Anywhere after the covariance or mean statement, the line
will request, after optimization, the computation of the covariance matrix of the parameters by numerical methods. Although not strictly accurate, this usually gives a fair impression of model identification through the ratio of the largest to smallest eigenvalues of the covariance matrix. Examination of the eigenvectors is used to detect which two parameters are most likely to be causing the underidentification problem. Note that spurious messages are possible, and that parameters identified in one observed covariance matrix may not be identified in another, even though the model and design are exactly the same.
Mike Neale, email@example.com, Medical College of VA