An experimental study of the kinetics of the initial stages of the pyrolysis of high-density polyethylene (PE) was performed. Quantitative yields of gas-phase products (C-1-C-8 alkanes and alkenes) and functional groups within the remaining polyethylene melt (methyl, vinyl, vinylene, vinylidene, and branching sites) were obtained as a function of time (0-20 min) at five temperatures in the 400-440 degrees C range. Gas chromatography and NMR (H-1 and C-13) were used to detect the gas- and condensed-phase products, respectively. Modeling of polyethylene pyrolysis was performed, with the primary purpose of determining the rate constants of several critical reaction types important at the initial pyrolysis stages. Detailed chemical mechanisms were created (short and extended mechanisms) and used with both the steady-state approximation and numerical integration of the differential kinetic equations. Rate constants of critical elementary reactions (C-C backbone scission, two kinds of H-atom transfer, radical addition to the double bond, and beta-scission of tertiary alkyl radicals) were adjusted, resulting in an agreement between the model and the experiment. The values of adjusted rate constants are in general agreement with those of cognate reactions of small molecules in the gas phase, with the exception of the rate constants of the backbone C-C scission, which is found to be approximately 1-2 orders of magnitude lower. This observation provides tentative support to the hypothesis that congested PE melt molecular environment impedes the tumbling motions of separating fragments in C-C bond scission, thus resulting in less "loose" transition state and lower rate constant values. Sensitivity of the calculations to selected uncertainties in model properties was studied. Values and estimated uncertainties of four combinations of rate constants are reported as derived from the experimental results via modeling. The dependence of the diffusion-limited rate constant for radical recombination on the changing molecular mass of polyethylene was explicitly quantified and included in the extended kinetic mechanism, which appears critical for the agreement between modeling and experiment, particularly the agreement between the experimental and the calculated activation energies for product formation rates. Calculations were performed to estimate the contribution to the overall rate of radical recombination of the "reaction diffusion" phenomenon, where recombination is driven not by the actual motion of the recombining radical sites but rather by the migration of the radical site through PE melt due to rapid hydrogen transfer; this contribution was shown to be negligible for the conditions of the current work.