A combination of fuzzy c-means clustering (FCM) and non-linear mapping (NLM) has been applied to a multi-parameter set of rock-magnetic data from the Pleistocene section of ODP Hole 904A in order to understand postdepositional alteration of the natural remanent magnetization (NRM) in marine sediments. On a per sample basis a 6 cluster model is coherent with a previous non-statistical study which interprets the natural remanent magnetization as a function post-depositional alteration of the magnetomineralogy. FCM permits recognition of intermediate cases between clusters which is important as diagenetic effects on the magnetic signal are inherently gradual. Critically, the multivariate analysis detects groups in the data set which may only be recognized by the consideration of all magnetic information simultaneously. A thorough consideration of the rock magnetic input parameters and their individual meaning for a specific site (sediment) is, however, imperative. We show that FCM and NLM greatly facilitate significant grouping of a rock magnetic data set to untangle diagenetic alterations of the NRM. Furthermore, this technique potentially provides additional information which may pass unnoticed with the application of "standard" rock-magnetic univariate and bivariate procedures.