Variability and scale-dependence in molecular evolution rate impacts interpretation of eukaryotic evolutionary histories

Molecular evolution is often modeled as proceeding at consistent rates over time, with some deviations accommodated by relaxed molecular clock models. Here, we quantify the full extent of variability in branch-specific substitution rates across relatively well-calibrated eukaryotic phylogenies, confirming that punctuated change at the molecular level underlies evolution at the morphological level where punctuated dynamics are more commonly recognized. We also show how inferred substitution rates decrease systematically when measured across increasing time intervals. This scale-dependence persists across alternative clock models, calibration strategies, and prior assumptions, but disappears in simulated data evolved under a constant rate - suggesting that the phenomenon arises from time-varying substitution rates and reflects genuine properties of evolutionary histories rather than model artifacts. The observed pattern is analogous to the Sadler effect in sedimentary geology, where time-averaged rates decline with increasing measurement interval because sedimentation is episodic, with longer hiatuses occurring less frequently. The recognized scale-dependent bias in molecular evolution is not captured in current molecular clock models and significantly impacts inferences of evolutionary history, such as estimating the age of Metazoa and understanding the timing and nature of the Cambrian Explosion. Significance StatementRates of molecular evolution are central to reconstructing the history of life, yet they are often assumed to be approximately constant over sufficiently long timescales. By analyzing relatively well-dated evolutionary trees of eukaryotes, we show how inferred rates of genome change systematically decrease as the timescale of measurement increases. This pattern is analogous to a well-known phenomenon in sedimentary geology where apparent sedimentation rates decline over longer intervals due to episodic processes. Our results demonstrate how long-term variability in evolutionary rates similarly produces a significant scale-dependent bias which is overlooked in current evolutionary models. Recognizing and quantifying this effect is important for dating key evolutionary events, such as the origin of animals, and for understanding the cadence of evolutionary processes.