5.1 Graphing Conventions: Line Graphs, Cumulative Records, and Standard Celeration Charts

The Scientific Rationale for Graphic Display in Behavior Analysis

In applied behavior analysis (ABA), graphic displays are not mere secondary summaries or decorative illustrations; they represent the primary medium through which behavior analysts organize, analyze, interpret, and communicate clinical data. Direct and repeated measurement of behavior across time generates a longitudinal time series. Transforming raw numerical counts and observation logs into visual displays allows practitioners to inspect dynamic behavioral changes as they unfold in real time.

B.F. Skinner (1953) emphasized that visual displays provide direct access to the ongoing interaction between an organism and its environment. Behavior analysis maintains five fundamental scientific and clinical justifications for prioritizing graphic displays over tabular numbers or post-hoc statistical summaries:

1. Immediate Visual Feedback: Plotting data immediately following or during observation sessions provides continuous visual contact with the behavioral trajectory. The clinician instantly sees whether responding is accelerating, decelerating, or remaining stable, enabling timely clinical adjustments.
2. Continuous Monitoring of Behavioral Dynamics: Unlike pre-test/post-test evaluations, continuous graphing reveals trends, cyclical fluctuations, day-to-day variability, and unexpected behavioral bursts that static numerical aggregates or summary averages completely conceal.
3. Conservative Standard for Scientific Judgment: Visual inspection represents a rigorous, conservative approach to discovering functional relations. Rather than relying on mathematical probabilities or inferential statistics ($p$-values) that can declare trivial, clinically negligible differences "statistically significant" with large sample sizes, visual inspection demands that independent variable manipulations produce obvious, robust, and socially valid behavioral changes apparent to the naked eye.
4. Facilitation of Stakeholder Communication: Visual displays provide an intuitive, objective, and transparent format for communicating progress to clients, parents, teachers, interdisciplinary professionals, and funding agencies without relying on technical statistical jargon.
5. Active Clinical Decision-Making Tool: Graphed data serve as an ongoing compass for behavioral programming. Clinicians utilize graphed trajectories to determine when to advance mastery criteria, thin reinforcement schedules, fade prompts, probe for generalization, or modify ineffective intervention packages.

---

Equal-Interval Line Graphs (Frequency Polygons)

The equal-interval line graph—technically designated a frequency polygon—is the most pervasive and versatile graphic display in behavior analysis. In an equal-interval line graph, equal physical distances along either axis represent equal absolute quantities of time or behavior.

Fundamental Anatomy of the Line Graph

To construct an equal-interval line graph that follows standard behavior-analytic conventions, an assistant behavior analyst must understand its essential structural components:

• Horizontal Axis (X-Axis / Abscissa): The horizontal line representing the passage of time or successive measurement occasions. Typical abscissa metrics include consecutive sessions, calendar days, weeks, or instructional trials. The horizontal axis is calibrated in equal increments from left to right.
• Vertical Axis (Y-Axis / Ordinate): The vertical line representing the dimensional measure of the target behavior (the dependent variable). Common ordinate metrics include response frequency, rate per minute, duration in seconds/minutes, latency, interresponse time (IRT), percentage of intervals, or trials-to-criterion. The vertical axis must be calibrated in equal increments, beginning at zero at the origin.
• Origin: The intersection coordinate $(0,0)$ where the abscissa and ordinate meet. The vertical axis should always begin at zero to avoid visual distortion of behavioral magnitude.
• Axis Labels and Numerical Calibration: Both axes must feature explicit, fully specified labels and numerical tick marks. The ordinate label must state both the behavioral dimension and the observation time unit (e.g., "Screaming Responses per Minute" or "Percentage of 10-Second Intervals with Off-Task Behavior"). The abscissa label must clearly designate the temporal unit (e.g., "Consecutive Daily Sessions").
• Condition Change Lines: Vertical lines extending upward across the data field to indicate environmental or experimental manipulations:
  • Solid Vertical Line: Indicates a major experimental phase change, such as transitioning from Baseline to an Independent Variable (e.g., DRA + Extinction), withdrawing an intervention, or switching experimental conditions.
  • Dashed Vertical Line: Indicates a minor procedural modification or parameter shift within an existing phase (e.g., thinning a reinforcement schedule from FR1 to FR3, increasing a prompt delay from 0s to 5s, or changing therapy staff).
• Condition Labels: Concise, descriptive titles placed horizontally above the data field and centered over each phase (e.g., "Baseline", "DRA + FCT", "Schedule Thinning"). Condition labels should identify the independent variable clearly without obscuring data points.
• Data Points: Discrete, precisely placed geometric symbols (typically solid dots, open circles, squares, or triangles) plotting the coordinate intersection of a specific temporal session (x-value) and its observed behavioral quantity (y-value).
• Data Paths: Solid straight lines connecting consecutive data points within an experimental condition. The data path illustrates the trajectory, slope, and continuity of responding over time.

