What Micro Research Cannot Do

1

We Cannot Establish Causation

Micro Research is observational. We document patterns, sequences, and associations. We do not establish causation in the experimental sense.

When we say 'children who experienced X showed Y,' we are reporting an association, not a causal claim. Causal inference requires experimental design with controlled manipulation of variables. That is not what we do.

Our longitudinal data can support causal hypotheses — but proving them requires different methods.

2

Our Findings Are Context-Specific

All our data comes from one context: a Montessori school in Hyderabad, India, with specific demographic, cultural, and pedagogical characteristics.

Findings may not generalize to:

• Non-Montessori settings
• Different cultural contexts
• Different socioeconomic populations
• Different age ranges than those we study

3

We Cannot Eliminate Bias

Embedded observation has advantages (zero observer effect) but also risks (observer bias). Our observers know the children. They have expectations. They have theoretical commitments.

We mitigate bias through:

• Rigorous training
• Inter-rater reliability testing
• Behavioral-only recording rules
• Multiple observers where feasible

4

We Do Not Replace Experimental Research

Micro Research complements experimental research — it does not replace it. Different questions require different methods.

5

Implementation Cannot Be Fully Controlled

While our cyclical model enables rapid implementation of findings, we cannot control implementation with laboratory precision. Guides interpret and adapt findings to their context.

This is a feature (practice-relevant implementation) but also a limitation (implementation varies). We document implementation approaches but do not claim experimental control over them.

6

Design-Research Entanglement

In our Innovation domain, curriculum and environment were co-designed. We cannot fully separate 'curriculum effect' from 'space effect' — this is by design (they are meant to work together), but it is a methodological limitation.

Findings apply to the integrated system, not necessarily to curriculum or space in isolation.

7

Sample Characteristics

Our longitudinal panel, while large (1045 children), represents families who chose Montessori education and can afford private school fees in an Indian urban context. This is not a representative sample of all children.

8

Observer Training Variability

While we enforce minimum reliability standards (≥80% inter-rater agreement), observer skill varies. Some observations may be more reliable than others. We document observer IDs to enable analysis of observer effects.