When discourse markers (DMs) occur in sequences of multiple markers, their linear order is generally quite rigid. Still, there are clear cases of reversible DM sequences. This article examines the reasons for DM reversibility and explores its implications for theories of DM sequencing. A corpus analysis of the English DMs so and and suggests that one reason for their reversibility is that the different orders achieve different effects. When so follows and (And so…) it more often marks the upcoming utterance as a result or conclusion (‘therefore’). When so precedes and (So and…) it is more often used to change the discourse topic. Crucially, both of these functions are part of so's known functional spectrum. This shows that functional accounts of DM syntax must be sensitive to DM polyfunctionality. Moreover, a close analysis of each marker's individual contribution to their cumulative effect as a sequence points to a general ordering principle: In DM sequences, the earlier DM constrains the interpretation of the later DM by providing a particular interpretive context for it. This explains why alternative sequence positions may go along with different functions. In this way, the temporality of spoken language acts as a general constraint on DM sequencing.

当话语标记 (DM) 出现在多个标记的序列中时，它们的线性顺序通常非常严格。尽管如此，仍有明显的可逆 DM 序列案例。本文探讨了 DM 可逆性的原因，并探讨了其对 DM 测序理论的影响。对英语 DM so和and的语料库分析表明，它们可逆性的一个原因是不同的顺序会产生不同的效果。当如此跟随并且（所以…… ）时，它更经常将即将到来的话语标记为结果或结论（“因此”）。当so出现在and ( So and...) 它更常用于改变话语主题。至关重要的是，这两个功能都是so已知功能谱的一部分。这表明 DM 语法的功能性帐户必须对 DM 多功能性敏感。此外，仔细分析每个标记对其作为序列的累积效应的个体贡献指向一般排序原则：在 DM 序列中，较早的 DM 通过为其提供特定的解释上下文来限制对后期 DM 的解释。这解释了为什么替代序列位置可能伴随不同的功能。通过这种方式，口语的时间性作为 DM 排序的一般约束。