For simple theoretical models of an underground explosion, two apparently different relationships between isotropic moment and volume change have been published. We show that there is no inconsistency, because these relationships are based on two different definitions of volume change at the explosion source. We report observations of three 25-ton chemical explosions conducted in 1997 at the former nuclear test site near Semipalatinsk, Kazakhstan. These show the theoretically predicted decrease of signal strength as the shot depth increases. We argue that the relationship between isotropic moment and volume change proposed by Müller (1973) is preferred.