Question, domain, controls, sampling principles, family rules, blind annotation, statistics and decision rules are immutable.
Unilanguage Experimental Research · UNI-EXP-002
W → WATER / WETNESS / WAVE-FLOW
Does W predict a semantic concentration across independent lexical families?
W → 水／湿润／波动—流动：W 是否在独立词汇家族层面预测语义集中？
Freeze status
Result Freeze v1.0 is complete
Source, preprocessing, exact power, Family Assignment and Sampling Manifests validate and are checksummed.
Machine-based exploratory blind annotation; 357/360 agreement and three adjudications.
W 7/120 vs Controls 9/240; RR=1.556; Fisher p=.4185.
设计、实施、机器盲标和结果现已分别冻结。主结果为不确定；次要分析不得覆盖这一结论。
Research question
A family-level enrichment test
P(Direct Target | W families) > P(Direct Target | pooled Controls)
The experiment tests a lexical distributional tendency in modern English. It does not test whether W itself means water, whether its shape originated as a wave, or whether the relevant words share an ancient source.
本实验只检验现代英语词汇分布中的统计性语义集中；不检验 W 本身是否表示水、字形是否起源于水波，或相关词是否具有共同远古来源。
Primary semantic endpoint
Direct meaning, not loose association
WATER
Water, bodies of water, or the presence and state of water.
水、水体，以及水的存在或状态。
WETNESS / CONTACT
Wetness, soaking, washing, or direct water/liquid-contact states and processes.
湿润、浸泡、冲洗，以及直接水／液体接触状态或过程。
WAVE / FLOW
Wave, flow, oscillation, or undulating movement encoded by the lexical sense.
词义直接编码波、流动、振荡或起伏运动。
Primary coding: `1 = Direct Target`; `2 = Associated/Borderline`; `0 = Not Target`; `U = Unclear`. Only `1` enters the primary endpoint. Habitat, use, cultural association and merely being a liquid do not qualify.
Sampling and independence
One lexical family, at most one vote
Frozen target
- SUBTLEX-US frozen local source
- 120 W families
- 240 pooled Control families
- Three frequency strata
- Master seed: 20260827002
Anti-inflation rule
Transparent derivatives and compounds inheriting the same target meaning are collapsed. Near-level documented stem families contribute once; speculative deep historical relations do not force merging.
透明派生词与复合词不得重复投票；有明确依据的近层级词干家族只计一次，推测性的深历史关系不强制合并。
Hypothesis-blind annotation
Independent judgments before aggregate results
Two annotators independently code standardized conventional senses in separately randomized orders. They do not receive the hypothesis, group labels, frequency strata, family outcome, aggregate counts, or each other's judgments. A third hypothesis-blind adjudicator resolves preserved disagreements under the same frozen rules.
两名标注者按不同随机顺序独立判断标准化常规义项；他们看不到假说、组别、频率层、词族统计贡献、总体计数和对方判断。分歧保留原记录，再由第三名假说盲裁决者依据相同冻结规则处理。
Terminology boundary · 术语边界： This is hypothesis-blind independent semantic annotation, not “fully double-blind” annotation, because annotators can see each lemma.
Primary decision
One test, four possible outcomes
Family-level, alpha=.05; report RR, RD and 95% intervals.
The planning reference RR=3.0 is not the success threshold.
All required conditions must hold.
Every outcome remains publicly reportable.
Read the immutable preregistration record
The full bilingual record separates design decisions from implementation artifacts and preserves the No Rescue and negative-result publication rules.
View Preregistration v1.0 · 查看预注册 v1.0 View Implementation Gate · 查看实施闸门 View Result Freeze v1.0 · 查看正式结果Current result: INCONCLUSIVE. Machine-based exploratory blind annotation — not human double-blind validation.