An experimental technique has been developed to identify the hierarchical importance of characteristic features in the complex sounds to which an auditory neuron is most selective. Single‐unit spike rates are measured in response to variation of component intensities and frequencies, using a subset of single and paired tone stimuli. The response profiles are used to rank the importance of nonlinear interactions between component frequencies as well as component intensity levels, and a multidimensional matrix is derived that enables prediction of the neuron's response to sounds in a more general, complex set. The derivation is based on weighted inner product space optimization (WIPSO) and singular valne decomposition (SVD) analysis. To test this general method, the activity of single auditory fibers in the eighth nerve of the green treefrog (Hyla cinerea) was recorded in response to computer‐synthesized stimuli. Results indicate that this is a very promising, powerful approach for characterizing the selectivity of individual auditory neurons to synthetic and natural sounds of interest. [Work supported by NINeDS.]