There have been several methods proposed in the literature for tuning feedback controllers in openloop unstable processes. All of these methods assume a transfer function model is available that consists of a steady-state gain, deadtime, and positive pole. In some methods an additional negative pole is considered. This paper points out that this type of transfer function is often not appropriate for tuning temperature controllers in openloop unstable chemical reactors. There are two reasons for this. First, many chemical reactors used in industry have more than one positive pole. For example, the commonly used jacket-cooled continuous stirred-tank reader has two positive poles even for the simple reaction A --> B. Second, in practical applications we may not have a model, just an operating reactor. Even if a model is available, it is usually much more complex than the simple transfer function used in the literature. Openloop identification is not possible because the process is openloop unstable. Usually the only information available is the ultimate gain and the ultimate frequency, which have been obtained from a closedloop relay-feedback test. In this paper the effect of conversion on openloop instability is explored for jacket-cooled continuous stirred-tank reactor systems. Both linear and nonlinear models are used to study controller tuning. The Tyreus-Luyben (TL) tuning method is demonstrated to give more robust control for this type of system than the Ziegler-Nichols method. A procedure for obtaining an approximate gain/deadtime/positive-pole model is presented so that the tuning methods proposed in the literature can be used. However, these tuning methods provide performance that is no better than the TL method.