Typically, nonlinear models such as those used in the analysis of covariance structures, are not globally identifiable. Investigations of local identifiability must either yield a mapping onto the entire parameter space, or be confined to points of special interest such as the maximum likelihood point. (Author/JKS)