Fitting correlated data

We discuss fitting correlated data, with the example of hadron mass spectroscopy in mind. The main conclusion is that the method of minimizing the correlated ${\chi }^2$ is unreliable if the data sample is too small.