Representing within-session response rates proportionally and entirely.
Plot cumulative responses as a percent of the total to spot within-session rate changes without picking time bins.
01Research in Context
What this study did
The author built a new way to graph within-session response rates. Instead of slicing the session into fixed time bins, he plotted each peck as a percent of the total.
He showed the method on pigeon data. The x-axis is session time. The y-axis is cumulative pecks shown as a percent of all pecks.
What they found
The graph lets you see local speed-ups or slow-downs without picking bin sizes. A steeper slope means faster responding at that moment.
Flat lines show pauses. Small bumps show brief bursts. All this pops out without extra math.
How this fits with other research
HERRICK (1965) gave rules to scale old-style cumulative records so rate changes look like clear angle shifts. The new percent plot keeps that visual perk but removes the need for angle formulas.
Chandler et al. (1992) showed that pigeon pecks rise then fall inside a session. The percent curve makes that bitonic shape easy to spot.
Richman et al. (2001) later used minute-by-minute cumulative plots in a single child. Their move extends the 1996 idea: watch local proportions, not just session totals.
Why it matters
If you track discrete trials or responses per minute, try plotting cumulative percent completed against session time. You will see exactly when motivation dips or surges without guessing bin width. One glance tells you where to place a break, a prompt, or extra reinforcement.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Add a running percent-of-total line to your usual cumulative graph and note where the slope changes.
02At a glance
03Original abstract
In this technical article, methods for collecting and representing response rates maintained by schedules of reinforcement are presented. First, the time in a session that each important event (e.g., responses, reinforcers) occurs is collected and stored by a computer. Another computer program is used, then, to convert each response to a percentage of the total responses in a session and to plot these percentages cumulatively as a function of the time in the session that they occurred. In this manner, response rates may be expressed proportionally (i.e., using the same y-axis scale regardless of absolute response rate) without requiring the arbitrary selection of an interval over which responses are aggregated and expressed relative to the entire-session rate. A property of these records is that deviations in the slope of the obtained record from the diagonal, which connects (x, y) = (start of session, 0%) to (x, y) = (end of session, 100%), occurring at any point and for any duration, represent changes in the local response rate from the entire-session rate. This method of representing ongoing responding is illustrated by several records of key pecking of a pigeon on a variable-interval 60-s schedule of food reinforcement. Relative local response rates were also computed from these data at several levels of resolution (i.e., the time over which responses were aggregated), including the level typically employed by those interested in within-session changes in response rates.
Journal of the experimental analysis of behavior, 1996 · doi:10.1901/jeab.1996.66-135