To determine the conditions under which elastic waves are radiated from crack sources, dynamic tensile cracks were propagated in glass samples in the double cantilever beam geometry. This geometry allows simple calculation of the strain energy release rate G at initiation from measured parameters of crack length, applied crack opening force, and crack opening displacement. Partial control over the strain energy state in the sample at fracture initiation, and hence G at initiation, was achieved by varying the geometry of the notch tip from which the fracture emanates. Elastic wave displacements were monitored with a broadband capacitance transducer with a pointlike probe. A single component of elastic wave displacement (parallel to the crack plane and perpendicular to the crack propagation direction) was measured. Two fracture configurations were investigated: (1) “primary fracture” in glass plates of dimension 305×102 ×12.7 mm and (2) “secondary fracture” in previously fractured glass plates of the same dimensions, bonded intermittently along the fracture plane. Primary fracture experiments afforded a means of investigating elastic wave radiation from mode I cracks in a highly brittle material, such that the strain energy released by the fracture is partitioned into fracture surface energy of the newly formed crack walls and radiated elastic wave energy; negligible energy is expended in ductile or frictional processes. Secondary fracture experiments afforded a means of investigating elastic wave radiation in the case of varying fracture surface energy along the crack path. For primary fracture, measurable elastic waves from the macrofracture were generated in 31% of the 16 dynamic fracture events monitored. The condition for radiation of measurable waves from these fractures appears to be a local abrupt change in the fracture path direction, such as occurs when the fracture intersects a surface flaw. For the five events with measurable elastic waves, the ratio of radiated elastic wave energy in the measured component to the fracture surface energy of the macrocrack was 0.0001–0.001. For secondary fracture, 100% of the 13 events monitored showed measurable elastic waves. The ratio of radiated elastic wave energy in the measured component to fracture surface energy was 0.001–0.01, or 10 times greater than for primary fracture. The observed value of G at crack initiation for both primary and secondary fracture ranged from 3 to 48 J m−2. When the time window for radiated elastic wave energy calculation was restricted to a few microseconds after the first arrival, a weak correlation of radiated elastic wave energy with initiation G value was observed for secondary fractures.