9.2 Descriptive Functional Behavior Assessment: Direct Observation and Data Collection

The Rationale and Role of Descriptive Functional Behavior Assessment

Descriptive functional behavior assessment (descriptive FBA) involves direct, systematic observation and repeated measurement of the target behavior and its surrounding environmental events as they naturally occur in the client's typical everyday ecology (e.g., classrooms, homes, group homes, job sites), without systematic manipulation or experimental control of independent variables (Bijou, Peterson, & Ault, 1968; Lerman & Iwata, 1993).

Descriptive assessment occupies a vital intermediate position on the functional assessment continuum. While indirect assessments rely on retrospective informant memory, descriptive assessments gather real-time, empirical observational data. Direct observation provides five critical advantages in clinical practice:

1. Direct Empirical Observation: Observers witness the actual topographies, intensities, and environmental contexts of behavior as they occur in real time.
2. Measurement of Baseline Parameters: Establishes objective quantitative baselines (frequency, duration, latency, interresponse time) against which intervention effects will subsequently be evaluated.
3. Refining Operational Definitions: Observers can witness subtle behavioral precursors, response variations, and idiosyncratic topographies, enabling refinement of inclusive and exclusive criteria.
4. Contextual and Environmental Ecological Mapping: Uncovers complex classroom, family, or workplace routines, physical layouts, noise levels, staffing ratios, and peer interactions that trigger or reinforce behavior.
5. Generating Concrete Functional Hypotheses: Identifies specific antecedent-behavior-consequence patterns that warrant confirmation through experimental functional analysis or targeted function-based treatment.

---

Primary Descriptive Assessment Methodologies

Applied behavior analysts utilize three primary direct descriptive methodologies: narrative ABC recording, structured (continuous) ABC recording, and scatterplot analysis.

1. Narrative ABC Recording (Anecdotal Recording)

In narrative ABC recording, the observer maintains an open-ended, continuous, written diary or log of environmental events occurring immediately before and immediately after each occurrence of the target behavior. Whenever the client emits the target response, the observer records:

• The precise time of day and physical setting.
• The immediate antecedent event (what occurred in the environment, who was present, what was said or presented immediately prior to the behavior).
• The exact behavioral topography and intensity.
• The immediate consequence (how staff, peers, or caregivers reacted, what items were provided or removed, what environmental changes took place).

Strengths and Inherent Weaknesses

• Clinical Strengths: Highly flexible and contextually rich. Captures unexpected, idiosyncratic, or complex environmental variables that pre-printed checklists cannot foresee. Highly useful during the earliest exploratory phases of assessment.
• Methodological Weaknesses: Substantially subjective; observers tend to record what they infer was important rather than objective facts. Writing descriptive prose in real time often causes observers to miss subsequent behavioral occurrences. Narrative data are notoriously difficult to quantify, convert into standardized metrics, or evaluate for inter-observer agreement (IOA).

2. Structured (Continuous) ABC Recording

To overcome the quantification limits of narrative logs, structured ABC recording utilizes a standardized, pre-determined data collection sheet featuring operationalized checkboxes or letter codes for anticipated antecedents, specific behavioral topographies, and immediate consequences.

• Antecedent Categories: Checklist items typically include Task/Demand Presented, Attention Diverted / Staff Interacting with Peer, Denied Access to Item / Tangible Removed, Transition between Activities, Alone / Barren Environment, Difficult/Non-preferred Academic Work, or Loud Ambient Noise.
• Behavior Categories: Pre-printed operational definitions of target responses (e.g., Head Banging, Vocal Screaming, Physical Hitting, Property Destruction).
• Consequence Categories: Checklist items typically include Verbal Reprimand / Redirection, Physical Guidance / Prompt Re-presented, Task Terminated / Break Provided, Preferred Item Delivered, Physical Comfort / Soothing, or Ignored / Planned Ignoring.

Strengths and Inherent Weaknesses

• Clinical Strengths: Objective, highly efficient, and easily completed by direct-care staff and Registered Behavior Technicians (RBTs). Quantifiable: enables calculating frequencies, proportions, and rigorous inter-observer agreement. Facilitates mathematical calculation of conditional and unconditional probabilities.
• Methodological Weaknesses: Constrained by the pre-selected categories; unlisted or unexpected antecedents and consequences may be ignored or forced into inaccurate codes. If multiple events occur simultaneously, observers may struggle to record the critical functional event.

3. Scatterplot Analysis (Touchette, MacDonald, & Langer, 1985)

A scatterplot is a visual grid or matrix that divides the observation day into discrete, uniform temporal intervals (typically 15-minute, 30-minute, or 60-minute blocks) across successive calendar days or weeks.

Recording Mechanics and Conventions

For each temporal interval throughout the day, the observer records the occurrence or non-occurrence of the target behavior using standardized cell fills:

• Open / Blank Cell: Zero occurrences of the target behavior during that time interval.
• Single Diagonal Slash / Half-Shaded Cell: Low-to-moderate rate of behavior (e.g., 1 to 2 occurrences, or below a pre-determined threshold).
• Fully Shaded / Cross-Hatched Cell: High-rate or high-intensity behavior (e.g., 3 or more occurrences, or exceeding the threshold).

