Reading graphs: trend, level and variability

Reading a behavior graph comes down to three features: trend (which way the data are heading), level (where the data sit on the y-axis) and variability (how much the points bounce around). Describe all three within each phase, then compare phases on either side of a phase change line.

What does task A.7 ask you to do with graphs?

Task A.7 asks you to describe behavior and environment using graphed data, which in practice means recognising trend, level and variability and noticing changes across phases. You are not deciding whether a treatment works; you are seeing what the graph shows and reporting meaningful patterns, especially unwanted ones, to your supervisor promptly.

How do trend, level and variability differ?

Trend is the overall direction of the data path: increasing, decreasing or flat. Level is the general height of the data, often judged by where most points cluster or by the mean. Variability is how far points spread from each other. A phase can have a flat trend at a high level with low variability, so judge each feature separately.

FeatureQuestion it answersExample dataHow to describe it
TrendWhich way is it going?14, 11, 9, 6, 4Decreasing trend
LevelWhere does it sit?19, 21, 20, 20, then 5, 6, 4Stable at about 20, then an immediate drop in level
VariabilityHow consistent is it?2, 15, 4, 18, 1, 16High variability, no clear trend

Reading a graph, step by step

School example: a graph of a student's call-outs shows 12, 13, 11 and 12 in baseline, then 9, 7, 6 and 4 after a hand-raising program starts. You would describe a stable baseline with low variability, then a decreasing trend and lower level in intervention, which is the direction the plan wants.

  • Read the axis labels and condition labels so you know what is measured and in which phase.
  • Check whether a higher value is good or bad for this target.
  • Within each phase, describe the trend, then the level, then the variability.
  • At each phase change line, compare the last few points before it with the first few after it.
  • Ask whether the direction matches the plan's goal, for example decreasing hitting or increasing independent steps.
  • Report patterns that run the wrong way or stall to your supervisor.

What does a stable baseline look like and why does it matter?

A stable baseline has little variability and no strong trend, such as 8, 9, 8, 9, 8. It gives a clear starting point, so any change after intervention is easier to attribute to the intervention. A baseline that is already trending in the hoped-for direction, or bouncing widely, makes the effect of a new procedure hard to see.

Worked scenario: in baseline, off-task data hover between 19 and 21. Right after a token system begins they sit between 4 and 6. That is a large, immediate change in level with low variability in both phases, which is a clear picture.

Second scenario: as prompts are faded, a learner's independent toothbrushing steps drop for three sessions. The trend runs opposite to the goal. A technician should report it to the supervisor, who may adjust the fading steps, rather than pausing the fading or re-adding prompts on their own.

Which graph-reading traps show up on the exam?

Watch for questions that mix up trend and level, or that describe highly variable data as increasing because the last point is high. Another trap is ignoring direction: a decreasing trend is good for self-injury but bad for correct responding. Also remember that a vertical dashed line marks a phase change, not a missed session or a reporting deadline.

What should the RBT report and what does the BCBA decide?

The technician notices and reports: a rising problem behavior, a skill that has stalled, sudden variability, or data that do not match what was seen in sessions. The supervisor interprets the graph, decides whether to change phases, adjust the program or collect more data, and explains the decision. Reporting early is part of the RBT's job; changing the plan is not.

Practice questions

Question 1 · Data collection and graphing

During intervention, the graph of Andre's daily self-injury shows 14, 11, 9, 6 and 4 instances. How is this pattern best described?

Question 2 · Data collection and graphing

In baseline, Nina's off-task data hover between 19 and 21 across sessions. Right after the phase line, intervention data hover between 4 and 6, with no upward or downward slope. Which describes the change?

Question 3 · Data collection and graphing

A graph of Tyler's correct responses shows 2, 15, 4, 18, 1 and 16 across six sessions. What stands out most about these data?

Question 4 · Data collection and graphing

Which set of baseline data points shows the most stable pattern?

Question 5 · Data collection and graphing

After prompts began to be faded, Hana's number of independent toothbrushing steps across five sessions was 8, 7, 7, 5 and 4. Which description is accurate?

Questions people ask

Can there be a change in level without a change in trend?

Yes. Data can jump to a new level and stay flat there, so the trend is flat in both phases while the level changed.

Is high variability always a problem?

Not always, but it often points to an uncontrolled variable, such as sleep or staff changes, that your supervisor will want to hear about.

How many points show a trend?

There is no fixed number; generally three or more points heading the same way suggest a trend worth noting.

How ready are you?

Ten questions, about 3 minutes, no account. Get your readiness score and your weakest domains.

Check my readiness