General case analysis

Start with the instructional universe

The instructional universe is the full range of situations in which the target response should and should not occur. It includes people, places, materials, wording, timing, sensory features, response forms, and consequences that matter in natural use.

A plan for requesting assistance might include stores, buses, school, work, unfamiliar staff, unavailable items, broken equipment, and several communication forms. “Ask for help anywhere” is too broad to guide teaching.

Separate relevant from irrelevant variation

Relevant features change which response is appropriate. A locked door versus an open one may change whether a key request makes sense. Irrelevant features can vary while the same response remains appropriate, such as the door's color.

The analysis should include both. Teaching only one surface form can create overly narrow stimulus control. Varying every feature randomly can bury the relation the learner needs to detect.

Examples and nonexamples work together

Examples show conditions where the target response applies. Nonexamples show conditions where it does not or where a different response fits.

If teaching safe street crossing, examples might include a walk signal with a clear path. Nonexamples can include a walk signal while an emergency vehicle enters or while a trusted partner signals a specific hazard. Safety expertise and local rules must shape the actual program.

Select a representative teaching set

Choose the smallest practical set that samples the important variation and difficult boundaries. Sequence examples so the controlling features are visible.

A general case programming guide describes defining situations, assessing their variation, selecting teaching examples, and testing across the instructional universe. Another review of instructional design emphasizes careful comparative examples and nonexamples.

Keep teaching and testing cases distinct

Teaching cases provide instruction, prompts, feedback, and reinforcement. Novel test cases evaluate whether performance extends to untrained examples.

Label cases before instruction. If a failed novel case is then taught, it becomes a teaching case for later analyses. Preserve that phase change rather than continuing to call it novel.

A fictional community example

Rosa wants to order a chosen snack at several community counters using speech or AAC. The team maps counter height, menu format, noise, staff position, item availability, payment sequence, wait time, and response options.

They select eight teaching cases and six reserved test cases. Rosa responds effectively in 7 of 8 teaching cases and 3 of 6 novel cases. One novel case lacks an accessible menu, so the team reports it as a system-access failure and keeps it visible rather than scoring Rosa's skill as incorrect.

After the menu is made accessible, the remaining five valid novel cases yield 3 of 5 effective responses. Report both the access failure and revised eligible denominator.

Define the response range too

Generalization can involve varied responses that serve the same outcome. A person might request help through speech, text, sign, gesture, or a communication device.

Specify which forms are effective and recognizable in each setting. Requiring one topography when another works can make a program less useful and less accessible.

Test the boundaries deliberately

Generalization checks should sample typical cases, rare high-impact cases, and near-boundary cases. Include situations where the response should stop, change, or transfer to another person.

Measure opportunity, independence, accuracy, latency, help received, safety, preference, and partner performance as relevant. A percentage without a case list cannot show which variation remains untested.

Multiple exemplars and case planning differ

Multiple-exemplar teaching means using more than one example. The case-planning method supplies a way to choose examples from a defined universe. A large convenience set can still miss the decisive variation.

The BACB outline covers procedures that promote generalization and maintenance. It does not prescribe this planning matrix or establish authority for a particular goal.

Build a case matrix

For every planned case, record:

• setting, person, material, and relevant stimulus features
• expected response and acceptable variants
• conditions requiring a different response
• access supports and partner actions
• teaching, probe, or maintenance status
• risk and stop condition
• result, prompt, error, and next decision

Review the matrix with the person and relevant stakeholders. Add real-world cases discovered during use without erasing the original test set.

Revisit the universe when settings, tools, communication systems, or priorities change. Version the matrix so a later reviewer can see which cases were planned, taught, tested, newly discovered, or appropriately retired.