Models of the β-delayed neutron emission (βn) assume that neutrons are emitted statistically via an intermediate compound nucleus post β decay. Evidence to the contrary was found in an 134Inβ-decay experiment carried out at ISOLDE CERN. Neutron emission probabilities from the unbound states in 134Sn to known low-lying, single-particle states in 133Sn were measured. The neutron energies were determined using the time-of-flight technique, and the subsequent decay of excited states in 133Sn was studied using γ-ray detectors. Individual βn probabilities were determined by correlating the relative intensities and energies of neutrons and γ rays. The experimental data disagree with the predictions of representative statistical models which are based upon the compound nucleus postulate. Our results suggest that violation of the compound nucleus assumption may occur in β-delayed neutron emission. This impacts the neutron-emission probabilities and other properties of nuclei participating in the r-process. A model of neutron emission, which links the observed neutron emission probabilities to nuclear shell effects, is proposed.