Graphing and Visual Interpretation

The Equal-Interval Line Graph and Its Conventions

The equal-interval line graph (the simple line graph) is the workhorse of ABA data display, because it makes behavior change visible across time and across conditions. "Equal-interval" means each equal distance on the y-axis (ordinate) represents an equal absolute change in the dependent variable, and each equal distance on the x-axis (abscissa) represents an equal passage of time or successive sessions. The exam expects you to recognize its parts:

• x-axis (abscissa): the horizontal axis, almost always time or sessions.
• y-axis (ordinate): the vertical axis, the dependent variable (rate, count, percent, duration).
• Data points and data path: points are connected within a condition to form the path; do not connect across phase-change lines.
• Condition (phase) change lines: vertical lines separating conditions (e.g., baseline vs. treatment).
• Phase / condition labels: text identifying each condition.
• Figure caption and axis labels: identify what is plotted.

A convention candidates miss: data paths are broken (not connected) across condition-change lines and across missing-data gaps, so a reader does not infer continuity that was not observed.

Other Display Formats: Cumulative, Bar, and the Standard Celeration Chart

Different questions call for different displays:

The cumulative record (used heavily by Skinner) accumulates responses, so the line can only rise or stay flat. The slope encodes rate: a steep slope is fast responding, a flat segment is no responding. On the Standard Celeration Chart, the y-axis is logarithmic, so celeration (acceleration or deceleration of rate across time) plots as a straight line; a ratio like x2 per week means rate doubles weekly. Because the SCC standardizes the scale, performance can be compared across behaviors and learners. Bar graphs are best for summary comparisons but sacrifice the trial-by-trial detail of a line graph.

Visual Analysis: Reading the Graph Conservatively

Visual analysis is how behavior analysts judge whether a change occurred and whether the intervention is responsible. Six features structure the judgment, analyzed within each phase first, then between adjacent phases:

• Level: the average value of the data within a condition.
• Trend: the direction and slope of the data path (increasing, decreasing, flat).
• Variability: how much data bounce around the trend; high variability blocks confident conclusions.
• Immediacy of effect: how quickly the data change after a condition change; immediate change strengthens causal claims.
• Overlap: the proportion of data points in adjacent phases that share the same range; less overlap = stronger effect.
• Consistency: whether similar conditions produce similar patterns across replications.

Decision rules the exam rewards are conservative: continue if data are improving and acceptable; modify if data are flat with adequate integrity; collect more data if variability prevents interpretation; and check graph construction (axis scaling, labels, broken paths) if the display itself distorts the pattern.

Avoid conclusions from a single data point unless the behavior is severe and the decision is risk-based. When level, trend, variability, overlap, and immediacy conflict, the defensible answer usually asks for more data or a design change before claiming experimental control, because replicated changes, not one attractive point, demonstrate effect.

How Axis Scaling and Construction Distort Interpretation

A graph can be technically accurate yet visually misleading, and Domain C tests whether you catch construction problems before drawing conclusions. The most common distortion is y-axis scaling. Compressing the ordinate (large range) can flatten a real, meaningful change into a visually trivial wiggle; stretching the ordinate (small range) can magnify trivial noise into an apparent dramatic effect. Before interpreting any change, confirm the axis range and units are appropriate to the behavior and decision. A second distortion is a truncated or non-zero baseline origin, which can exaggerate differences between conditions.

Other construction issues to scan for:

• Unequal x-axis spacing, where sessions are not evenly spaced, hides gaps in time and can fabricate a smooth trend.
• Connecting data across phase lines or missing-data gaps, implying continuity that was not observed.
• Inconsistent data-point symbols across conditions, making series hard to distinguish.
• Missing axis labels, units, or phase labels, leaving the reader to guess what is plotted.

When a question shows a striking effect, a disciplined analyst first asks whether the construction produced the appearance. If the y-axis is compressed and the change still looks small, the effect may actually be large once rescaled; if the y-axis is stretched, an impressive-looking jump may be clinically trivial.

Finally, match the display to the question. Use a line graph for ongoing decisions about trend and level; a cumulative record when total output and rate-via-slope are the focus; a bar graph for summary comparisons across conditions, participants, or settings; and the Standard Celeration Chart when fluency and celeration (proportional rate change) are the targets.

Choosing the wrong display, such as a bar graph that erases within-condition variability when variability is exactly what the decision needs, is itself a measurement error the exam will penalize. The defensible reader interprets only after verifying that the graph type and its scaling faithfully represent the behavior.