Two approaches for noise source identification based on theory for multiple‐input systems have been investigated. The concepts of the frequency response function and the coherent residual spectral density function were used to estimate the spectra of the noise sources that were accounted for in the multiple‐input model. The key factors which determine the applicability of the techniques for noise source identification are found to be the degrees of coherence between themeasuredinputs. The experimental results showed that, for a noise system with multiple sources which were slightly coherent to each other, the techniques presented might be able to estimate the spectra of the noise sources, provided there was only modest measurement contamination.