This paper presents a recurrent fuzzy neuron (RFN) which facilitates nonlinear mapping from an input space to an output space. The synaptic junctions are characterized by a set of IF-THEN rules and recurrent characteristics provide dynamic properties to the neuron, allowing its application to on line modelling for a variety of nonlinear systems. The effectiveness of this neuron to synthesize complex nonlinear models, is illustrated by simulation results related to on line prediction of chaotic behavior and modelling of time varying nonlinear systems.