We present proton-proton multiple-quantum investigations on a series of monomodal and strongly bimodal end-linked poly(dimethylsiloxane) model networks. A robust pulse sequence characterized by a well-defined double-quantum Hamiltonian along with a specific normalization approach is used to obtain double-quantum build-up curves. These curves are analyzed in terms of the spin dynamics of a local subsystem of monomer-fixed spins, where analytical fitting functions yielding residual dipole-dipole coupling constants are derived on the basis of exact solutions provided by simulations. Further employing the novel experimental strategy of double-quantum preselection of elastically active network chains, it is shown that the network response is purely heterogeneous, and that the data can be analyzed in terms of distributions of local dynamic order parameters using different models. The results yield consistent proof that local chain order in bimodal networks obeys a linear mixing law of short- and long-chain components. The order parameter distribution in a long-chain monomodal network is found to be surprisingly narrow, with a rather high average order parameter. Implications on the validity of present theories used to explain order and dynamics in networks are discussed. (C) 2003 American Institute of Physics.