Phase-change line

The line separates conditions

Examples of planned conditions include baseline, teaching, generalization, maintenance, and two named versions of a procedure. The phase line helps readers see which observations occurred under each arrangement.

The condition label should be concise and traceable. “Visual schedule v2” is more useful than “Treatment” because it points to a specific record. The full protocol still carries components, parameters, roles, training, consent, and safeguards.

Place the boundary between points

If sessions 1 through 8 used the first condition and session 9 used the next condition, place the line between sessions 8 and 9. Running it through session 9 creates uncertainty about which condition governed that point.

Use the actual first implementation, not merely the date a plan was signed, approved, trained, or entered into software. Those dates can differ. Preserve each in the source record and use the one relevant to the graph.

When a condition changes midway through an observation, apply the program’s declared rule. The record may split the observation, assign it based on a defined exposure threshold, or mark it as a transition. Explain the choice.

Training and implementation can also begin gradually. When some partners use the new procedure while others continue the prior one, a single clean boundary may overstate the change. Mark a transition period, separate series by implementer, or delay the new phase until the stated implementation criterion is met. The choice should follow the analysis plan and actual records.

Retrospective graphs require the same discipline. Reconstruct the boundary from dated service notes, implementation logs, and protocol history. When the exact session remains uncertain, show the uncertainty instead of choosing the most favorable boundary.

Missing data require a different mark

A gap between observations does not mean a treatment phase changed. Leave missing data visibly missing and annotate the reason when appropriate. Plotting zero claims an observation with a measured value of zero.

Holidays, school closures, medication changes, staffing events, and illnesses can matter to interpretation. Use an event marker or note rather than presenting each as a designed condition. If an event leads to a planned procedural change, record both facts separately.

A fictional schedule-version example

Evan’s morning-routine graph shows independent steps across 14 school days. Sessions 1 through 8 use “Picture schedule v1.” Version 2 begins before session 9 with larger symbols and an added choice step.

The phase-change line falls between sessions 8 and 9. The label over the new segment reads “Picture schedule v2.” Training occurred two days earlier, but the graph uses the first day the new procedure operated.

Session 6 is absent because school was closed. The graph shows a gap at session 6 rather than another phase line. The team retains the closure in the event log and excludes the day from the session-based measure under its predeclared rule.

Bundled changes limit attribution

If version 2 begins alongside a new implementer, a different time of day, and a new reinforcement arrangement, one phase line marks a bundled condition change. The graph may show a clear effect while leaving the active component uncertain.

Name the package accurately and disclose concurrent changes. A component analysis or another design may help when clinically warranted, safe, consented to, and feasible. Graph labels alone cannot isolate cause.

Reversals need ethical and clinical judgment

Some single-case designs reintroduce an earlier condition to test replication. Reversal may be unsuitable when withdrawing an effective support risks harm, causes meaningful loss, conflicts with assent, or allows irreversible learning to carry forward.

The design decision belongs to an appropriately qualified clinician working within law, ethics, consent, payer, and organizational requirements. A graphing convention never authorizes a condition change.

Audit phase lines mechanically

For every boundary, verify:

• last observation under the prior condition
• first observation under the new condition
• actual effective date and protocol version
• label wording and component record
• mid-observation assignment rule
• missing-data and event-marker distinction
• graph legend and accessible text alternative

Phase-boundary accuracy can be calculated as boundaries matching every required source field divided by boundaries due for review. Keep mismatches visible until the graph and underlying record agree.

The BACB BCBA Test Content Outline, sixth edition includes graphing, visual analysis, experimental design, and data-based decisions. A precision-teaching synthesis describes within- and between-condition analysis. Both support careful condition identification; neither prescribes one universal line style.