The QUENCH bundle experiments together with pertinent separate-effects tests are run to investigate the hydrogen source term resulting from water injection into an uncovered core of a light water reactor for emergency cooling. The test bundle consists of 21 fuel rod simulators, 20 of which are heated electrically over a length of 1024 mm. The center rod is either an unheated fuel rod simulator or a control rod containing B4C absorber material. The Zircaloy-4 rod cladding and the grid spacers are identical to those used in pressurized water reactors, whereas the fuel is represented by ZrO2 pellets. After transient heating to 2000 K and above, cooling of the test bundle is accomplished by injecting water or steam into the bottom of the test section. Hydrogen generation during cooling was found either to stop almost immediately or to increase for a certain time. Increased hydrogen generation was found in those tests in which local melting occurred, probably as a result of oxidation of the melt containing zirconium. Hydrogen release in the flooding/cooling phase of all QUENCH experiments performed so far seems to be insensitive to the coolant (water or steam) under similar test conditions.