Baker Rodrigo Ocumpaugh Monitoring Protocol¶

Collect time-synchronized quantitative field observations of learner behavior and affect by momentarily sampling predetermined individuals with trained observers, explicit codes, and inter-rater reliability controls.

Version: v2 · 2026-09-06 · History
Domain-specific #: 1339
Origin domain: education
Subdomain: learning analytics
Aliases: BROMP, BROMP 2.0, Baker–Rodrigo–Ocumpaugh Monitoring Protocol

Core Idea¶

The Baker Rodrigo Ocumpaugh Monitoring Protocol (BROMP) is a standardized method for rapid quantitative field observation of learner behavior and affect. An observer follows a predetermined sequence of students, watches one learner briefly—normally no more than about twenty seconds under the 2.0 manual—and records the first behavior and affect state that meet the active coding definitions. The observer then moves to the next learner rather than lingering on conspicuous cases. This momentary time-sampling structure is intended to produce a less salience-biased sample of classroom states while retaining natural-setting context.^[1]

BROMP is a protocol family, not merely a list of emotion labels. It couples sampling order, observation duration, holistic coding, operational definitions, observer training, inter-rater reliability, contextual annotation, and synchronized timestamps. Coding schemes can be adapted to a study, but observers must learn the distinctions and demonstrate agreement against a certified coder; the 2.0 materials use Cohen's kappa as the principal reliability gate. The protocol may record behavior and affect simultaneously, but each code remains an observer judgment under the specified scheme, not direct access to a learner's internal state.^[2]

HART, the handheld observation application associated with BROMP, accelerates entry and synchronizes observations with software logs or other streams. It is an implementation aid rather than the abstraction's identity: paper forms or another compliant recorder can instantiate the protocol if timing, sampling, coding, and reliability commitments survive. Conversely, using HART without the observation rules does not constitute BROMP. The official project materials document use across classrooms, educational software, and informal learning settings, and describe certification and cross-cultural adaptation as continuing methodological work.^[3]

The method produces sampled estimates and labeled episodes suitable for descriptive analysis, studying pedagogy, or training automated detectors. It does not provide continuous surveillance, a clinical affect diagnosis, a causal estimate by itself, or a representative sample of every temporal phenomenon. Brief observation can miss short events, cultural display rules can affect judgments, observer presence can alter behavior, and kappa can be influenced by prevalence. Those limitations are part of the measurement package rather than reasons to treat the protocol as a software product or proper-name artifact.

Structural Signature¶

Naturalistic learning setting. Observation occurs in the learner's field environment rather than only in a laboratory task.
Predetermined traversal. Learners are visited in an established sequence to constrain salience selection.
Momentary sample. Each focal observation occupies a short bounded interval.
First qualifying state. The first behavior and affect matching the codebook are recorded rather than an observer-selected dramatic state.
Operational codebook. Behavior and affect categories have explicit definitions and exclusions.
Holistic judgment. Observable posture, action, context, and interaction can jointly inform the code.
Observer training. Coders practice examples and field discrimination before independent collection.
Reliability gate. Agreement with another qualified observer is quantified, typically using Cohen's kappa.
Timestamp synchronization. Observations can align with learning-system events or other data streams.
Context record. Classroom activity, software use, and other framing information are retained when the design calls for them.
Sampling-based inference. Recorded moments estimate state prevalence or association within declared coverage limits.
Adaptation control. Cultural or setting-specific changes require recertification and documented code definitions.

What It Is Not¶

Not HART. The app supports recording and synchronization; it is not the protocol itself.
Not continuous observation. BROMP samples brief moments and therefore has a defined temporal coverage limit.
Not self-report. Codes are observer judgments rather than a learner's own stated experience.
Not automatic affect detection. Its labels may train or validate detectors, but BROMP itself uses human field observers.
Not clinical diagnosis. Operational classroom affect labels do not establish psychiatric or medical conditions.
Not an experimental design by itself. Sampling observations does not create random assignment or causal identification.
Not arbitrary classroom note-taking. Predetermined sampling, code definitions, training, and reliability are constitutive.
Not universally culture-free. Expression and interpretation can vary, so adaptations require evidence and calibration.

Scope of Application¶

BROMP is literal when trained observers need rapid, synchronized, quantitatively analyzable samples of learner behavior and affect in authentic educational settings.

