The production of vanillin from sodium lignosulfonate under highly alkaline conditions, catalyzed by Cu2+ and at elevated temperature and pressures, has been studied in two sizes of stirred reactors. The larger reactor (3 L) was operated in both the dead end and the gas throughflow modes; the sparged gas was nitrogen and "simulated air" in the former case and air in the latter. The smaller reactor (300 mL) was only operated in the batch mode with oxygen. In the 3 L reactor, with nitrogen gas alone in the dead end mode, vanillin was produced by hydrolysis. With the other conditions, both hydrolysis and oxidation occurred and the amount of vanillin produced was greater. In addition, for the first time since this process was first introduced in 1936, the composition of the gas phase produced by the reaction was investigated, too. The measurements were made on samples taken from the headspace of the 300 mL batch reactor and the headspace of the 3 L reactor in the dead end mode and in the exhaust gases in the throughflow mode. It was found that, whenever vanillin was produced, hydrogen was detected in the gas phase. In the 3 L reactor in the dead end mode, the amount of H-2 formed was so great (similar to 7% by volume) in the case of "simulated air" that the production of vanillin ceased as no further air (oxygen) was able to enter the reactor. In the throughflow mode, the concentration of hydrogen detected in the exit gas was much lower as it was flushed out in the exhaust. As a result of the different levels of oxygen and hydrogen in the reactor in the dead end and throughflow modes, the amount of vanillin produced was greater in the latter case. Thus, it is difficult to use studies in the dead end mode to predict the behavior in throughflow, the mode generally used industrially.