Periodic Reporting for period 4 - ROCKS-PARADOX (Dissecting the paradox of stasis in evolutionary biology)
Période du rapport: 2025-07-01 au 2025-12-31
One of biology's enduring puzzles is the paradox of stasis: most species appear morphologically unchanged in the fossil record across millions of years, yet living populations can evolve rapidly in response to environmental pressures. Why does evolution seem to slow over geological timescales? This is the central question that the ROCKS-PARADOX project set out to address.
Understanding how evolution operates across different timescales touches on fundamental questions about the predictability and pace of biological change, knowledge that is increasingly important in an era of rapid environmental transformation. If we can identify what governs rates of evolutionary change across timescales, we gain deeper insight into the resilience and adaptability of life in response to environmental pressures.
The ROCKS-PARADOX project pursued four interconnected objectives: (1) developing a comprehensive statistical framework for analysing evolutionary change within lineages across timescales; (2) applying this framework to the world's largest collection of fossil time series data to characterise patterns of evolutionary change; (3) assessing how genetic constraints and evolvability shape evolution beyond short-term microevolutionary timescales; and (4) pioneering new methods for high-throughput data collection from the fossil record using artificial intelligence.
Over the course of the project, these objectives were met. The project produced a statistical modelling framework – evoTS – that is now the most comprehensive tool available for studying evolutionary change in time series data. A landmark study published in the journal Science demonstrated that a key microevolutionary parameter – genetic variance – estimated from fossil samples predicts evolutionary divergence across millions of years, confirming a central hypothesis of the project. Multiple further publications advanced our understanding of evolutionary rates, stasis, and the causal drivers of change across timescales. The project also pioneered the use of AI-based phenomics pipelines to extract morphological data from fossils at unprecedented scale and speed.
Understanding how evolution operates across different timescales touches on fundamental questions about the predictability and pace of biological change, knowledge that is increasingly important in an era of rapid environmental transformation. If we can identify what governs rates of evolutionary change across timescales, we gain deeper insight into the resilience and adaptability of life in response to environmental pressures.
The ROCKS-PARADOX project pursued four interconnected objectives: (1) developing a comprehensive statistical framework for analysing evolutionary change within lineages across timescales; (2) applying this framework to the world's largest collection of fossil time series data to characterise patterns of evolutionary change; (3) assessing how genetic constraints and evolvability shape evolution beyond short-term microevolutionary timescales; and (4) pioneering new methods for high-throughput data collection from the fossil record using artificial intelligence.
Over the course of the project, these objectives were met. The project produced a statistical modelling framework – evoTS – that is now the most comprehensive tool available for studying evolutionary change in time series data. A landmark study published in the journal Science demonstrated that a key microevolutionary parameter – genetic variance – estimated from fossil samples predicts evolutionary divergence across millions of years, confirming a central hypothesis of the project. Multiple further publications advanced our understanding of evolutionary rates, stasis, and the causal drivers of change across timescales. The project also pioneered the use of AI-based phenomics pipelines to extract morphological data from fossils at unprecedented scale and speed.
The ROCKS-PARADOX project was organised around six work packages (WPs) spanning model development, empirical data analysis, and dissemination.
Modelling framework (WP1): evoTS, an R package implementing a comprehensive suite of models for analysing evolutionary change within lineages using time series data, is now the most complete and flexible modelling framework available for this purpose. It enables researchers to compare competing models of evolutionary dynamics – including random walks, directional change, stasis, and multi-phase models – and to draw causal inferences from fossil and contemporary data.
Model adequacy and the fossil record (WP2): New tests of model adequacy assess not only which model fits best but whether any model fits adequately. Applied to the largest fossil time series dataset ever assembled, results indicate the fossil record contains more evolutionary change than traditionally recognised.
Rates of evolution (WP3): A published study demonstrated that the negative correlation between evolutionary rates and measurement timescale is a genuine biological signal, combining absolute and relative model fit for the first time when investigating this pattern.
Explaining stasis (WP4): A submitted manuscript systematically evaluates alternative explanations for stasis in phenotypic time series, representing the first in-depth application of evoTS across a wide range of timescales.
Evolvability in the fossil record (WP5 & WP6): A peer-reviewed manuscript examines how evolvability affects within-lineage evolution in the fossil record; a second manuscript is in preparation. These are the first studies to analyse how genetic variance-covariance structures change through time in the fossil record. They also introduce the steginator: an AI-based phenomics pipeline automating morphological measurement from fossil images with high accuracy and throughput.
Large-scale analysis and synthesis: The PI is finalising two manuscripts: one presenting quantitative results on patterns of evolutionary change, and a perspective paper synthesising insights on within-lineage evolution and implications for bridging microevolution and macroevolution.
Dissemination: Results have been presented at international conferences and published in high-impact journals, including Science. The evoTS package is publicly available, and teaching activities have extended the project's reach to students and early-career researchers internationally.