Educational software studies. Observations are synchronized with tutor logs and learner actions.
Classroom engagement research. On-task, off-task, gaming, confusion, boredom, and related coded states are sampled.
Pedagogy comparison. State prevalence is related to classroom activity or teaching format without claiming causal identification automatically.
Detector development. Human codes serve as bounded labels for training and validating automated models.
Cross-cultural adaptation. Codebooks and training are recalibrated for different educational contexts.
Informal learning. Museums or science activities can be studied when the traversal and code rules remain meaningful.
Process evaluation. Changes in observed engagement can accompany iterative learning-product refinement.
Multimodal analysis. Timestamps link field judgment with clickstream, sensor, or contextual records.

Clarity¶

State the protocol version, setting, learner population, observation schedule, traversal rule, focal duration, behavior and affect codebooks, training process, certification criterion, number of observers, reliability statistic, and time-synchronization method. Report missing observations and whether a category was unavailable, unobservable, or absent. Distinguish prevalence across sampled moments from duration, incidence, learner-level persistence, and causal effect. Name any departure from the first-state or predetermined-order rules. Explain cultural adaptation and recertification. Preserve uncertainty around affect: a code is an operational inference based on observed evidence, not a transparent reading of mental experience. Keep the data-capture application, coding scheme, sampling protocol, and analysis model as separate layers.

Manages Complexity¶

BROMP turns a fast-moving classroom into a structured stream of short, comparable observations. Predetermined traversal prevents the observer from concentrating only on unusual learners; brief windows support coverage; codebooks reduce semantic drift; training and kappa make disagreement visible; timestamps support fusion with machine logs. The compression is valuable but lossy. A momentary code can omit sequence, duration, mixed affect, social meaning, and events between visits. Increasing the number of simultaneous code schemes divides attention. Reliability does not ensure construct validity, and synchronized data do not themselves identify causes. The method manages these tradeoffs by specifying what was sampled and how, leaving interpretation and causal design separate.

Abstract Reasoning¶

Define the educational question and decide whether momentary field observation addresses it.
Choose operational behavior and affect categories appropriate to the population and culture.
Construct a predetermined learner traversal and sampling schedule.
Train observers on category definitions, counterexamples, and field conditions.
Evaluate agreement against a qualified observer with a declared reliability statistic and threshold.
Observe each learner for the bounded window and record the first qualifying states.
Timestamp observations and annotate relevant activity context.
Audit missingness, observer drift, prevalence effects, and deviations from traversal.
Estimate the intended prevalence or association at the correct sampling level.
Keep causal claims, automated models, and learner-level diagnoses outside what observation alone warrants.

Knowledge Transfer¶

The strict parent is Measurement. BROMP maps time-bounded learner behavior and affect attributes onto categorical scales through a standardized observer procedure, producing coded values whose uncertainty is partly represented by reliability evidence. Measurement supplies the transferable instrument–attribute–scale relation. The educational accent is momentary classroom sampling, learner traversal, affect/behavior codebooks, observer certification, and synchronization. Monitoring is adjacent, but BROMP need not continuously detect deviations or trigger corrective action.

Examples¶

Canonical¶

During a class using an intelligent tutor, a certified observer follows a fixed seating order. For each student, the observer watches briefly and records the first behavior and affect categories supported by the manual, along with a timestamp. The resulting observations are joined to tutor logs to estimate how often confusion coincides with particular interaction patterns. The code is a sampled operational judgment; it is not a continuous record or a clinical claim.^[1]

Mapped back: predetermined learner sequence + bounded momentary observation + operational behavior/affect codes + trained reliability → synchronized categorical measurements.

Applied / In Practice¶

A museum study adapts the codebook to an informal science activity. Researchers establish local exemplars, retrain observers, and retest inter-rater agreement before collection. They retain the traversal rule and timestamps, then compare sampled engagement across exhibit contexts. Because visitors move unpredictably, the report describes coverage and missingness rather than assuming each visitor received identical observation probability.^[2]

Mapped back: new learning context → controlled code adaptation → renewed reliability evidence → momentary samples → bounded comparison.

