The levels-of-processing effect is extraordinarily, even puzzlingly, robust. It occurs even when a memory test is expected, ample study time is given, and deep processing is encouraged regardless of orienting task. Thus, processing appears to be "fixed" by the requirements of the orienting task. This enigma is explored in a selective remembering procedure involving the recall of words of arbitrarily varying values. After verifying that selectivity is substantially localized at the encoding rather than the recall stage of the remembering process (Experiment 1), recall was found to be selective despite the imposition of an orienting task and regardless of whether item values were assigned according to the items' semantic category (Experiment 2) or entirely at random (Experiment 3). Indeed, no evidence was found for any effect at all of orienting tasks on selectivity (Experiment 4). Orienting tasks do not, after all, universally dominate the encoding process.