Relaxed clock methods account for among-branch-rate-variation when estimating divergence times by inferring different rates for individual branches. In order to infer different rates for individual branches, important assumptions are required. This is because molecular sequence data do not provide direct information about rates but instead provide direct information about the total number of substitutions along any branch, which is a product of the rate and time for that branch. Often, the assumptions required for estimating rates for individual branches depend heavily on the implementation of multiple fossil calibrations in a single phylogeny. Here, we show that the basis of these assumptions is often critically undermined. First, we highlight that the temporal distribution of the fossil record often violates key assumptions of methods that use multiple fossil calibrations with relaxed clocks. With respect to "node calibration" methods, this conclusion is based on our inference that different fossil calibrations are unlikely to reflect the relative ages of different clades. With respect to the fossilized birth-death process, this conclusion is based on our inference that the fossil recovery rate is often highly heterogeneous. We then demonstrate that methods of divergence time estimation that use multiple fossil calibrations are highly sensitive to assumptions about the fossil record and among-branch-rate-variation. Given the problems associated with these assumptions, our results highlight that using multiple fossil calibrations with relaxed clocks often does little to improve the accuracy of divergence time estimates.

中文翻译：

---

来自对使用多个化石校准和松弛时钟进行发散时间估计的经验分析和仿真得出的见解。

当通过为单个分支推断不同的速率来估计发散时间时，宽松的时钟方法会考虑分支速率之间的差异。为了推断各个分支机构的不同费率，需要进行重要的假设。这是因为分子序列数据不提供有关速率的直接信息，而是提供有关沿任何分支的取代总数的直接信息，这是该分支的速率和时间的乘积。通常，估计单个分支的速率所需的假设在很大程度上取决于在单个系统发育中执行多个化石校准的情况。在这里，我们表明，这些假设的基础经常被严重破坏。第一，我们强调指出，化石记录的时间分布通常违反使用多个带有宽松时钟的化石校准方法的关键假设。关于“节点校准”方法，该结论基于我们的推论，即不同的化石校准不太可能反映不同进化枝的相对年龄。关于化石的生死过程，这一结论是基于我们的推论，即化石的回收率通常是高度异质的。然后，我们证明了使用多个化石校准的发散时间估计方法对关于化石记录和分支速率变化的假设高度敏感。鉴于与这些假设相关的问题，