Identifying Trends in Graphed Data

Summary: A line graph tells you three things: which way the data is heading (trend), how high or low it’s sitting (level), and how much it’s bouncing around (variability). Your job as an RBT is to look at a graph and describe what you see in plain terms, “it’s trending up,” “it’s flat and noisy”, not to decide what to do about it. That decision belongs to the BCBA. This page teaches you to eyeball a graph honestly, recognize when a trend is therapeutic, and know exactly where your lane ends.

You’re going to spend a lot of your career making graphs without ever really reading them. You tally, you log, the software draws the line, and you move on to the next program. That’s fine, until your supervisor points at a graph during a meeting and asks, “what do you see here?” The RBT who can answer that cleanly looks like they know what they’re doing, because they do. And the third-edition Test Content Outline now expects you to. Describing what a graph shows is a task on the list, so it’s fair game for the exam.

The good news: you don’t need statistics. You need to look at a line and say three honest things about it.

The three things every graph tells you

Almost every graph you’ll meet in this field is a line graph. Behavior goes up the side (the vertical y-axis), time goes across the bottom (the horizontal x-axis), usually one dot per session or per day. Connect the dots and you’ve got a line that moves. Reading it means answering three questions.

Trend: which way is it going?

Trend is the overall direction of the data over time. Picture the general slope of the dots if you squinted and drew one straight line through them. There are really only a few answers.

• Increasing (ascending): the dots climb left to right. Behavior is going up.
• Decreasing (descending): the dots fall left to right. Behavior is going down.
• Flat (zero-celerating): the dots hold roughly level. No real direction. “Zero-celerating” is just the technical way of saying the line isn’t accelerating up or down.

That’s it for direction. The word “celerating” trips people up, so park it: a zero-celerating trend is a flat one. An increasing trend is accelerating, a decreasing trend is decelerating, and a flat trend is doing neither.

The thing to hold onto is that trend is about the whole stretch, not one dot. A single high day in the middle of a downward slide doesn’t reverse the trend. Step back, look at the path the data is walking, and call the direction.

Level: how high or low is it sitting?

Level is where the data lives on the y-axis, the typical value the line hovers around. You can have two graphs with the exact same trend but very different levels. One learner is tantruming around 10 times a session and trending down. Another is tantruming around 2 times a session and trending down. Same direction, very different level, and the kid sitting at 2 is in a different place than the kid sitting at 10 even though both lines slope the same way.

When you describe a graph, level is the “how much” and trend is the “which way.” You usually want both. “It’s trending down and now sitting around two per session” says far more than “it’s trending down.”

Variability: how much is it bouncing?

Variability is how spread out the dots are, how much the line jumps from session to session. Low variability looks like a tidy line that moves smoothly. High variability looks jagged, big swings up and down even if the overall trend is clear.

Variability matters because a noisy graph is hard to trust. If a learner scores 2, 9, 3, 8, 1, 10 across six sessions, the average might look fine but something is going on. Maybe a different staff member runs Tuesdays. Maybe the behavior depends on whether the kid slept. A clean, low-variability line lets everyone believe the trend. A jagged one tells the team to dig before they conclude anything.

On the exam: Know all three terms cold, because a question will hand you a described or pictured graph and ask which one it’s testing. Trend is direction, level is the typical value on the y-axis, variability is the bounce. The classic trap is calling something a “trend” when the data is really just variable, six dots sawing up and down with no net direction is high variability, not a trend.

How to eyeball a graph fast

You won’t compute anything in a session, and you won’t on the exam either. Here’s the look-don’t-calculate routine.

1. Read the axes first. What’s on the bottom (usually sessions or dates) and what’s on the side (the behavior and its unit, like “tantrums per session” or “percent correct”). Skip this and you can misread everything that follows.
2. Find the phase lines. A vertical line on the graph marks where something changed, baseline ending and intervention starting, or one intervention swapping for another. You read each phase on its own, then compare across the line. More on this below.
3. Squint for direction. Within a phase, is the cloud of dots heading up, down, or holding flat? That’s your trend.
4. Note the level. Roughly what value are the dots sitting around in this phase versus the last one?
5. Check the bounce. Smooth or jagged? High variability means hold your conclusions loosely.

Do those five and you can describe any graph your supervisor puts in front of you.

Phase lines are the whole point of comparing

That vertical phase-change line is where the story lives. To the left is what behavior looked like before the BCBA changed something (often baseline, the “before” picture). To the right is what it looked like after. You compare the two sides on trend, level, and variability, and that comparison is how the team sees whether the change did anything.

A baseline that’s flat or trending the wrong way, then a clear shift after the line, is the cleanest evidence an intervention is working. Don’t try to read across the phase line as one continuous trend. Read each side, then compare.

What a therapeutic trend actually looks like

Here’s where “up” and “down” stop being enough. Whether a trend is good news depends entirely on what the program is trying to do, and that splits into two camps.

Acquisition targets: up is good

Acquisition programs teach a new skill, tacting, manding, matching, following instructions, washing hands. You want more of the behavior, so an increasing trend is therapeutic. The line should climb toward mastery: percent correct rising session over session, more independent responses, fewer prompts needed over time. A flat line on an acquisition program means the learner is stuck, and a downward line means the skill is slipping. Neither is what you want.

So on a skill-acquisition graph, up is the win. You’re rooting for the line to go up and then hold high once it gets there.

