4.2 Discontinuous Measurement Systems & Measurement Artifacts

Principles of Discontinuous Measurement

While direct continuous measurement captures every occurrence of a target behavior, direct continuous observation is not always clinically feasible. In many applied environments—such as busy public school classrooms, group vocational centers, community job sites, and residential living facilities—a practitioner must deliver academic instruction, provide physical care, or manage a group of learners while concurrently recording data. When a solo practitioner cannot maintain uninterrupted, continuous visual contact with a client, discontinuous measurement systems are deployed.

Discontinuous measurement (also known as time sampling) divides an observation session into discrete, equal time intervals and records whether the behavior occurred during, throughout, or at the boundary of each interval. By sampling time rather than tracking every response from onset to offset, discontinuous measurement dramatically reduces observer burden.

The Fundamental Sampling Trade-Off

Discontinuous measurement involves an inherent scientific compromise: it sacrifices dimensional precision in exchange for practical feasibility. When behavior analysts convert continuous behavioral streams into dimensionless percentages of scored intervals, they introduce systematic measurement error. An assistant behavior analyst must thoroughly understand the direction and magnitude of these errors to select the appropriate system and avoid clinical misinterpretation.

                  [ SPECTRUM OF DISCONTINUOUS TIME SAMPLING ]
                                       |
         +-----------------------------+-----------------------------+
         |                             |                             |
         v                             v                             v
  [ PARTIAL-INTERVAL ]          [ WHOLE-INTERVAL ]         [ MOMENTARY TIME SAMPLING ]
 Scored if behavior occurs     Scored ONLY if behavior     Scored ONLY if behavior
 at ANY point during interval  persists ENTIRE interval    occurs at TERMINAL instant
         |                             |                             |
 Systematic OVERESTIMATION    Systematic UNDERESTIMATION   NO directional duration bias;
 of duration; undercounts     of duration & rate           prone to random sampling error
         |                             |                             |
 Target: BEHAVIOR REDUCTION    Target: BEHAVIOR INCREASE    Target: BUSY MULTI-TASKING

---

Partial-Interval Recording (PIR)

In Partial-Interval Recording (PIR), the observation period is divided into brief, equal intervals (typically 5 to 15 seconds). The observer records a positive score ($+$) for an interval if the target behavior occurs at any point whatsoever during that interval, even if the behavior persists for only a fraction of a second.

Systematic Directional Bias: Overestimation of Duration

PIR produces a profound, systematic overestimation of the overall duration of behavior. If an observation session is divided into 10-second intervals, a client who emits a single 0.5-second vocal tic will receive a positive score for that entire 10-second interval (100% of that interval is scored). If the client emits that brief 0.5-second vocalization once every 10 seconds, PIR will report that the behavior occurred in 100% of intervals—giving the false impression that the behavior was occurring continuously throughout the entire session when it actually occurred for less than 5% of the total time.

Furthermore, PIR underestimates the count of high-rate behaviors because it cannot differentiate between an interval containing 1 occurrence and an interval containing 10 occurrences: multiple responses within the same interval are collapsed into a single positive mark.

Clinical Indication: Behaviors Targeted for Reduction

Despite its upward distortion, PIR is the primary discontinuous measurement system recommended for behaviors targeted for reduction or elimination (e.g., self-injurious behavior, physical aggression, property destruction, elopement, vocal stereotypy).

• Why PIR is Clinically Indicated for Reduction: In clinical intervention, overestimating problem behavior represents an ethically sound, conservative strategy. PIR ensures that even brief, fleeting instances of severe behavior are captured and not overlooked. If an intervention successfully drives PIR data down to near-zero levels, the clinician can be fully confident that the behavior has genuinely decreased across all temporal dimensions.

---

Whole-Interval Recording (WIR)

In Whole-Interval Recording (WIR), the observation period is divided into equal time intervals (typically 5 to 15 seconds). The observer records a positive score ($+$) for an interval only if the target behavior occurs continuously throughout the entire uninterrupted duration of that interval.

Systematic Directional Bias: Underestimation of Duration and Rate

WIR produces a severe, systematic underestimation of both overall duration and rate. If an individual engages in continuous on-task academic engagement for 9.5 seconds of a 10-second interval, but glances away or drops their pencil for the final 0.5 seconds, the entire interval must be scored as a non-occurrence ($-$). Consequently, an individual could spend 80% of an entire observation session engaged in appropriate behavior, yet receive a WIR score of 0% if their engagement was briefly interrupted during every single interval.

Clinical Indication: Behaviors Targeted for Acquisition and Increase

WIR is the discontinuous system of choice for behaviors targeted for increase or skill acquisition (e.g., sustained on-task academic study, continuous cooperative play, attending during group instruction, remaining seated during mealtime).

• Why WIR is Clinically Indicated for Increase: Underestimating desirable behavior is a conservative clinical strategy. By demanding that the client sustain the behavior across the entire interval without interruption, WIR prevents the clinician from declaring premature mastery. If a student achieves high scores under WIR (e.g., on-task in 85% of intervals), the educational team can be certain that high-level, sustained task engagement has truly taken root.

---

Momentary Time Sampling (MTS)

In Momentary Time Sampling (MTS), the observation session is divided into intervals (ranging from 10 seconds to several minutes). The observer does not watch the client during the interval. Instead, the observer looks at the client strictly at the exact instant the interval terminates (e.g., upon hearing a quiet timer chime or vibrating watch) and records whether the target behavior is occurring at that precise split-second.

