Visual analysis

Visual analysis begins before pattern reading

Confirm what each point measures. Read the operational definition, unit, eligible opportunity, observation duration, and recording method. Verify dates, missing-data rules, axis scales, condition labels, phase-change timing, and data-entry accuracy.

A graph can look orderly while combining unequal session lengths, changed definitions, missing communication access, or values entered under the wrong condition. Resolve those issues or mark their limits before interpreting the pattern. Preserve the raw observations and audit trail.

The current BACB BCBA Test Content Outline includes choosing displays, interpreting graphed data, single-case design features, threats to validity, procedural integrity, and data-based decisions. It identifies examination content and does not supply one mandatory visual-analysis protocol.

Review each condition from left to right

Within each condition, examine:

• Level: where the observations generally sit vertically.
• Trend: the direction and rate of systematic change over time.
• Variability: the spread and sequence of values around the level or trend.

Describe the raw pattern before assigning meaning. A mean or trend line can help when its method is stated, yet the summary may hide cycles, outliers, or a changing slope. Note floor and ceiling effects, gaps, small samples, and contextual events.

The analyst may then project the current pattern forward as a tentative prediction. In a single-case design, a condition change tests whether subsequent observations differ in the predicted way. That logic depends on a design capable of verification and replication.

Compare adjacent and repeated conditions

Across a condition boundary, examine:

• Immediacy: how quickly level, trend, or variability changes near the boundary.
• Overlap: how many observations from one condition occupy the range or distribution of another.
• Consistency: whether similar conditions produce similar patterns at different times or across tiers.

The What Works Clearinghouse single-case technical documentation uses these six dimensions and looks for repeated demonstrations of an effect at different points in time when judging design evidence. One AB comparison can describe change. Replicated, design-controlled comparisons are needed for stronger causal inference.

Expected latency and carryover affect the boundary. Some outcomes may change at the next opportunity; others need repeated practice or time. State the expected timing and its basis before viewing the result when possible.

A fictional repeated-condition example

A fictional graph reports independent help messages during ten eligible opportunities per session. AAC access and the partner-response procedure are available in every scored session.

• Baseline A1: 3, 4, 3, 4, 3, mean 3.4.
• Support B1: 7, 8, 7, 9, 8, mean 7.8.
• Baseline A2: 4, 4, 5, 4, 5, mean 4.4.
• Support B2: 8, 9, 8, 9, 9, mean 8.6.

Both B phases have higher levels than their neighboring A phases, little range overlap, and changes beginning near each boundary. B1 and B2 show similar patterns. The two A phases differ somewhat, which the reviewer should report rather than erase.

This repeated pattern strengthens a hypothesis that the arranged condition influenced the measured response under these circumstances. The design details, assignment of phase changes, integrity, measurement reliability, concurrent events, and carryover still matter. The graph does not establish generalization, social importance, comfort, or a durable effect after support changes.

A protocol makes judgment inspectable

The authors of Systematic Protocols for Visual Analysis identified variation in the content and sequencing of published protocols. A single-case design analysis review likewise separates within-phase, between-phase, and design-level reasoning.

A review of visual-analysis training and reliability describes disagreement as a recognized challenge and evaluates structured training approaches. Agreement among reviewers does not prove the conclusion is correct, while unexplained disagreement weakens transparency. Record each independent judgment, the evidence used, and how differences were resolved.

Connect the pattern to a decision carefully

Ask whether the outcome reflects the person's priorities, communication, comfort, participation, health, or safety. Include assent and dissent when applicable, direct client report, family input, feasibility, burden, adverse effects, and useful supports. A visually clear shift can have little practical value; a modest change can matter greatly to the person.

Software may calculate summaries, flag boundaries, and display source data. An appropriately qualified professional selects the analysis, interprets the evidence, evaluates clinical meaning and risk, and records the decision within scope.

A repeatable review record

1. Lock the question, measure, graph, and eligible data.
2. Review level, trend, and variability within every condition.
3. Review immediacy, overlap, and consistency across conditions.
4. Evaluate design replication, latency, carryover, and validity threats.
5. Check measurement quality, procedural integrity, and concurrent changes.
6. Include client report, access, clinical importance, burden, and risk.
7. Record the conclusion, uncertainty, decision, owner, and recheck date.