Graphic Standards and the Aspect Ratio Rule

A common convention is a vertical-to-horizontal axis ratio of about 5:8 to 3:4, so the horizontal axis is roughly 1.33 to 1.6 times longer than the vertical axis. Adhering to this proportional standard is vital because manipulating axis lengths alters the visual slope of the data path:

• Vertical Distortion (Stretched Y-Axis): If the vertical axis is stretched disproportionately tall while the horizontal axis is compressed, minor, clinically trivial fluctuations appear as steep, dramatic behavioral shifts.
• Horizontal Distortion (Stretched X-Axis): If the horizontal axis is stretched excessively wide while the vertical axis is flattened, steep, rapid, and clinically urgent behavioral trends appear falsely gradual or flat.

Keeping the ratio in this range helps ensure that visual analysis reflects authentic behavioral dynamics across clients, settings, and research studies.

Strict Plotting and Data-Path Rules

Assistant behavior analysts must follow standardized plotting rules when drawing data paths:

1. Never connect data points across condition change lines: Data paths must NEVER bridge a solid or dashed condition change line. Condition change lines demarcate different environmental conditions; drawing a line across the boundary creates a false visual impression of continuous contingency across distinct experimental phases.
2. Never connect data points across significant time breaks: If an extended interruption occurs in data collection—such as a three-week client vacation, an extended hospitalization, a school holiday break, or clinic closure—the data path must be broken. Connecting points across an extended time gap misleads the viewer into assuming continuous, uninterrupted measurement.
3. Never connect data points across distinct measurement scales: If different observation durations or incompatible recording systems are used, data paths must not connect across the transition.
4. Never connect data points to axes or condition change lines: Data paths connect data point to data point only. They must never drop to touch the horizontal axis or anchor to condition lines.
5. Connect identical consecutive values horizontally: When consecutive sessions yield identical values, connect the data points with a straight horizontal data path.
6. Omit data points during missed sessions: If an observation session is cancelled or missed, leave a blank space on the abscissa corresponding to that date; NEVER plot a zero. Plotting zero indicates that an observation took place and zero instances of behavior occurred, which is inaccurate data recording.

---

Bar Graphs (Histograms) in Applied Behavior Analysis

A bar graph (or histogram) is a graphic display consisting of discrete vertical or horizontal rectangular bars whose heights or lengths correspond to the magnitudes of the data. Unlike line graphs, the horizontal axis of a bar graph does not represent a continuous temporal dimension.

Primary Clinical Applications

Bar graphs serve specific, targeted functions in ABA:

• Displaying Discrete Categorical Data: When the dimensions under comparison represent separate categories, participants, settings, or tasks rather than continuous time.
• Summarizing Preference and Reinforcer Assessments: Bar graphs represent the standard format for displaying stimulus preference assessments, such as plotting the percentage of trials each stimulus was selected in a Paired-Choice or Multiple Stimulus Without Replacement (MSWO) assessment.
• Displaying Group or Aggregate Performance: Useful for summarizing program-wide performance, multi-student benchmarks, or comparing initial baseline averages across multiple classroom settings.

