Partial-Interval Recording and Estimates of Duration in Meta-analyses: Insights from Self-Monitoring Research
Small guesses about how long a behavior lasts can swing a meta-analysis that uses partial-interval data.
01Research in Context
What this study did
King et al. (2025) looked at 30 years of self-monitoring studies that used partial-interval recording (PIR). They wanted to know how guesses about behavior duration changed the final meta-analytic numbers.
The team ran two sets of meta-analyses. One set used the raw PIR scores. The other set added a math fix that accounts for how long the behavior actually lasts.
What they found
The math fix moved the overall effect only a little when the guess about duration was close to perfect. In every other case, the effect size bounced around so much that the meta-analytic verdict could flip.
In short, a small change in the assumed duration can turn "it works" into "it does not work."
How this fits with other research
Cook et al. (2020) also warn about time-sampling error, but at the single-case level. They tell you to collect a few real duration checks so you can spot drift. King’s team scales that same warning up to meta-analyses.
Davis et al. (2018) found that picking a different effect-size index can flip the meta-analytic conclusion for sensory interventions. King shows the same flip can happen within one index just by tweaking a duration assumption.
Sasson et al. (2018) give a reporting checklist so future studies can enter meta-analyses cleanly. King’s work adds one more must-report item: raw duration per behavior.
Why it matters
If you sit on a review team or use meta-analytic evidence to pick interventions, email the original authors for duration data before you trust a PIR-based effect. When you publish your own single-case work, attach a simple duration-per-occurrence column so your study never becomes the outlier that skews the next meta-analysis.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Add a note to your next PIR data sheet that says how many seconds each behavior lasted.
02At a glance
03Original abstract
Abstract Time sampling, wherein an observer indicates whether a behavior occurred within a specified time interval, frequently appears in single-case designs. Partial-interval recording (PIR), the most common form of time sampling, is known to introduce error when used to estimate the duration of a behavior. Recent scholarship has suggested that PIR can be corrected statistically. However, few studies report the data needed to adjust for PIR in meta-analyses conducted to identify effective practices. This could present issues in identifying evidence-based practices in areas, such as research involving children with emotional and behavioral disorders (EBD), where PIR is commonly used. The present study examined the results of a previously published meta-analysis concerning the effect of self-monitoring on the challenging behavior of students with EBD (Bruhn et al. in J Positive Behav Interv 24(2):156–168, 2022). Analyses involved (a) calculating effects for each study based on three imputed average durations of their occurrence and (b) conducting meta-analyses to evaluate the impact of average duration on overall estimates of effect. Results suggest that correcting for PIR resulted in marginal changes when observation intervals were consistent with recommended lengths or when results were conclusive, but otherwise were indicative of a wide range of possible outcomes. A discussion of potential changes to reporting and analysis practices follows a description of results.
Journal of Behavioral Education, 2025 · doi:10.1007/s10864-025-09595-7