Finite-rotation parameters for 17 isochrons from anomalies 1-4A for the Juan de Fuca-Pacific plate pair are reported. Comparison of confidence intervals for plate rotations derived by different techniques indicates that a single-plate misfit criterion significantly underestimates the confidence region. Defining the confidence interval by a threshold level of misfit determined by an F statistic gives similar results to the technique of bootstrap resampling for internally consistent data. Bootstrap resampling offers far greater robustness in the case of inconsistent data. Precision of about 1 km in the direction normal to isochrons is available for the younger anomalies. This precision allows confident recognition of 2--5 km of eastward deformation of the Pacific plate, 50--100 km south of the active Sovanco transform. Slight deformation of the northern Juan de Fuca plate adjacent to the Nootka fault zone is also apparent. East of the northern Gorda ridge, younger (<3 Ma) anomalies do not show measurable deformation near the Blanco fracture zone, but older anomalies are up to 30 km east of their position predicted by rigid behavior of the Juan de Fuca plate. Reconstruction of the overlap geometry of past propagators shows no evidence that the extend of overlap has ever exceeded 70 km, even with offset of 75 km in one case and 140 km in another. New techniques for testing for intervals of constant plate motion require at least two changes in instantaneous motion at 99% confidence: a clockwise change in motion direction at about 5 Ma and an increase in the gradient of spreading rate since 2 Ma. Two other changes in gradient since 5 Ma are required at 95% confidence. ¿ American Geophysical Union 1993