Inherent Limitations of Bar Graphs

Despite their visual clarity for categorical summaries, bar graphs have severe limitations in behavior analysis:

• Complete Masking of Variability and Trend: A bar graph collapses all observation sessions into a single aggregate measure (such as a mean or total percentage). Two clients may both exhibit a mean rate of 10 aggressive outbursts per hour across 10 sessions. However, Client 1 may show a steady flat line, while Client 2 exhibited an alarming upward trend from 1 to 25 outbursts per hour. A bar graph displays them as identical bars, completely obscuring the emerging behavioral crisis.
• Inability to Evaluate Dynamic Functional Relations: Because bar graphs omit continuous longitudinal tracking, they cannot reveal immediacy of effect, behavioral contrast, extinction bursts, or treatment degradation over time.

---

Cumulative Records: Mechanics, Slope, and Schedule Analysis

The cumulative record was invented by B.F. Skinner in the 1930s as the primary continuous recording apparatus for the Experimental Analysis of Behavior (EAB). Skinner designed the cumulative recorder to automate the direct, real-time plotting of operant behavior in experimental chambers.

Mechanical Operation of the Cumulative Recorder

The classic electromechanical cumulative recorder operates via a motor-driven transport roll and an inking pen:

1. Constant Paper Transport: A continuous roll of graph paper unrolls at an invariant, constant speed beneath the pen from right to left (representing the uniform passage of time along the horizontal axis).
2. Upward Pen Increments: Every time the organism emits a target operant response (e.g., a lever press or key peck), an electrical microswitch energizes an escapement mechanism that steps the inking pen upward across the paper by a fixed, uniform distance.
3. Continuous Cumulative Summation: If the organism responds rapidly, the pen steps upward repeatedly in quick succession while the paper moves slowly, generating a steep upward line. If the organism stops responding, the pen remains stationary while the paper continues to roll horizontally, drawing a flat, horizontal line.
4. Pen Reset: Because the physical paper roll has finite width, when the pen reaches the top margin of the chart, an automated trip switch releases the pen carriage, resetting it instantly to the bottom baseline ($y = 0$), where cumulative stepping resumes.

Visual and Mathematical Interpretation of Slope

The fundamental interpretative rule of cumulative records is that the slope of the line reflects the rate of responding:

$\text{Slope} = \frac{\text{Change in Cumulative Responses}}{\text{Elapsed Time}} = \text{Rate of Responding } (R/t)$

• Steep Slope: Indicates a high, rapid rate of responding. The more vertical the line, the higher the response rate.
• Moderate / Shallow Slope: Indicates a low, slow rate of responding.
• Flat, Horizontal Line: Indicates a zero rate of responding. Responding has ceased entirely. Time continues to advance horizontally, but because no responses occur, the pen receives zero upward steps.
• The Unbreakable Cumulative Law: The slope of a cumulative record NEVER decreases or points downward. An organism cannot "un-emit" or subtract responses that have already occurred. Therefore, a negative slope is physically impossible on a cumulative record. When responding decelerates or ceases, the slope flattens toward horizontal; it never slopes downward.

Characteristic Response Patterns on Cumulative Records

Cumulative records are famous for uncovering schedule-induced response topographies:

• Fixed Ratio (FR) Schedules: Yield a "break-and-run" pattern consisting of a post-reinforcement pause (flat horizontal segment immediately following reinforcement) followed by an abrupt transition to a steep, steady run of responding until the ratio requirement is completed.
• Fixed Interval (FI) Schedules: Produce the classic FI scallop: a post-reinforcement pause after delivery of the reinforcer, followed by a gradual, accelerating upward curve as the end of the interval approaches.
• Variable Ratio (VR) Schedules: Generate a remarkably steep, steady, and uniform slope with virtually no post-reinforcement pauses, reflecting rapid, persistent responding.
• Variable Interval (VI) Schedules: Produce a uniform, moderate, stable slope without pauses or scalloping.

Applied Clinical Utility

In applied clinical practice, cumulative graphs are frequently plotted electronically to track cumulative skill acquisition. For example, a clinician tracking cumulative vocabulary words mastered, sight words acquired, or math facts retained plots cumulative milestones over weeks. Cumulative records provide an inspiring visual demonstration of continuous, non-decreasing learning progress across long-term educational curricula.

