不定式、现在分词和过去分词可以把完整命题降级为句内成分。它们省去时态和显式主语，却保留动作、参与者和逻辑关系。
看到 to do, doing, done 时，如何判断逻辑主语、时间关系和主动被动？关键不是背名称，而是还原被压缩的命题。
-ing 常保留主动或过程视角；过去分词常保留被动或结果状态。The team plans to replicate the study.
to replicate 是 plans 的内容，逻辑主语由 the team 控制。
The team asked the laboratory to replicate the study.
逻辑主语改由宾语 the laboratory 控制，是实验室重复研究。
Using administrative records, the researchers tracked migration.
句首分词的逻辑主语与主句主语一致：研究者使用记录。
Data collected before the reform show a different pattern.
过去分词压缩 data that were collected，保留被动关系。
Having excluded incomplete cases, the team reran the model.
完成分词明确排除发生在重新估计之前，逻辑主语仍是 team。
To interpret the coefficient correctly, one must know the reference group.
不定式表示目的，逻辑执行者与主句泛指主语 one 一致。
The intervention appears to have reduced absenteeism.
完成不定式把减少置于 appears 的判断时刻之前。
The sample is too small to support a precise estimate.
too...to 表示程度阻止结果实现：样本小到无法支持精确估计。
Seeking to reduce congestion, the city introduced a fee charged to vehicles entering the center. Drivers wishing to avoid the fee could travel outside peak hours or use public transport. Having observed an initial decline in traffic, officials declared the policy successful. Yet the data needed to determine whether trips had shifted elsewhere were not collected.
段落包含目的、被动修饰、主动意愿、先行动作和必要性修饰。最后一句提醒：压缩结构能提高密度，也可能让“谁需要、谁收集”变得不显眼。
训练语境：学术科研
还原：participants assigned to treatment。
可还原为 participants who were assigned to treatment。过去分词与 participants 是被动关系，不能理解为参与者分配了别人。
训练语境：学术科研
找逻辑主语：The model allows researchers to estimate spillovers.
to estimate 的逻辑主语是宾语 researchers，不是 model。结构是 allow + 人 + to do。
训练语境：学术科研
解释先后：Having completed the survey, the team began analysis.
完成分词表示完成调查先于开始分析；两个动作共享主语 team。若调查尚未完成，应使用普通 While conducting... 等结构。
训练语境：学术科研
诊断：Using panel data, the effect became smaller.
按默认控制关系，似乎是 effect 使用面板数据，语义不合理。应写 Using panel data, the researchers obtained a smaller estimate，显式提供执行者。
训练语境：学术科研
比较 to measure 与 to be measured。
前者让逻辑主语执行测量；后者让逻辑主语成为被测对象。主动被动变化会交换参与者角色，不能只译作“测量”。
训练语境：学术科研
解释：The result is unlikely to have occurred by chance.
to have occurred 表示结果发生早于当前概率判断；unlikely 评估该过去事件由偶然产生的可能性。
训练语境：真实生活
区别：I stopped smoking 与 I stopped to smoke。
动名词作 stopped 的对象，表示戒烟；不定式表目的，表示停下原来的活动去抽烟。形式微小，事件结构相反。
训练语境：真实生活
还原：the people waiting outside。
相当于 the people who are waiting outside。people 是 wait 的执行者，分词短语限定哪些人。
训练语境：真实生活
解释：I need the report finished by noon.
finished 是宾语 the report 的结果补语，表示需要报告处于完成状态。它不等于说“我需要完成报告”的执行者一定是 I。
训练语境：真实生活
诊断：Walking home, the rain started.
句法让 the rain 成为 walking 的逻辑主语，形成悬垂修饰。应写 While I was walking home, the rain started。
Seeking to reduce the noise in the dataset, the researchers excluded observations recorded during the first week of implementation. The decision, while defensible, makes the estimated effect harder to interpret, because the excluded cases may have captured the period in which users were still learning to navigate the interface. To treat the remaining data as fully representative is therefore to assume that adaptation itself is not part of the intervention's effect.
Seeking to reduce... 的逻辑主语是 the researchers。recorded during... 修饰 observations，表示这些观察值是在第一周记录的。while defensible 是压缩让步结构，完整意思近似 although the decision is defensible。To treat... is therefore to assume... 中两个不定式分别占据主语和表语位置。高阶语境：学术科研
判断 Seeking to reduce the noise in the dataset 的逻辑主语，并说明如果主语不一致会产生什么问题。
逻辑主语是 the researchers，因为是研究者试图降低噪声。如果后面主句主语换成 observations，就会变成“观察值试图降低噪声”，造成悬垂修饰。
高阶语境：学术科研
第三句中 To treat... is therefore to assume... 为什么适合学术写作？
它把一个方法选择转化为一个可审查的前提：把剩余数据视为代表性样本，就等于假设适应过程不属于干预效果。这个结构能暴露推理条件，而不是只陈述操作。
高阶语境：真实生活
用 while defensible 写一句：某个决定可以理解，但会带来解释困难。
可以写：The decision to skip the meeting, while defensible, made the later misunderstanding harder to resolve. while defensible 压缩了让步关系，主干仍是 The decision made... harder...。
每遇到非谓语，尝试补回“谁、在什么时间、主动还是被动”。若补回的主语荒谬，原句很可能存在控制错误或悬垂修饰。