This paper comments on selected aspects of identification issues of DSGE models. It suggests the singular value decomposition (SVD) as a useful tool for detecting local weak and non-identification. This decomposition is useful for checking rank conditions of identification, identification strength, and it also offers parameter space ‘identification patterns’. With respect to other methods of identification the singular value decomposition is particularly easy to apply and offers an intuitive interpretation. We suggest a simple algorithm for analyzing identification and an algorithm for finding a set of the most identifiable set of parameters. We also demonstrate that the use of bivariate and multiple correlation coefficients of parameters provides only limited check of identification problems.