Deep brain stimulation (DBS) is an effective symptomatic treatment in Parkinson's disease. High frequency stimulation (HFS) of the subthalamic nucleus elicits neurotransmitter release in multiple nuclei. Therefore, we tested the hypothesis that neurotransmitter release during HFS may be used to provide feedback control of the intensity and pattern of HFS. We studied the dynamic relationship between extracellular glutamate levels and HFS in and around the STN in anesthetized rats. We used a pseudorandom binary sequence (PRBS) of stimulation in the STN, the independent forcing function, while measuring extracellular glutamate in the same nucleus, the dependent variable. The PRBS consisted of 90 s periods during which stimulation (100 microA, 150Hz, 10% duty cycle) was either off or on. The stimulation and extracellular glutamate levels were fitted using an autoregressive exogenous model (ARX) to determine the transfer function between HFS and the extracellular glutamate concentration in the STN. The ARX model fit the dynamics of extracellular glutamate levels well (correlation coefficients ranged from 0.74 to 0.99; n=11). The transfer function accurately predicted extracellular glutamate levels in the STN even when the pattern of HFS was modified. We used the transfer function to develop a feedback controlled stimulation algorithm. Feedback controlled HFS maintained extracellular glutamate concentrations at any predefined level, but only intermittent HFS was required. We conclude that the transfer function between HFS and neurotransmitter levels in the brain can be used to design DBS protocols that generate specific temporal patterns of glutamate release in the STN.