Clinical Utility and Information Gained

Scatterplots do not identify specific controlling antecedents or consequences; instead, they reveal temporal predictability and environmental routines:

• Uncovers times of day consistently associated with high behavioral density (e.g., severe self-injury consistently clustering between 12:30 PM and 1:30 PM during cafeteria lunch).
• Identifies daily transitions, specific staff shifts, or instructional subjects correlated with behavioral escalations.
• Enables targeted deployment of descriptive ABC observations or experimental functional analyses during high-probability temporal windows.

---

Quantitative Analysis of Descriptive Data: Conditional Probabilities

Rather than relying on informal visual inspection of ABC sheets, behavior analysts quantify descriptive data by calculating conditional probabilities and comparing them to unconditional background probabilities (Lerman & Iwata, 1993; Vollmer et al., 2001).

Calculating Conditional Probability

A conditional probability measures the mathematical likelihood that a specific consequence ($C$) will immediately follow a specific target behavior ($B$):

$P(C \mid B) = \frac{\text{Total instances where Target Behavior is followed by Consequence } C}{\text{Total instances of Target Behavior emitted}}$

Similarly, behavior analysts calculate the conditional probability of behavior given a specific antecedent ($A$):

$P(B \mid A) = \frac{\text{Total instances where Antecedent } A \text{ is followed by Target Behavior}}{\text{Total instances of Antecedent } A \text{ presented}}$

Comparing Conditional vs. Unconditional Probabilities

To determine whether a consequence is truly contingent upon behavior, the conditional probability $P(C \mid B)$ must be compared to the unconditional background probability of that consequence occurring in the absence of behavior, $P(C \mid \text{No } B)$:

• If $P(C \mid B)$ is substantially higher than $P(C \mid \text{No } B)$, a positive correlation exists between behavior and the consequence, supporting a functional contingency hypothesis.
• If $P(C \mid B) \approx P(C \mid \text{No } B)$, the consequence occurs at an identical rate whether the behavior occurs or not, demonstrating that the consequence is merely a noncontingent background event.

---

Major Limitations of Descriptive FBA: The Correlational Trap

Despite its direct observational nature, descriptive functional assessment possesses a severe, inherent methodological limitation: Descriptive assessments are strictly correlational, NEVER causal.

The High-Frequency Background Stimulus Trap (False Positives)

In natural environments, certain environmental events occur at exceptionally high base rates. Caregivers, teachers, and therapists are socially conditioned to provide frequent verbal attention, ongoing conversational commentary, and repetitive instructional demands.

When a client emits a problem behavior, the probability that the caregiver will deliver attention is remarkably high simply because attention is delivered almost continuously. Vollmer, Borrero, Wright, Van Camp, and Lalli (2001) and Thompson and Iwata (2007) empirically demonstrated that descriptive ABC data routinely yield false-positive indications of social attention:

• In descriptive observations of individuals whose problem behavior was actually maintained by automatic reinforcement or escape, ABC data often showed caregiver attention following a large share of problem behavior (reprimands or soothing).
• Because attention followed behavior with high frequency, naive observers incorrectly hypothesized an attention function.
• When subjected to experimental functional analysis, attention was revealed to be an irrelevant bystander; the actual maintaining variable was escape or sensory stimulation.

Because descriptive assessments cannot manipulate environmental variables or hold background events constant, they cannot distinguish between accidental coincidences, caregiver habits, and genuine reinforcing contingencies. Therefore, descriptive data must be interpreted as hypotheses to be verified, never as definitive proofs of behavioral function.

---

Comparative Matrix of Descriptive Assessment Methodologies

The following table compares the operational mechanics, data collection mechanisms, primary findings, clinical strengths, and major confounding risks of the primary descriptive FBA methods:

---

Clinical Troubleshooting and Common Descriptive Traps

• Trap 1: Confusing Correlation with Causation: Observing that a student screams whenever a math worksheet is presented ($P(\text{Scream} \mid \text{Math}) = 0.90$) does not prove an escape function. It is possible that the math teacher wears a strong perfume that triggers an allergic headache, or that the student screams because math class is the only time peer attention is unavailable.
• Trap 2: Ignoring Setting Events During ABC Recording: An observer records an antecedent as "Teacher said open your book" and consequence as "Student was redirected", failing to record that the student arrived at school after zero hours of sleep and skipped breakfast. Proximal antecedents can only be properly interpreted within the context of distal setting events.
• Trap 3: Misinterpreting Scatterplot Patterns: A scatterplot shows high rates of aggression every day between 10:00 AM and 10:30 AM. Clinicians must not assume the intervention is simply to cancel 10:00 AM activities; the scatterplot merely signals where to conduct focused ABC data collection to uncover the actual operant variables.