We study the sensitivity to physics beyond the standard model of precise top-quark pair production measurements at future lepton colliders. A global effectivefield-theory approach is employed, including all dimension-six operators of the Warsaw basis which involve a top-quark and give rise to tree-level amplitudes that interfere with standardmodel e + e − → tt → bW +b W − ones in the limit of vanishing b-quark mass. Four-fermion and CP-violating contributions are taken into account. Circular-collider-, ILC-and CLIClike benchmark run scenarios are examined. We compare the constraining power of various observables to a set of statistically optimal ones which maximally exploit the information contained in the fully differential bW +b W − distribution. The enhanced sensitivity gained on the linear contributions of dimension-six operators leads to bounds that are insensitive to quadratic ones. Even with statistically optimal observables, two centre-of-mass energies are required for constraining simultaneously two-and four-fermion operators. The impact of the centre-of-mass energy lever arm is discussed, that of beam polarization as well. A realistic estimate of the precision that can be achieved in ILC-and CLIC-like operating scenarios yields individual limits on the electroweak couplings of the top quark that are one to three orders of magnitude better than constraints set with Tevatron and LHC run I data, and three to two hundred times better than the most optimistic projections made for the high-luminosity phase of the LHC. Clean global constraints can moreover be obtained at lepton colliders, robustly covering the multidimensional effective-field-theory space with minimal model dependence.