An iterative method is described for least curve fitting, when both the abscissae and ordinates are subject to error. The ordinate and abscissa are fitted by parametric functions, using a constructive non-linear optimisation procedure. The order of the data points is maintained, and hence multi-valued data may be fitted.