AbstractMitotic chromosome segregation is a self-organising process that achieves high fidelity separation of 46 duplicated chromosomes into two daughter cells. Chromosomes must be captured by the microtubule-based spindle, aligned at the spindle equator where they undergo oscillatory motion (metaphase) and then pulled to opposite spindle poles (anaphase). These large and small-scale chromosome movements are driven by kinetochores, multi-protein machines, that link chromosomes to microtubules and generate directional forces. Through automated near-complete tracking of kinetochores at fine spatio-temporal resolution over long timescales, we produce a detailed atlas of kinetochore dynamics throughout metaphase and anaphase in human cells. We develop a hierarchical biophysical model of kinetochore dynamics and fit this model to 4D lattice light sheet experimental data using Bayesian inference. We demonstrate that location in the metaphase plate is the largest factor in the variation in kinetochore dynamics, exceeding the variation between cells, whilst within the spindle there is local spatio-temporal coordination between neighbouring kinetochores of directional switching events, kinetochore-fibre (K-fibre) polymerization/depolymerization state and the segregation of chromosomes. Thus, metaphase oscillations are robust to variation in the mechanical forces throughout the spindle, whilst the spindle environment couples kinetochore dynamics across the plate. Our methods provide a framework for detailed quantification of chromosome dynamics during mitosis in human cells.