Decide: are partial (aborted) dyads usable for analysis?

[htsukamoto5]

Phase 1 — unblocked, and a decision rather than code. Settle this first, because it changes what the dropout abort path does.

When a dyad dies partway through, is the completed portion usable?

With C&WG's 6 trials, a dyad that dies at trial 4 still has 3 clean trials — and the words-per-figure effect is steepest across exactly those early trials, so partial dyads may carry real signal. With Hawkins' 72 trials the same logic applies per block.

Why it matters beyond analysis

Effective attrition cost depends directly on the answer:

• If partial dyads are usable, the abort path must flush data, record n_trials_completed, and the cost model improves substantially over the READINESS.md section 5 estimate.
• If they are not, the abort path only needs to pay people and exit, and the recruitment target has to absorb the full loss of every broken dyad.

Work

• Decide the minimum usable trial count for C&WG (and for Hawkins, minimum complete blocks).
• Decide whether a partial dyad enters the primary analysis or only a robustness check.
• Write the decision and its rationale into READINESS.md so the analysis plan and the abort path stay consistent.

See READINESS.md sections 2 and 5.