In this paper, a memristor-based Pavlov associative memory circuit accorded with the biological feature is proposed. The proposed circuit mainly consists of neuron module, synapse module with synaptic weight measurement function and feedback module. Variety of learning and forgetting functions are considered. The condition of learning is that the food neuron and ring neuron are activated simultaneously. There are three types of forgetting conditions, which represent different forms of forgetting, respectively. The first (second) type of forgetting condition is that only food (ring) neuron is activated. The third type of forgetting condition is that neither food neuron nor ring neuron is activated. In addition, the rates of learning and forgetting are variable. After loops of learning and forgetting, further learning will be more skillful. In other words, the rate of learning a forgotten knowledge will be faster than before. This bionic feature is similar to human. It is implemented by changing learning and forgetting voltages by adding the feedback voltage from feedback module to them. Functions mentioned above are demonstrated by simulation results in PSPICE software.