AbstractThree approaches towards a fracture theory are discussed, namely the statistical, the continuum‐mechanical, and the micro‐morphological approach. The latter recognizes that (amorphous or semicrystalline) high‐polymer solids are composed of chain molecules which show a close range order and/or a superstructure. Information from morphological and spectroscopic investigations is utilized to characterize the different elementary molecular processes (slip of chains, segment rotation, chain scission, crystal deformation) which may lead to fracture initiation, crack growth, or yielding.The statistical variability of the elementary processes and the correlation of these processes with each other determine the statistical variability of the cumulative process,e.g., time to failure or fracture strength. Thus, a relation between the results of the micro‐morphological approach and the terms of the statistical approach (underlying distribution) is obtained. A mathematical study of the effect of these processes (network degradation, stress relaxation) on the expectation values of strength or time to failure is carried out for networks of different degrees of orientation.The results obtained for one‐phase networks are applied to critically stressed regions of heterogeneous polymers (amorphous regions in oriented fibers, plastically deformed zone at tip of growing crack). Thus, a correlation between continuum‐mechanical terms (stress intensity factor, surface energy) and molecular terms (network properties, activation energies) is attempted. A failure criterion