Assessment & Research

Further generalization technology: Accounting for natural covariation in generalization assessment.

Pigott (1987) · Journal of applied behavior analysis 1987
★ The Verdict

Plot baseline covariation first or you may call pre-existing links "generalization."

✓ Read this if BCBAs who run generalization probes or write single-case reports.
✗ Skip if RBTs only collecting data and not interpreting it.

01Research in Context

01

What this study did

Pigott (1987) wrote a how-to paper, not an experiment.

He shows how to draw a scatterplot of two untreated behaviors before you start any teaching.

If the points line up, the behaviors already move together, so later "generalization" might be an illusion.

02

What they found

The paper gives step-by-step graphs.

No new data; it simply warns: check baseline covariation or you may credit your program for change that was going to happen anyway.

03

How this fits with other research

Kelly (1973) made the same warning with numbers. He re-analyzed old data with baseline-adjusted ANCOVA and flipped the conclusion. Pigott (1987) turns that idea into a quick picture you can eyeball.

Touchette et al. (1985) used scatterplots to spot stimulus control. Pigott (1987) uses the same tool for a new job: spotting linked behaviors before treatment starts.

Rasing et al. (1992) reviewed dozens of social-skills studies and found most skipped any baseline link check. The scatterplot fix plugs the hole they complained about.

04

Why it matters

Next time you plan a generalization probe, open Excel. Plot the untreated behavior against a similar untaught skill during baseline. If you see a tight diagonal, plan for that shared trend instead of claiming your teaching later caused it. One extra graph can save you from a false positive and a red face in team meeting.

Free CEUs

Want CEUs on This Topic?

The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.

Join Free →
→ Action — try this Monday

Add a scatterplot of two untreated behaviors to your next baseline phase.

02At a glance

Intervention
not applicable
Design
methodology paper
Finding
not reported

03Original abstract

In recent years, the assessment of generalization effects has become a major priority of applied behavior analysis. In this paper we propose a set of procedures to increase the accuracy of generalization assessments by accounting for the degree of natural covariation between treated and untreated behaviors. Scatterplot analyses were used (a) to assess the amount of baseline and postbaseline covariation between behaviors, (b) to determine if the observed generalization effect was due to a preexisting covariation between the behaviors, and (c) to assess if there is a significant change in the strength of the relationship between the behaviors as a function of the intervention. Six hypothetical sets of data are used to demonstrate how these procedures provide more accurate and detailed generalization assessment.

Journal of applied behavior analysis, 1987 · doi:10.1901/jaba.1987.20-273