Three classes of automated tuning for PID controllers are discussed and compared. These are a Ziegler-Nichols-based autotuner, a waveform-analysis adaptive tuner and a leastsquares identification-based self-tuner. The results of a program of experiments to evaluate the relative performance of the three approaches are presented, together with a discussion of their relative merits in the light of these experiments. Two representative industrial temperature-controlled processes are used in these experiments: a plastic-onwire multizone extruder barrel and a laboratory furnace. Between them, these processes display typical characteristics such as nonlinearity and loop coupling.