder judgments of goals, procedures and outcomes.
DMaintenance criteria test whether performance lasts after teaching stops.
Laid out the way a failing BACB score report lists your percentage per content area. A practice percentage is not a pass prediction: the BACB publishes no passing percentage.
When the record opens, note the task codes behind your misses and reread only those phases. Then take the C chip on the 60-trial practice exam, which narrows it to its 8 Section C trials, before the full paced run.
09Onward
Where Section C leads next
The graphs you read here are the evidence behind Section D, Experimental Design, and the data-based decisions of Section H. The other section guide covers Section G, Behavior-Change Procedures, the heaviest at 22 items. For booking and pass/fail scoring, read how the BCaBA exam works; for the credential as a whole, start at the home page.
01Does partial-interval recording overestimate or underestimate behavior?
It overestimates: any moment of behavior scores the full interval. In the worked minute above, the same behavior reads 83% partial and 17% whole against 43% actual.
02How do you calculate IOA?
Match the formula to the data. Total count: smaller over larger, times 100, so 24 and 30 give 80%. Interval-by-interval: agreements over agreements plus disagreements, so 17 of 20 gives 85%. The IOA table lists nine types.
03What is a cumulative record?
A graph where each response adds to a running total, so the line only rises or stays flat. Its slope is the rate: steep means fast responding, flat means none.
04What is the difference between latency and IRT?
Latency runs from a stimulus to the start of the response; IRT runs from the end of one response to the start of the next. An instruction at 0:00 and writing at 0:06 is a 6-second latency.
05What is momentary time sampling?
A discontinuous method that scores the behavior only if it is happening at the instant each interval ends. It frees the observer between moments but misses brief behavior, and its error runs either way.
06How many Section C questions are on the BCaBA exam?
20 of the 150 scored items, 13%, under the 6th-edition outline per the current BCaBA Handbook (updated June 2026). Only Section G, with 22, carries more.