---

Standard Celeration Charts (SCC) & Precision Teaching

The Standard Celeration Chart (SCC) was developed by Ogden R. Lindsley, a doctoral student of B.F. Skinner, as the cornerstone measurement system for Precision Teaching. Lindsley recognized that conventional equal-interval line graphs suffer from a fundamental visual flaw: identical absolute changes in behavior look identical on equal-interval paper, even though their clinical significance varies wildly depending on baseline magnitude.

For example, an increase of 5 responses per minute represents a doubling of performance (×2.0) for a child reading 5 words per minute (from 5 to 10), but represents a negligible 5% change for a child reading 100 words per minute (from 100 to 105). On an equal-interval line graph, both changes produce an identical vertical rise of 5 units. The Standard Celeration Chart eliminates this distortion through semi-logarithmic scaling.

Semi-Logarithmic Scaling Mechanics

The Standard Celeration Chart is a semi-logarithmic chart:

• Horizontal Axis: Calibrated in an equal-interval linear scale representing calendar time (standardized at 140 calendar days or 20 weeks). Because the abscissa tracks real-world calendar time rather than arbitrary instructional sessions, missed days and weekends are visually depicted, preventing false impressions of learning speed.
• Vertical Axis: Calibrated in a logarithmic scale covering six cycles of 10 (a $10^6$ range), spanning from $10^{-3}$ (0.001 responses per minute, or 1 response per 1,000 minutes / ~16 hours) up to $10^3$ (1,000 responses per minute).

Proportional Changes on Semi-Logarithmic Scales

On a logarithmic vertical axis, equal vertical distances represent proportional (multiplicative or dividing) changes, rather than additive numerical differences:

• The physical distance from 1 to 2 is identical to the distance from 10 to 20, 50 to 100, or 500 to 1,000 (all representing a multiplication of ×2.0).
• The physical distance from 10 to 5 is identical to the distance from 100 to 50 (both representing a division of ÷2.0).

The Concept of Celeration

In Precision Teaching, learning is defined as a change in behavioral frequency over time. Celeration is the fundamental measure of learning:

$\text{Celeration} = \frac{\text{Rate of Responding}}{\text{Unit of Time}} = \text{Rate per Minute per Week}$

Celeration quantifies how rapidly a learner's fluency is multiplying or dividing per week:

• Standard Slope Angles: On the standardized SCC dimensions, a straight line drawn at a 34-degree upward angle represents a doubling of rate per week (a celeration of ×2.0). A 34-degree downward angle represents a halving of rate per week (a deceleration of ÷2.0).
• Standardized Learning Visuals: Because the chart dimensions and logarithmic scaling are standardized globally, any behavior analyst or teacher can glance at any SCC, anywhere in the world, and instantly evaluate whether learning is progressing at an acceptable velocity, regardless of the learner's age, disability, or target skill.

---

Comparative Matrix of Graphic Displays in Behavior Analysis

The following table contrasts the defining features, functions, and clinical applications of the primary graph types used across applied and experimental behavior analysis:

---

Clinical Troubleshooting and Common Graphing Traps

• Trap 1: Connecting Data Paths Across Phase Boundaries: Connecting baseline to intervention data points with a line is one of the most frequent errors on the BACB exam. Drawing a line across a phase change line misrepresents the experimental design by implying continuity across distinct environmental contingencies.
• Trap 2: Misinterpreting Horizontal Lines on Cumulative Records: A horizontal line on a cumulative record does NOT mean responding is occurring at a steady, unchanging rate; it means responding has completely stopped (rate = 0). On an equal-interval line graph, a horizontal data path indicates steady responding at that non-zero rate.
• Trap 3: Believing Cumulative Records Can Have a Negative Slope: Because cumulative records plot the ongoing sum of total responses emitted, the count can never decrease. A downward-sloping cumulative record line is physically impossible.
• Trap 4: Plotting Missed Sessions as Zero: If a client misses a session due to illness, plotting a data point at zero records false clinical data. Missed sessions must be left as a blank space on the abscissa.