Modelling framework (WP1): evoTS, an R package implementing a comprehensive suite of models for analysing evolutionary change within lineages using time series data, is now the most complete and flexible modelling framework available for this purpose. It enables researchers to compare competing models of evolutionary dynamics – including random walks, directional change, stasis, and multi-phase models – and to draw causal inferences from fossil and contemporary data.
Model adequacy and the fossil record (WP2): New tests of model adequacy assess not only which model fits best but whether any model fits adequately. Applied to the largest fossil time series dataset ever assembled, results indicate the fossil record contains more evolutionary change than traditionally recognised.
Rates of evolution (WP3): A published study demonstrated that the negative correlation between evolutionary rates and measurement timescale is a genuine biological signal, combining absolute and relative model fit for the first time when investigating this pattern.
Explaining stasis (WP4): A submitted manuscript systematically evaluates alternative explanations for stasis in phenotypic time series, representing the first in-depth application of evoTS across a wide range of timescales.
Evolvability in the fossil record (WP5 & WP6): A peer-reviewed manuscript examines how evolvability affects within-lineage evolution in the fossil record; a second manuscript is in preparation. These are the first studies to analyse how genetic variance-covariance structures change through time in the fossil record. They also introduce the steginator: an AI-based phenomics pipeline automating morphological measurement from fossil images with high accuracy and throughput.
Large-scale analysis and synthesis: The PI is finalising two manuscripts: one presenting quantitative results on patterns of evolutionary change, and a perspective paper synthesising insights on within-lineage evolution and implications for bridging microevolution and macroevolution.
Dissemination: Results have been presented at international conferences and published in high-impact journals, including Science. The evoTS package is publicly available, and teaching activities have extended the project's reach to students and early-career researchers internationally.
The ROCKS-PARADOX project has advanced the state of the art on multiple fronts.
First, the evoTS modelling framework is the most developed and flexible toolset available for studying within-lineage evolutionary change using time series data. By implementing both new and established models in a single coherent R package, the project has made rigorous evolutionary analysis accessible to the broader research community.
Second, the project introduced novel tests of model adequacy that shift attention from which model fits best to whether any model fits adequately – a conceptually important distinction that had been largely missing from the field. These tests provide a safeguard against drawing firm conclusions from models that do not, in fact, describe the data well.
Third, the landmark publication in Science (Holstad et al. 2024) demonstrated for the first time that genetic variance estimated from fossil samples predicts phenotypic divergence across millions of years of evolution. This confirms that a central parameter of microevolutionary theory is relevant at macroevolutionary timescales, directly addresses the paradox of stasis at the heart of the project, and opens entirely new avenues for using fossil data to inform evolutionary theory.
Fourth, the AI-based phenomics pipeline (the steginator) breaks new ground in palaeobiology by enabling automated, high-throughput collection of morphological data from fossil images. This technology dramatically reduces the time and cost of data collection, making possible future studies at scales that were previously impractical.
Fifth, by applying evoTS to the largest fossil time series dataset ever assembled, the project has provided the most comprehensive empirical characterisation of within-lineage evolutionary patterns in the fossil record. Results suggest it contains considerably more evolutionary change than decades of literature have recognised – a finding with broad implications for how we understand and model long-term evolution.
Taken together, the ROCKS-PARADOX project has established a new methodological and conceptual foundation for studying evolution across timescales. The tools and insights developed are expected to drive advances in evolutionary biology and palaeontology for years to come, and the research group has emerged as a leading centre for this line of enquiry, attracting students and collaborators from across the world.
First, the evoTS modelling framework is the most developed and flexible toolset available for studying within-lineage evolutionary change using time series data. By implementing both new and established models in a single coherent R package, the project has made rigorous evolutionary analysis accessible to the broader research community.
Second, the project introduced novel tests of model adequacy that shift attention from which model fits best to whether any model fits adequately – a conceptually important distinction that had been largely missing from the field. These tests provide a safeguard against drawing firm conclusions from models that do not, in fact, describe the data well.
Third, the landmark publication in Science (Holstad et al. 2024) demonstrated for the first time that genetic variance estimated from fossil samples predicts phenotypic divergence across millions of years of evolution. This confirms that a central parameter of microevolutionary theory is relevant at macroevolutionary timescales, directly addresses the paradox of stasis at the heart of the project, and opens entirely new avenues for using fossil data to inform evolutionary theory.
Fourth, the AI-based phenomics pipeline (the steginator) breaks new ground in palaeobiology by enabling automated, high-throughput collection of morphological data from fossil images. This technology dramatically reduces the time and cost of data collection, making possible future studies at scales that were previously impractical.
Fifth, by applying evoTS to the largest fossil time series dataset ever assembled, the project has provided the most comprehensive empirical characterisation of within-lineage evolutionary patterns in the fossil record. Results suggest it contains considerably more evolutionary change than decades of literature have recognised – a finding with broad implications for how we understand and model long-term evolution.
Taken together, the ROCKS-PARADOX project has established a new methodological and conceptual foundation for studying evolution across timescales. The tools and insights developed are expected to drive advances in evolutionary biology and palaeontology for years to come, and the research group has emerged as a leading centre for this line of enquiry, attracting students and collaborators from across the world.