Structural Tensions¶

Coverage vs. depth. Brief visits reach more learners but lose sequence and nuance. Diagnostic: Does the research question concern sampled prevalence or sustained episodes?
Reliability vs. validity. Observers can agree on a code that imperfectly represents the construct. Diagnostic: What evidence links the operational category to the intended attribute?
Predetermined order vs. dynamic access. A traversal reduces salience bias but classrooms move. Diagnostic: Which learners or moments become systematically unobservable?
Holistic judgment vs. reproducibility. Context supports inference while expanding discretion. Diagnostic: Are definitions, exemplars, and disagreement procedures adequate?
Cross-cultural use vs. category stability. Expressions and classroom norms differ. Diagnostic: Was adaptation followed by local calibration and reliability testing?
Protocol vs. software product. HART facilitates collection but is replaceable. Diagnostic: Would the sampling and coding commitments survive a different recorder?
Autonomous method vs. generic Measurement. All coded observation measures something. Diagnostic: Are traversal, momentary sampling, first-state coding, and certification jointly present?

Structural–Framed Character¶

Momentary traversal, bounded observation, operational codes, observer training, reliability evidence, and synchronization are structural. The exact codebook, setting, learner population, device, sampling schedule, contextual fields, and downstream model are framed. BROMP yields coded observational evidence, not direct access to internal affect, causal attribution, exhaustive behavioral history, clinical diagnosis, or permission to monitor learners without appropriate ethics and governance.

Structural Core vs. Domain Accent¶

The transferable skeleton is attribute measurement by a trained human instrument under a sampling and reliability contract. The domain accent is authentic educational settings, predetermined learner traversal, short observations, first qualifying behavior and affect, HART-like synchronization, and adaptation across classroom cultures. Removing those features yields Measurement or systematic observation rather than BROMP.

Measurement is the strict parent because BROMP maps observed learner attributes to defined categorical values through a calibrated procedure and retains uncertainty evidence through inter-rater agreement. The edge is composition/presupposes: BROMP is a protocol for measurement, not a subtype of every measurement act.

The prospective workspace queue contains one strict upward edge to prime:measurement. No live DAG mutation is authorized.

Relationships to Other Abstractions¶

Current abstraction Baker Rodrigo Ocumpaugh Monitoring Protocol Domain-specific

Parents (1) — more general patterns this builds on

Baker Rodrigo Ocumpaugh Monitoring Protocol is a kind of Measurement Prime

Measurement is the strict parent because BROMP maps observed learner attributes to defined categorical values through a calibrated procedure and retains uncertainty evidence through inter-rater agreement.

Hierarchy path (1) — routes to 1 parentless root

Baker Rodrigo Ocumpaugh Monitoring Protocol → Measurement

Neighborhood in Abstraction Space¶

Baker Rodrigo Ocumpaugh Monitoring Protocol sits in a sparse region of the domain-specific corpus (96^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (1565 abstractions)

Nearest neighbors

Demonstration (teaching) — 0.77
Neural coding — 0.76
Passive Learning (Passive Engagement) — 0.76
Thin Description — 0.75
Reinforcement learning — 0.75

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With¶

Momentary Time Sampling. A broad sampling family that does not supply BROMP's learner traversal, paired code schemes, training, and synchronization.
HART. A data-entry and synchronization application used with BROMP.
Experience Sampling Method. Usually solicits participant self-reports at sampled times.
Classroom Assessment Scoring System. A different observation framework with different units and dimensions.
Automated Affect Detection. A model-inference task for which BROMP data may serve as labels.
Continuous Behavioral Recording. Tracks occurrences or duration rather than sampled moments.
Experimental Design. Establishes comparison and causal structure beyond the observation protocol.

References¶

[1] Jaclyn Ocumpaugh, Ryan S. Baker, and Ma. Mercedes T. Rodrigo, Baker Rodrigo Ocumpaugh Monitoring Protocol (BROMP) 2.0 Technical and Training Manual (2015), University of Pennsylvania Learning Analytics, https://learninganalytics.upenn.edu/ryanbaker/BROMP.pdf. registry ↩a ↩b

[2] Ryan S. Baker, Jaclyn L. Ocumpaugh, and Juan Miguel Andres, “BROMP Quantitative Field Observations: A Review,” in Learning Science: Theory, Research, and Practice (McGraw-Hill, 2018), https://learninganalytics.upenn.edu/ryanbaker/BROMPbookchapter.pdf. registry ↩a ↩b

[3] University of Pennsylvania Learning Analytics, “BROMP,” official project and certification page, https://learninganalytics.upenn.edu/ryanbaker/bromp.html, accessed August 29, 2026. registry ↩