Cumulative record

Every plotted value includes the past

Suppose session counts are 2, 0, 3, and 1. The cumulative values are 2, 2, 5, and 6. The second point stays level because the total did not grow during that interval. The third point rises by three.

This transformation makes the total-to-date immediately visible. It also means the height of a point is not the response count for that single interval. Reading 6 at the fourth point as “six responses in session four” would be an error.

Slope carries the main information

On a graph with equal time intervals, the slope represents how quickly responses accumulate. A vertical rise of 20 across one hour indicates a higher rate than the same rise across four hours. State the horizontal unit before comparing slopes.

A flat segment means zero newly counted responses during covered time. It can also reflect missing observation if the graphing rule silently carries the prior total forward. Those states must look different. Use a gap, symbol, annotation, or accompanying table for missing or ineligible periods.

The line does not fall

A running total cannot become smaller unless the chart resets or someone corrects an earlier error. If the recorded line drops from 42 to 37, investigate the data transformation, reset convention, or correction history.

Some cumulative charts reset at the start of a day, week, phase, or page. Mark each reset clearly. A new origin creates a new cumulative series; it should never appear to be a sudden reduction in the measured behavior.

The reset period should match the decision. A daily reset can support day-level production review, while one continuous total may better show progress toward a long-term count. Retain the source counts so either view can be reconstructed.

If an earlier count was wrong, preserve the original entry and correction trail according to the applicable documentation policy. Recalculate downstream totals rather than drawing an unexplained descending segment.

A fictional community-work example

Andre is practicing asking for clarification during a community job routine. The team observes four one-hour blocks and counts 3, 5, 0, and 4 defined requests. The cumulative record shows 3, 8, 8, and 12.

The rise is steepest during hour two, level during hour three, and rises again during hour four. The final height shows 12 total requests across four observed hours. The original interval counts remain necessary for calculating hourly rates and examining context.

During hour three, Andre completed familiar tasks and had no eligible uncertainty events. Calling the flat segment a performance decline would be misleading. The record needs an opportunity measure or contextual annotation so readers understand what was available to occur.

Keep raw and cumulative data together

A build-ready record includes:

• the defined response and counting rule
• interval start and end times
• raw count for each interval
• cumulative total after each interval
• observation duration and eligible opportunities
• condition and phase labels
• missing-data and reset rules
• corrections and their authorship

Audit cumulative accuracy as plotted totals that equal the sum of eligible source counts divided by totals due for review. Keep every error in the denominator until corrected.

Choose the display for the question

Cumulative records are useful when total production and changes in rate matter. They can make short-term variability harder to see because every point carries earlier history. A standard noncumulative line graph may better show session-by-session level, trend, and variability.

A bar graph may suit discrete category comparisons. A standard celeration chart may suit repeated frequency data when proportional change is central. The same data can be viewed more than one way, provided the team keeps one verified source and states each transformation.

Use phase markings carefully

Place a phase-change line between the last interval under the old condition and the first interval under the new one. The cumulative line may continue across the change, yet interpretation should compare slopes within each condition.

When a phase begins with a reset, disclose that choice. Otherwise the visual jump back to zero can be mistaken for behavior change.

Training sources and limits

The BACB BCBA Test Content Outline, sixth edition covers graphing, visual analysis, experimental design, and data-based decisions. It is training and examination content, not a required cumulative-record format.

A precision-teaching synthesis describes frequent measurement, charting, and rate-based interpretation. Its examples help explain why time units and chart conventions need to stay stable.