The Micro-Research Framework.

Question: What material do children choose first when entering the prepared environment?

Protocol: Record child's age (years + months), first material touched, time of entry.

Time Cost: 10 seconds per child.

Question: When do children help each other without adult prompting?

Protocol: Record helper age, recipient age, type of help, adult presence.

Time Cost: 2 minutes per incident.

Zoo Research brings children to labs or exposes them to unfamiliar observers. The setting is controlled but unnatural — children know they're being studied, so they perform. Jungle Research observes children in their everyday environment with familiar adults present. Nothing changes. Behavior stays authentic. Access spans years, not hours. We only conduct Jungle Research. Any study requiring artificial conditions or external observers gets rejected at the design stage — not as preference, but as policy.

Four checkpoints every study must pass before data collection begins. Gate 1 (Longitudinal): Does this connect to children we've observed before? Gate 2 (Naturalistic): Can we observe without disrupting the environment? Gate 3 (Specificity): Is the question bounded and precise — not vague? Gate 4 (Micro): One question, under three weeks of collection, under five minutes per observation, single output. Fail any gate, the study gets redesigned or rejected. No exceptions. The gates exist because loose questions produce unusable data.

Most research produces snapshots — isolated observations at single points in time. We produce something closer to cinema — the same children observed across developmental phases, year after year. This reveals what episodic observation misses: how behaviors emerge, how they evolve, what triggers transitions, what persists. Institutions with rotating subjects and temporary access cannot replicate this. Continuity is our primary methodological asset.

Three ways. First, the See/Hear Rule: record only what you can see or hear. Actions, words, timing, context — nothing else. 'Child was frustrated' fails. 'Child pushed materials away, said I can't do this' passes. Second, inter-rater reliability: Fellows record the same footage, we compare sheets, discrepancies reveal drift into interpretation. We require 80% agreement minimum before deployment and reassess every quarter. Third, disclosure: every publication states that embedded observers have perspectives we reduce but do not eliminate.

Per BEOP v1.0: professional Montessori credential (AMI, AMS, or equivalent), minimum three months working in the specific environment, eight hours of BEOP Observer Training, demonstrated inter-rater reliability at 80% or higher, quarterly reliability checks, and annual ethics refresher. We don't use untrained volunteers. Embedded observation requires trained observers — that's the trade-off.

Observation records behavior: 'Child attempted task four times; completed on fifth attempt.' Interpretation assigns meaning: 'Child struggled.' We capture actions, exact words, timing, context. We don't record emotions, motivations, or judgments — those belong in analysis, clearly separated from the raw record. This separation is what makes our data usable by other researchers.

Correlations, patterns, sequences, temporal relationships. We observe that X precedes Y, or that children who do A tend to also do B. We do not claim X causes Y — that requires experimental manipulation, which we don't conduct. Our contribution is pattern detection across time: seeing what emerges over years of continuous observation. Causal claims belong to controlled experiments. Descriptive claims grounded in extensive naturalistic data belong to us.

It's a selection effect we state explicitly. Our panel consists of children whose families opted into this educational approach — that's not representative of all children. Findings may differ in other contexts, populations, or pedagogies. We note this in every publication. Generalization requires evidence from multiple settings; we provide one data point, not universal claims.

They go into the archive like everything else. Methodological honesty means documenting what we observe, not what we expected. The Montessori tradition gives us our observation culture — systematic watching, careful recording, pattern recognition. It doesn't predetermine our conclusions.

Zenodo provides DOIs, version control, and permanent archival — the infrastructure required for research that will be cited and built upon. Every micro-study becomes a citable, permanent record. We also pursue peer review for work that warrants it. Zenodo and peer-reviewed journals serve different functions; we use both.

It gets documented with 'no significant pattern observed' as the finding. Null results are results — they stop other researchers from chasing the same dead end. We record what we hypothesized, what we observed, and why the data didn't converge. The archive includes failures. Selective publication of only positive results is a known way to corrupt an evidence base; we avoid it.

Yes. Opted-out children are excluded from all data collection. Their behavior is never recorded, even if a protocol is running in their environment. This applies retroactively: if a parent withdraws consent, we remove that child's data from any unpublished study. The consent process is documented in MREF v1.0.

Names become codes at point of collection — not later. Campus names become Site A, Site B. Age is recorded in years and months, never birthdates. Anonymization happens during data capture, not during publication prep. The Child Data Classification Standard (CDCS v1.0) defines four tiers of data sensitivity with handling requirements for each.

Internal ethics review is mandatory before any publication. The review checks consent compliance, anonymization completeness, and whether limitations are accurately stated. We follow MREF v1.0 standards and document compliance in every publication. The ethics framework itself is published — anyone can assess whether we follow our own rules.