Within the currently accepted mechanism of the Li-salt-free Wittig reaction, the phenomenon of stereochemical drift re- mains the one remaining ¿loose end¿ in an otherwise intern- ally consistent explanation of a large body of diverse obser- vations. The term describes the nonstereospecific decompo- sition of the oxaphosphetane (OPA) intermediate in reactions of certain alkylides with certain aldehydes. In this paper, it is shown that the previous examples in which drift occurs are not merely isolated aberrations from the observed norm, but rather that there is a general phenomenon in reactions of eth- ylides with benzaldehydes. Variable-temperature NMR (VTNMR) spectroscopy was used to establish that the amount and diastereomeric ratio of the OPA intermediates do not change below a certain temperature. At and above the temperature at which OPA decomposition to alkene and phosphine oxide begins to occur, the alkene shows a dif- ferent diastereomeric ratio to the OPA, which indicates the occurrence of stereochemical drift. In one example, owing to an apparent remarkable coincidence of rates, the diastereomeric ratio of the OPA does not change above the de- composition temperature, even though stereochemical drift occurs in the formation of the alkene product. An alternative mechanism for drift involving its catalysis by aldehyde was not confirmed. Drift was also shown not to occur in similar Wittig reactions of structurally related longer-chain alkylides by stereospecific decomposition of OPA intermediates generated from β-hydroxyphosphonium salts (β-HPSs). The extremely useful (and generally applicable) NMR techniques, 1H¿31P HMBC and selective 1H{31P}, which we have utilised to establish kinetic diastereomeric ratios, are described in full for the first time. Details of the determination of the relative stereochemistry of two β-HPSs (derived from acid quenching of OPAs) by X-ray crystallography are also given.