Kinetic patterns of glucose-stimulated insulin secretion from the in vitro perfused pancreas were used to test different types of secretion models of similar complexity. A storage-limited, two-compartment model, modified slightly from that previously, was compared with signal-limited models incorporating delta or feedback characteristics. Mathematical relationships for all models were fixed by single-step, dose-response experiments and models were compared in a series of glucose test patterns including steps, step restimulations, staircases, pulses, ramps, and ramp restimulations. The work quantifies previously unreported characteristics of hypersensitivity and low-glucose rest/restimulation behavior in the pancreas. All models simulated staircase and ramp experiments. The two-compartment model contains an inherent hypersensitivity factor required for repeated pulse-type experiments. However, the kinetics of hypersensitivity were too rapid to be simulated in all types of pulse and ramp/pulse experiments by the restricted refilling characteristics as written into this model. The signal-limited, delta-feedback model did not inherently produce potentiation, but required add-on modification that then more closely simulated some pulse and ramp/pulse experiments. This model simulated experimental negative spikes, whereas the storage-limited model would require additional complexity to do so. These and other results suggest that both storage- and signal-limited models, although currently insufficient, could be elaborated to simulate available data. Therefore, a choice between the two to describe the underlying physiological mechanism of multiphasic insulin secretion is premature. The alternate possibility that the secretion mechanism may be reflected by a combination of the two models is presented.