Boundary Conditions, Scope & Identifiability (psychrev-boundary-conditions)

When to trigger

• Your theory reads as if it explains everything (a red flag to reviewers)
• You have not said where the theory stops holding
• For a formal model: you have not checked whether distinct parameter settings are distinguishable
• A reviewer will ask "what would falsify this?" or "can you even estimate that parameter?"

Why scope is a contribution, not a confession

At Psychological Review, stating where a theory holds and where it fails is part of the theory itself, not a limitations paragraph tacked on at the end. A theory that "explains everything" explains nothing — unbounded scope signals an unfalsifiable model. Editors read explicit boundaries as a sign of theoretical maturity. Three kinds of limit must be stated.

1. Scope (the explanatory boundary)

• Phenomena in scope vs. phenomena explicitly out of scope (left to other processes).
• Population / domain limits — does the theory claim to hold across development, species, cultures, tasks, or only within a stated range?
• Level of analysis — computational, algorithmic, or implementational (Marr); be consistent, and theorize level shifts rather than sliding between them.
• Conditions of breakdown — name the regime where the mechanism should stop producing the phenomena, and treat that prediction as a test of the theory.

2. Identifiability (for formal/computational models)

This is the modeling-specific boundary reviewers probe hardest:

• Structural identifiability — can the parameters, in principle, be recovered from the kind of data the theory addresses, or do different settings produce identical predictions (a mimicry problem)?
• Parameter recovery — demonstrate, by simulation, that fitting the model to data it generated recovers the true parameters; report where recovery degrades.
• Model mimicry — can your model and a rival mimic each other on the available data? If so, the comparison is not diagnostic; say what data would separate them.
• Sloppiness / sensitivity — note parameters the predictions barely depend on (and resist over-interpreting them).

3. What it does NOT explain

A short, explicit list of phenomena the theory deliberately does not address, with one line each on why (out of scope vs. genuinely open). This pre-empts the "but it can't handle Y" reviewer objection by conceding Y on your own terms.

Checklist

• Phenomena in scope and explicitly out of scope are both listed
• Population/domain/level limits are stated (development, species, culture, task, Marr level)
• A breakdown regime is named and treated as a test, not a disclaimer
• (Formal) structural identifiability is discussed; mimicry risk addressed
• (Formal) parameter recovery is demonstrated by simulation, with degradation noted
• (Formal) data that would separate the model from a mimicking rival are specified
• A short "what this theory does not explain" list is included with reasons

Anti-patterns

• A theory presented as universal, with no stated breakdown condition
• Boundary conditions written as apologies ("a limitation is...") rather than as theory
• Skipping identifiability for a model with many free parameters
• Claiming parameters are meaningful without ever showing they can be recovered
• Ignoring that a rival model mimics yours on the available data
• Burying scope limits in a final paragraph instead of theorizing them up front

Output format

【In scope】[phenomena explained]
【Out of scope】[phenomena left to other processes, with reasons]
【Domain limits】[development / species / culture / task / Marr level]
【Breakdown regime】[where the mechanism should stop — stated as a test]
【Identifiability】structural: ok/at-risk | recovery: demonstrated/degraded where [...] | mimicry: [rival], separating data: [...]
【Does NOT explain】[short explicit list]
【Next step】psychrev-conceptual-exhibits (diagram + simulation figures) → psychrev-contribution-framing