Absence of Systematic Directional Duration Bias

Unlike PIR (which systematically overestimates duration) and WIR (which systematically underestimates duration), MTS does not exhibit a systematic directional bias regarding overall duration when measured across a large number of samples. Long-term empirical studies confirm that the percentage of intervals scored via MTS closely approximates the true cumulative duration percentage obtained from continuous duration recording.

Limitations: Random Sampling Error and Low-Rate Sensitivity

While MTS lacks directional duration bias, it introduces substantial random sampling error, particularly when interval lengths are long or the number of intervals is small. Most critically:

• MTS is completely contraindicated for low-rate, brief behaviors. A client who engages in dangerous self-injurious head-hitting that lasts 2 seconds per episode will almost certainly be missed by an observer checking only at the 60-second chime, reporting 0% occurrences when dangerous behavior actually transpired.
• MTS is optimal for high-rate or continuous behaviors (e.g., in-seat behavior, thumb-sucking, manipulating task materials) monitored by busy classroom teachers or residential staff who must maintain other duties throughout the day.

---

Planned Activity Check (PLACHECK)

The Planned Activity Check (PLACHECK) is a group-observation variation of Momentary Time Sampling developed by Doke and Risley (1972). It allows a single practitioner to measure the collective engagement of a group of individuals without needing separate observers for each person.

Protocol and Calculation

1. At the scheduled interval boundary (e.g., every 5, 10, or 15 minutes), the observer initiates a standardized scan of the target setting.
2. The observer visually scans the room from left to right, counting the number of individuals who are actively engaged in the assigned target activity at the instant they are viewed.
3. The observer records both the number of engaged individuals and the total number of individuals present in the area.
4. Group engagement percentage is calculated using the following formula:

$\text{PLACHECK Percentage} = \frac{\text{Number of Individuals Engaged in Target Activity}}{\text{Total Number of Individuals Present}} \times 100\%$

• Clinical and Administrative Applications: PLACHECK is widely used in day habilitation programs, inclusive preschools, specialized autism classrooms, and psychiatric inpatient units to measure the overall environmental quality, teacher instructional delivery, and student participation during structured group routines.

---

Measurement Artifacts and the Impact of Interval Length

A measurement artifact is an apparent experimental finding, trend, or outcome that results from the measurement procedure itself rather than reflecting genuine changes in the target behavior. In applied behavior analysis, measurement artifacts frequently mislead clinicians into believing an intervention is succeeding or failing when the data are merely reflecting the mathematical distortions of the recording system.

The Interval Length Parameter

The primary driver of measurement artifacts in discontinuous systems is the length of the observation interval. Consider the dramatic impact of varying interval lengths across identical physical behavior:

   Physical Reality: Client screams for 2 seconds at minute 0:15, 1:15, 2:15, 3:15, and 4:15.
   True Session Duration: 10 seconds of screaming across 5 minutes (3.33% true duration).
   
   Condition A: 10-Second Partial-Interval Recording (PIR)
   - 30 intervals total; screaming occurs in 5 distinct intervals.
   - PIR Result: 5 / 30 = 16.7% of intervals scored.
   
   Condition B: 60-Second Partial-Interval Recording (PIR)
   - 5 intervals total; screaming occurs in all 5 intervals.
   - PIR Result: 5 / 5 = 100.0% of intervals scored!

As demonstrated in the comparison above, expanding the interval from 10 seconds to 60 seconds transforms the reported data from 16.7% to a staggering 100%—a sixfold artifactual inflation—while the client's actual behavior remained completely unchanged.

Critical Parameter Selection Rules

1. In Partial-Interval Recording: Shorter intervals (e.g., 5 to 10 seconds) minimize overestimation. As intervals expand beyond 15–30 seconds, overestimation artifacts grow rapidly.
2. In Whole-Interval Recording: Shorter intervals minimize underestimation. If intervals are too long (e.g., 60 seconds), the probability of an uninterrupted 60-second bout of engagement approaches zero, yielding massive underestimation artifacts.
3. Rule of Thumb: If high precision is required to evaluate subtle independent variable manipulations, utilize continuous measurement. If discontinuous systems must be deployed, select the shortest interval length that can be reliably executed by the observer.

---

Selecting a Measurement System (Task C.7)

Choosing a system is a two-part decision: which dimension answers the clinical question, and what the logistics of observing allow.

1. Pick the dimension first. Use count or rate for discrete responses whose frequency matters (hits, mands); duration for responses that last (tantrums, on-task); latency when the delay to start matters (compliance); IRT when spacing matters (eating pace); and permanent products when the behavior leaves a reliable product (completed work).
2. Prefer continuous measurement when an observer can watch the whole session and each response has a clear onset and offset.
3. Use a discontinuous system only when continuous recording is impossible: partial-interval recording for behavior targeted for reduction, whole-interval recording for behavior that should be sustained, and momentary time sampling for ongoing behavior when the observer has other duties or is watching several people (PLACHECK for groups).
4. Check feasibility and validity. Can the person collecting data do so accurately while working? Can a second observer collect IOA? Will the system detect the change you expect? A system that a busy teacher cannot run accurately produces worse data than a simpler one run well.

---

Comparing Discontinuous Measurement Systems

The following table contrasts the operational definitions, scoring mechanics, directional estimation biases, clinical use cases, and primary pitfalls across discontinuous measurement systems: