Data-based decision making: the impact of data variability, training, and context.

The team asked three groups to judge how long a baseline should run. Experts, certified BCBAs, and novices looked at the same graphs. Some graphs had jumpy data. Some had labels that said what the axes meant.

Each person picked the point where they would stop baseline and start treatment. The researchers counted how often people in each group agreed with each other.

Experts and BCBAs picked similar baseline lengths. Novices picked very different lengths from each other. When the data line bounced around, agreement dropped for everyone.

Giving the axes simple labels like “time-out minutes” made people choose shorter baselines. Jumpy data made novices exit even sooner.

Rader et al. (2021) later showed that even doctoral-level BCBAs misread functional-analysis graphs about one third of the time. Together the two papers say: letters after your name do not protect you from visual errors.

Falligant et al. (2020) added math aids called dual-criteria rules. Their work shows that false alarms go up when baseline data swing wide, matching the 2011 warning about jumpy lines.

Carey et al. (2014) found another trap: sampling only the first few trials each session hides true mastery. All four studies tell the same story — our eyes need help.

You probably decide when to start treatment every week. If you train new staff, show them calm and jumpy examples side by side. Give clear axis labels, then teach them to wait through the bounce. One extra week of data now can save months of guessing later.