How Should Generalization Data Be Compared Across People and Places?

Compare ABA generalization data

Build a table with person, place, date, activity, opportunities, independent responses, prompted responses, ordinary supports, partner response, observer, and client feedback. Mark which conditions were taught and which were probes. Keep a condition separate when definitions or exposure differ materially.

The table should let a reader see whether the person, the setting, or both changed. “School” can contain several classrooms, partners, and routines. “Caregiver” can refer to people with different familiarity and response patterns. Use the level of detail needed for the decision.

Match the comparison fields

Before comparing rates, check:

• the same response definition and independence rule
• the same type of eligible opportunity
• the same ordinary access supports
• comparable prompt and response windows
• partner preparation and expected response
• observation coverage and data reliability
• plan version and teaching history
• client choice, health, and relevant context

If the conditions differ, preserve the result while qualifying the comparison. A difference can still guide support without proving that one person or place caused it.

Define the conclusion carefully

A combined rate can answer an overall access question when its components are compatible. It can also hide a setting with no transfer. Preserve each numerator and denominator, then state why an aggregate is useful. Weighting every setting equally may also mislead when opportunity counts differ.

Avoid two common pooling errors

First, averaging percentages gives each setting equal weight regardless of opportunity count. Home at 4 of 5 is 80%, and community at 1 of 1 is 100%. Their simple average is 90%, while pooled performance is 5 of 6, or 83.3%. Both calculations answer different questions.

Second, pooling raw counts can hide an important setting. A high-volume home result may dominate a low-volume school result. Show each setting first. Use an aggregate only when the combined denominator and intended question are explicit.

Compare client and partner behavior separately

The client may respond while an unfamiliar partner fails to notice or honor the message. Report client response and partner response with separate denominators. Partner training or access repair may be the next step, rather than more client teaching.

Likewise, missing AAC or materials belongs in an operational readiness measure. Do not interpret a setup failure as evidence that the skill disappeared.

Use current clinical and measurement sources

The CASP public summary places assessment, planning, implementation, and evaluation within its autism-treatment scope. The BACB Ethics Code addresses client involvement, consent and assent when applicable, assessment-based intervention, risk, documentation, and evaluation for covered behavior analysts.

The BCBA Test Content Outline covers programming for generalization and maintenance as examination content. It does not create a universal probe count, maintenance interval, independence definition, or authority to practice.

Use the historical framework with current judgment

Stokes and Baer organized generalization-programming tactics in a 1977 conceptual paper. It remains historically influential, while current clinical use still requires individualized goals, evidence, supports, client choice, and review.

Keep ordinary access supports available

The ASHA AAC portal supports continuous communication-tool access. AAC, visual schedules, mobility supports, and other ordinary access tools can remain part of independent daily participation when they fit the person and goal.

The same tool should not be labeled independent in one setting and prompted in another without a documented reason. Define the support's role once per plan version and show any approved setting-specific variation.

A practical example

Priya requests help in 4 of 5 home opportunities, 2 of 5 school opportunities, and 1 of 2 community opportunities. The pooled result is 7 of 12, or 58.3%, while the setting rates remain 80%, 40%, and 50%.

At school, Priya's AAC is available in all five opportunities. She sends two messages, while the partner responds to only one. The team reports client requests as 2 of 5 and school partner response as 1 of 2 messages. It does not describe the remaining three opportunities as partner failures because no message occurred.

The school partner practices waiting and acknowledging the AAC message. The next sample uses the same definition and five opportunities. Priya requests in four and the partner responds to all four. The team preserves both phases and states that partner coaching, time, and other conditions changed together.

Choose an action for each condition

Assign a specific owner, repair, and next measure. One setting may need materials, another partner preparation, and another no change because the client does not value expansion there. A single universal homework assignment can miss every actual barrier.

Check measurement quality across conditions

Different observers may apply the response or opportunity definition differently. Ask who collected data, how observers were trained, whether matched observations were checked, and whether camera or visibility differed. A setting difference can reflect measurement rather than client performance.

Treatment fidelity also matters. If staff use the correct wait time in one place and prompt immediately in another, compare implementation before interpreting transfer. Report fidelity and generalization outcomes as separate measures.

Display the data so families can see the decision

A useful chart or table keeps people and places in separate rows with raw counts. Add columns for ordinary supports, partner response, teaching history, and client feedback. Mark phase changes when a definition, plan, or support changes.

Avoid a graph that shows only percentages when denominators vary. Two of two and eight of ten are both 100% and 80%, while the opportunity counts affect confidence and practical interpretation. Show both.

Keep client experience beside performance

A person may perform well in a setting they dislike or choose fewer opportunities in a setting they value for another reason. Ask about privacy, comfort, burden, sensory conditions, travel, and social context. Performance should not override a clear choice to stop expansion.

The person may also prefer different response forms with different partners. Speech at home and AAC at school can both meet a communication outcome when the plan defines them. Comparing only one form would create an artificial setting gap.

Review missing settings

If a planned condition has no data, list it as pending, declined, unavailable, or outside scope with the reason. Do not let it disappear from a completed-condition denominator. Assign a next step only when the client still wants the setting and the probe is feasible.

Preserve the next review date and owner.

Questions families can use

Ask whether definitions and supports matched, how many opportunities occurred in each condition, which were independent, whether partner response differed, what the client reported, and whether pooling hides a useful pattern.