Summarizing single-subject research. Issues and applications.
PND was once hailed as the go-to merge tool for single-case graphs, but better options have since overtaken it.
01Research in Context
What this study did
Gaily et al. (1998) wrote a narrative review. They asked: can we pool single-subject graphs into one number?
They focused on Percentage of Non-Overlapping Data (PND). This index counts how many treatment points beat the best baseline point.
The authors concluded that PND-based meta-analysis is a fair way to sum up single-case findings.
What they found
The paper found that PND gives results that match the original graphs. A high PND usually means the graph looks good to the eye.
They said the method is simple, quick, and keeps the spirit of visual analysis.
How this fits with other research
Reid et al. (1999) built on this idea. They widened the defense to any quantitative integration, not just PND.
Nasr et al. (2000) pushed back. They warned that any average is a group comparison and can hide individual quirks.
Cohn et al. (2007) later showed IRD beats PND in tests on 165 graphs. The field moved past the 1998 vote of confidence.
DeHart et al. (2019) offered mixed-effects modeling as a modern route that needs no aggregation at all.
Why it matters
If you write literature reviews, this paper marks the moment PND gained early respect. Know its limits: newer indices and mixed-effects models now give sharper answers. Use PND only for a quick scan, then check IRD or mixed-effects for stronger evidence.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Run IRD on your last single-case graph instead of PND and note any difference in conclusion.
02At a glance
03Original abstract
In this article, literature concerning the quantitative synthesis (meta-analysis) of single subject research literature is reviewed. First, the general rationale for such an approach is discussed. Next, procedures for synthesizing single-subject literature are described, followed by comments and critiques of those procedures. Finally, a review is presented of the results of applications of those procedures. The authors suggest that procedures based on percentage of nonoverlapping data (PND) between baseline and treatment are justifiable, meaningful, and--across nine applications--have produced results that are highly meaningful and faithful to the original research reports.
Behavior modification, 1998 · doi:10.1177/01454455980223001