Behavior analysts make most of their decisions by looking at graphs. A well-built graph shows what behavior is doing at a glance. A poorly built one can hide a real change or invent one that never happened.
Why it matters
Visual analysis is the main way behavior analysts judge whether an intervention is working. That only works if the graph is honest: the right display for the question, a sensible scale, and clear marks for when conditions changed. As a BCBA, you will design graphs, review ones that others produce, and present data to families and funders.
Practice items on this task ask you to choose a display for a purpose, spot a construction error, or read a cumulative record correctly.
Key ideas
- Line graphs are the default. An equal-interval line graph plots a measure of behavior across time. It shows level, trend, and variability, which is why it is the standard display for ongoing clinical data and single-case designs.
- Time usually goes on the horizontal axis. The x-axis shows sessions, days, or another unit of time. The y-axis shows the measure of behavior, such as rate, duration, or percentage correct.
- Condition changes get a vertical line. A condition change line marks where the independent variable changed, such as baseline to intervention. Label each condition above the data, and do not connect the data path across the line. Also break the path across a long stretch when behavior was not measured, such as a school break.
- Bar graphs summarize. A bar graph compares discrete sets of data, such as the average for each condition or each class period. Because bars collapse many sessions into one value, they hide session-to-session variability and trend. Use them as a summary, not as your only decision-making display.
- Cumulative records show a running total. Each response is added to everything that came before, so the line can only rise or stay flat. The slope shows the rate of responding: steeper means faster, flat means no responses.
- Scale shapes the story. A y-axis that starts well above zero, or tiers drawn on different scales, can make a small change look large or a large change look small.
Definitions
- Line graph: a display in which data points for successive observations are plotted against time and joined by a line.
- Data path: the line connecting successive data points within one condition.
- Condition change line: a vertical line showing when the independent variable changed. Some texts call it a phase change line.
- Bar graph: a display that uses the height of separate bars to compare summary values across discrete categories or conditions.
- Cumulative record: a graph in which each data point is the total number of responses so far, so its slope reflects rate.
- Equal-interval scale: an axis on which equal distances stand for equal amounts, such as each gridline adding 5 responses.
Examples from a session
- Line graph at a day program. Amara, a BCBA, graphs the rate of hand-biting for Jordan, age 16, at a day program. Baseline runs for 8 sessions, a condition change line follows, and then the intervention phase begins. The baseline and intervention data paths are drawn separately, so the drop in level at the change is easy to see.
- Bar graph for a school team. Mr. Ruiz, a BCBA consultant, wants to show a middle school team which class period has the most outbursts. He keeps the daily line graph for decisions but adds a bar graph with one bar per period showing the monthly average. Period 3 stands out right away.
- Cumulative record at home. Lena, age 7, practices reading sight words with her grandmother. Each word she reads correctly is added to a running total, plotted by calendar day. In week 1 the line climbs steeply. During a family trip with no practice sessions, the line runs flat. When practice resumes, it climbs again more gently, showing that she is still reading words, just more slowly than before.
- Adult services. A job coach tracks the percentage of packing steps Dev, age 32, completes independently. A line graph across work shifts shows a steady upward trend, which a single average bar would have hidden.
Common mistakes / exam traps
- Connecting data across a condition change line. That suggests one continuous condition and blurs the effect of the change.
- Reading a rising cumulative record as "behavior is increasing." The line rises whenever any responding occurs. Only a steeper slope means the rate went up. A gentler slope means the rate went down, even though the total keeps growing.
- Reading a flat segment as steady responding. On a cumulative record, flat means zero responses during that period.
- Using a bar graph to judge progress. Averages can hide a downward trend at the end of a phase or wide day-to-day swings.
- Truncating or mismatching scales. Starting the y-axis near the data, or giving tiers different ranges without flagging it, distorts how big the change looks.
Exam trap
Line graph vs. bar graph: check what is on the x-axis. If it is time (sessions, days), use a line and keep it for decisions. If it is separate categories (class periods, settings, staff), use bars; a line joining categories implies a change from one to the next that never happened.
Exam trap
On a cumulative record, ask "how steep?" rather than "how high?" Height is the total so far. Slope is the rate.
Quick check
A BCBA shows caregivers a graph of percentage of correct responses that ranged from 62% to 68%. The y-axis runs from 60% to 70%, and the caregivers believe their child made a huge leap. What is the main problem with this graph?