When estimate the factor scores in SEMs by AMOS, I find a phenomenon that the factor score weights calculated by regression method always sum up to one, when the SEM model includes only one factor. And when a model include two or more factors, any factor's score weights sum up to one too, if ignore weights of other observable variables which are no belong to this factor.

 I don't know why. There seems no any constraint in the process of calculating score weights by regression method that can promise the sum of weights is one.