Reduction targets: down is good

Behavior-reduction programs aim to decrease something, aggression, elopement, self-injury, calling out. You want less of it, so a decreasing trend is therapeutic. The line should fall: fewer instances, lower rate, shorter durations as the intervention takes hold. A flat or increasing line on a reduction program is the worrying one.

The mirror image of this is worth saying out loud, because programs often run in pairs. When you’re reducing a problem behavior, there’s usually a replacement behavior being taught at the same time. So you might watch two graphs side by side: aggression trending down (good) while mands trending up (also good). Seeing both move the right way is stronger evidence than either alone, because it shows the learner swapping the problem behavior for an appropriate one rather than the problem just vanishing.

Common mistake: Reading a downward trend as automatically “bad” or an upward trend as automatically “good.” Direction means nothing until you know the goal. A line crashing toward zero is fantastic news on a self-injury program and a disaster on a hand-washing program. Always ask “what is this program trying to do?” before you judge the trend.

A scenario to make it concrete

You’re running two programs with Marcus, a 7-year-old.

The first is a tacting program, naming animals from flashcards. His graph shows percent correct: week one he’s around 40%, then 55, then 60, then 72, then 80. The dots are a little jagged but the climb is unmistakable. You’d describe it as an increasing trend with low-to-moderate variability, level rising from the 40s into the 80s. Therapeutic, because this is acquisition and up is the goal.

The second is a reduction program for grabbing items off the table instead of asking. That graph shows grabs per session: 12, 11, 13, 9, 6, 5. Decreasing trend, dropping from a level around 12 down toward 5, fairly clean. Therapeutic, because this is reduction and down is the goal. Meanwhile his “asks for the item” graph is climbing from 2 to 9 over the same weeks. The replacement is taking over.

Now suppose week six on the grabbing graph spikes back to 11. One dot. You don’t panic and you don’t announce the program failed. You report it: “grabbing was trending down nicely but jumped back up this session.” That’s the honest read. What happens next isn’t your call, and that’s the next thing to get straight.

Your lane: report what you see, don’t decide what to do

This is the part the exam cares about most, and it’s the part new RBTs blur. You observe and describe. The BCBA analyzes and decides. Those are two different jobs, and the line between them is bright.

You read a graph and say what’s there: “trending up,” “flat for two weeks,” “really variable since we switched rooms,” “she hit mastery criterion three sessions in a row.” That’s accurate, useful, and squarely within your scope. You bring it to your supervisor, especially when something looks off, a behavior climbing on a reduction program, a skill stalling, a sudden burst of variability.

What you don’t do is change the program off your own read. You don’t decide the intervention isn’t working and quietly stop running it. You don’t add a prompt, drop a step, raise the mastery bar, or switch reinforcers because the line looked discouraging. Program changes are clinical decisions, and clinical decisions belong to the BCBA, who’s weighing things you may not see: the assessment, the history, other settings, the family’s goals, the research behind the procedure.

Why the hard line? Because data-based decision-making is only as good as the person trained to make it. The BCBA reads your graph alongside everything else and decides whether to continue, modify, or end a program. If RBTs started making those calls from the dots alone, you’d get changes based on one bad day, or a program dropped right before it would’ve turned the corner. Your accurate reporting is what makes the BCBA’s decision possible. That’s not a small role. It’s the foundation the whole decision sits on.

On the exam: When a scenario shows you a graph and asks what the RBT should do, the answer is almost always “collect and report the data to the supervisor,” not “change the intervention.” Any option where the RBT modifies a program, ends it, or makes a clinical judgment is the trap. Describe and report stays in scope; decide and change does not.

Reading trend honestly when the line is messy

Real graphs rarely cooperate. You’ll get a few patterns that are easy to misread, so know them.

• A trend hiding under variability. The dots bounce a lot, but if you squint, the bounces are happening at steadily lower values. That’s a decreasing trend with high variability, and it’s still a real trend. Don’t let the noise convince you nothing’s happening.
• No trend at all. Six sessions sawing 4, 8, 3, 9, 4, 8 with no net movement isn’t a trend in any direction. It’s a flat, highly variable line. The honest description is “no clear trend, high variability,” and that itself is information the BCBA wants.
• A short run. Two or three dots don’t make a trend. You need enough data points for a direction to mean anything. If the phase just started, the honest answer is “too early to call.”
• One outlier. A single dot way off the line doesn’t flip the trend. Note it, mention it, but don’t let it rewrite the direction the rest of the data is clearly walking.

When you’re not sure, say you’re not sure. “It might be trending down but it’s only three sessions and they’re all over the place” is a better thing to tell your BCBA than a confident call you can’t back up.

What to walk away with

• Every line graph answers three questions: trend (direction), level (typical value), variability (bounce).
• Trend can be increasing, decreasing, or flat (zero-celerating). It’s about the whole stretch, not one dot.
• Read the axes and find the phase lines first; compare each phase to the one before it.
• A therapeutic trend depends on the goal: up is good for acquisition, down is good for reduction.
• High variability means hold your conclusions loosely and tell the BCBA something’s noisy.
• Your job is to describe what the graph shows and report it. The BCBA makes the data-based decision to continue, change, or end a program.
• When a scenario asks what the RBT should do about a graph, “report it to the supervisor” beats “change the program” almost every time.