情态表达的不是单纯时间，而是可能性、必要性、能力、许可和推断强度。学术阅读中，忽略 `may`, `must`, `would` 会把谨慎结论误读成确定事实。
同一个 must 为什么有时表示义务，有时表示推断？would have 为什么常在讨论没有发生的路径？关键是识别命题世界与说话者立场。
may, must, cannot。must, should, may。would 常建立假设世界、过去视角中的将来或反复行为。The discrepancy may reflect measurement error.
may 提供一种可能解释，不承诺这是唯一或最可能原因。
The discrepancy must reflect measurement error.
这里的 must 是基于证据的强推断，不是要求差异履行义务。
Participants must provide written consent.
制度语境把 must 解释为义务。
If the assumption were false, the estimate would be biased.
句子建立一个与当前假定相距较远的世界，并说明在该世界中的后果。
If the team had preregistered the analysis, the criticism might have been avoided.
过去完成形式把条件标为未实现；might have 仍只表达可能避免，而非保证。
That cannot be the whole explanation.
cannot 否定的是解释成立的可能性，通常表示说话者认为证据排除了它。
The policy should reduce waiting times, provided that staffing levels remain stable.
should 是有条件的预期；provided that 明确预期依赖人员稳定。
A larger sample would not necessarily eliminate the bias.
would 进入假设情境，not necessarily 只否定必然性，并不声称扩大样本毫无作用。
The association could indicate a causal effect, but it could also arise from selection. If more motivated students are both more likely to attend tutoring and more likely to improve, the observed difference would overstate the program's contribution. A randomized design might reduce this ambiguity, although imperfect compliance could still complicate interpretation. The evidence therefore supports a possibility, not a certainty.
段落连续使用 could, would, might 建立多条可能路径。作者不是犹豫不决，而是在区分观察到的事实与尚需识别的因果解释。
训练语境：学术科研
The result may be driven by attrition 承诺了什么？
它只把样本流失列为可能机制，没有断言流失已经得到证明。严谨复述应保留“可能”，不能改写成“结果由流失导致”。
训练语境：学术科研
解释：The pattern cannot be explained by chance alone.
cannot 表达强认识判断：仅靠偶然不足以解释。alone 很关键，因为句子没有排除偶然因素参与，只排除它作为完整解释。
训练语境：学术科研
比较 must submit 与 must have submitted。
must submit 通常规定未来或一般义务；must have submitted 通常根据现有线索推断过去已经提交。完成形式把事件放在推断时刻之前。
训练语境：学术科研
分析反事实：Had the control group been larger, the interval would have been narrower.
倒装的 Had... 等于 If ... had been，表示控制组实际上没有更大。would have been 给出未实现条件下的过去结果。
训练语境：学术科研
should 是否表示道德义务：The estimator should converge as the sample grows.
这里表示基于理论的预期，不是道德要求。是否收敛仍受模型条件约束，因此不能把 should 读成无条件事实。
训练语境：学术科研
解释否定范围：The intervention may not improve every outcome.
最自然解释是“可能并非改善每个结果”，否定落在 improve every outcome 上。它不等于 may not 表示禁止，也不等于所有结果都不会改善。
训练语境：真实生活
You must be tired 是命令吗？
通常不是。说话者从旅程、工作时间等证据推断对方疲惫，是认识情态；语调和情境决定这一解释。
训练语境：真实生活
比较 If it rains, we will stay home 与 If it rained, we would stay home。
第一句把下雨视为现实开放条件；第二句把它放得更远、更假设化，可能暗示当前不太可能下雨。差别是说话者对条件世界的距离判断。
训练语境：真实生活
解释：You could have called me.
完成情态指向过去未实现的可能行动，常带有责备或遗憾。它不是单纯说明过去具备打电话的能力。
训练语境：真实生活
That might not be necessary 与 That must not be necessary 是否等价？
不等价。前者表示“也许没有必要”；后者若作认识解释非常生硬，若作义务解释又接近“不得成为必要”。自然表达强推断“不可能有必要”通常用 cannot be necessary。
The policy might appear successful if success is defined only as a short-term increase in enrollment. Yet a different conclusion would follow if the analysis asked whether the students who enrolled under the policy could have achieved the same outcomes without the subsidy. The counterfactual is difficult to observe directly, so the paper treats matched districts as a proxy rather than as proof. This distinction matters because a policy can be associated with an outcome without having caused it.
might 降低断言强度，并把成功限定在一个定义里。would follow if... 建立反事实推理路径。might appear successful 不是说政策一定成功，而是“在某定义下看起来成功”。would follow if... 表示结论依赖一个未实际采取的分析问题。could have achieved 指向过去可能性：没有补贴时是否也能达到同样结果。rather than as proof 把证据地位压低：matched districts 是 proxy，不是 proof。高阶语境：学术科研
找出本段中最关键的反事实问题，并说明它为什么比“入学率是否提高”更深。
关键问题是 whether the students who enrolled under the policy could have achieved the same outcomes without the subsidy。它问的是没有补贴时结果是否仍会发生，直接对应因果判断；“入学率是否提高”只描述政策后发生了什么，不能排除其他原因。
高阶语境：学术科研
解释 might appear, would follow, could have achieved 三个情态结构的功能差异。
might appear 表示谨慎判断；would follow 表示在某条件成立时会推出某结论；could have achieved 表示过去未观测情形中的可能性。三者分别承担“弱断言、条件推论、反事实可能性”。
高阶语境：真实生活
朋友说“我喝了咖啡所以考试考好了”。请用英文写一句反事实追问。
可以写：Would you have performed just as well if you had not drunk the coffee? 这句话不是否定咖啡可能有用，而是要求比较“实际发生”和“没有咖啡时可能发生”的结果。
读到情态词时，依次问：它评估真假还是规定行动？强度多大？是否受条件限制？否定落在情态词还是命题上？