A new technique is proposed for analyzing neutron noise signals to identify two-phase flow patterns in an experimental reactor. Based on the modeling of neutron signals by autoregressive moving average models via an optimization technique, two-phase flow patterns can be identified by a single index known as the “dynamic signature.” A computer code is set up in an IBM-PC/XT microcomputer by which ∼90% of the experimental cases have been successfully identified among 86 data sets. This technique is recommended for the analysis of boiling water reactor neutron signals to generate a data base for designing a core flow monitoring system. This technique would be useful in understanding thermohydraulic phenomena in an operating power reactor.