The representativeness of observational samples of different durations.
Short sessions can erase low-duration behaviors from your data—match the window to the behavior's real bout length.
01Research in Context
What this study did
The team ran computer simulations of adult residents with intellectual disability. They asked: how short can an observation window be before it hides low-duration behaviors?
They varied session length and watched which behaviors showed up in the final data set.
What they found
Short samples left out the brief, low-duration responses. Only longer windows caught the full pattern.
The authors say you must test session length against the behavior's real bout length or your graph lies.
How this fits with other research
Aragona et al. (1975) first showed that whole-interval sampling always under-counts and partial-interval always over-counts. Mace et al. (1990) widen the warning: even the total session length can mislead if it is too brief.
Flood et al. (2011) later proved the danger is real. With hens, short sessions created higher response rates and steeper demand curves. Their lab data back up the 1990 simulation.
McLennan et al. (2008) found that within-session shifts in reinforcer frequency also change the numbers you see. Together these papers say the same thing: time choices shape the data before the pen hits the paper.
Why it matters
Before you write a treatment plan, run a one-day probe. Watch the behavior across several natural cycles. Pick the shortest window that still captures the low, quick responses you care about. Your baseline will be safer and your decisions faster.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Time five natural cycles of the target behavior today, then set your observation window at the longest bout plus a small buffer.
02At a glance
03Original abstract
The representativeness of behavioral observation samples with durations of less than the whole time of interest was investigated. A real-time recording system was developed to quantify the behavior of 5 profoundly mentally retarded physically handicapped adult students in an institutional training setting. Behavior was observed using six mutually exclusive and exhaustive categories during 2.5-hr observation sessions. Sample observation sessions with durations ranging from 15 to 135 min were computer simulated from the whole-session (150-min) records. It was found that the representativeness of these samples, when compared to whole-session records, was a function of the relative duration of the behavioral categories and of sample duration. The occurrence of relatively high-duration behaviors (lasting for more than 50% of the session) was estimated to within 20% error by samples of less than 60 min, but low-duration behaviors (1 to 3% of the session) were inadequately quantified even from 135-min samples. Increasing irregularity of bouts of behavior in the low-duration behaviors is suggested as the cause of the functions obtained. Implications of the findings for applied behavior analysis are discussed, with the recommendation that the adequacy of observational session durations be empirically assessed routinely.
Journal of applied behavior analysis, 1990 · doi:10.1901/jaba.1990.23-323