Deconvolution analysis of hormone data poses special problems in view of the sparse, noisy, and short data series typically available for analysis; the unknown true nature of the underlying secretory event; and potentially large variations in dissipation or clearance kinetics in different settings. Consequently, deconvolution techniques, which concern themselves with the estimation of hormone secretion and/or clearance based on serial circulating hormone concentration measurements, face a particular challenge. Ideal features of deconvolution algorithms are summarized in Table IV. Specific deconvolution techniques available to analyze hormone data include both waveform-defined procedures and waveform-independent algorithms. These approaches should be viewed as complementary rather than antagonistic. All deconvolution techniques are subject to individual limitations and specific strengths. Independently of the method employed, error propagation is necessary so as to define the statistical uncertainty intrinsic to the estimate of secretion and clearance. Such calculations of experimental uncertainty should include error inherent in the sample collection, processing, and assay as well as error in the kinetic constants and/or anticipated departures of the biological process from the algebraic structure of the convolution formulation. Moreover, more complex convolution statements will be required to describe the full range of behavior of hormone data in a systems view. The applications of such newer convolution methods as well as currently available techniques include model synthesis, model testing, and analysis of the interactions among multiple pulse generators.