Wide species crosses often result in uniparental genome elimination and visible failures in centromere function.Crosses involving lines with mutated forms of the CENH3 histone variant that organizes the centromere/kinetochore interface have been shown to have similar effects,inducing haploids at high frequencies.Here,we propose a simple centromere size model that endeavors to explain both observations.It is based on the idea of a quantitative centromere architecture where each centromere in an individual is the same size,and the average size is dictated by a natural equilibrium between bound and unbound CENH3 (and its chaperones or binding proteins).While centromere size is determined by the cellular milieu,centromere positions are heritable and defined by the interactions of a small set of proteins that bind to both DNA and CENH3.Lines with defective or mutated CENH3 have a lower loading capacity and support smaller centromeres.In cases where a line with small or defective centromeres is crossed to a line with larger or normal centromeres,the smaller/defective centromeres are selectively degraded or not maintained,resulting in chromosome loss from the small-centromere parent.The model is testable and generalizable,and helps to explain the counterintuitive observation that inducer lines do not induce haploids when crossed to themselves.