Multiple Baseline Undermined by a District-Wide Rollout
A BCBA in a school district uses a multiple baseline across three classrooms to evaluate a teacher-delivered token system for out-of-seat behavior (responses per 20 minutes). Baseline lengths were set in advance at 5, 8, and 11 sessions. Classroom 1's intervention begins at session 6 and out-of-seat drops from 8.4 to 2.1. In classroom 2, out-of-seat falls from 9.0 to 3.5 during sessions 6–8, before its intervention is scheduled to start at session 9. In session 7 the district launches a school-wide PBIS kickoff with daily announcements and class-level rewards. Classroom 3's baseline holds steady at 7–9 throughout.
1. What is the threat to internal validity, and how does it affect the tiers differently?
Read the worked answer
Multiple-baseline logic requires behavior to change when-and only when-the independent variable is applied to that tier. Classroom 2 changed three sessions before its intervention began, which violates that logic. Two accounts compete:
- History/confound: the session-7 PBIS kickoff (announcements, class-level rewards) is a plausible alternative cause and coincides with classroom 2's improvement. This is the more parsimonious explanation.
- Indirect generalization or reactivity: classroom 1's token system bled into classroom 2 through shared students, teacher conversation, or teachers' awareness of being observed.
Either way, classroom 2 no longer demonstrates experimental control, and because the PBIS event is district-wide, it threatens all three tiers, not just the one where the change happened to appear first.
2. The team wants to claim the token system worked because classroom 1 improved. What can you actually conclude at this point?
Read the worked answer
Classroom 1's change began at session 6, before the session-7 kickoff, so its effect is temporally cleaner - but one tier is one instance of behavior change, not a demonstration of a functional relation. Classroom 3 is now the critical test and it is compromised: its baseline is stable, but PBIS has been in place since session 7.
- If classroom 3 stays at 7–9 through session 11 and then drops when the tokens begin, that argues against PBIS as the sole cause and strengthens the case.
- If classroom 3 drops without intervention, PBIS or another shared variable is likely responsible.
On present data the honest conclusion is: possible effect in classroom 1, functional relation undetermined. Writing 'the token system reduced out-of-seat behavior' overstates the design.
3. What would you do next to salvage a defensible evaluation?
Read the worked answer
- Extend classroom 3's baseline. Do not intervene early to keep the schedule; the staggered length is the design.
- Graph the confound. Add a phase-change line at session 7 for the PBIS kickoff on all three graphs and describe it in the report.
- Consider a withdrawal in classroom 1 (return to baseline conditions) to see whether out-of-seat recovers - only if the teachers will accept it and the behavior is not dangerous.
- Replicate in a fourth classroom whose baseline begins after PBIS has stabilized, or run an alternating-treatments comparison within classroom 1 to isolate the token system from PBIS components.
- Document that the original design was disrupted by an administrative decision, rather than presenting the data as if the design had been carried out as planned.