The invention relates to a method for identifying one or more spectral features in a spectrum (4, 5) of a sample for a constituent analysis of the sample, comprising providing the spectrum (4, 5), predefining an approximation function (6), which is a continuously differentiable mathematical function, respectively forming an (n−1)-th order derivative (7, 8, 9) of the spectrum (4, 5) and of the approximation function (6), wherein the number n>1, generating a correlation matrix (10) from the two (n−1)-th order derivatives (7, 8, 9), and respectively identifying the spectral feature or one of the spectral features in each case as a function of a local extremum (i) of the correlation matrix (10) for at least one extremum (i) of the correlation matrix (10) in order to simplify the constituent analysis of the sample.