In memory for subject-performed tasks (SPTs), subjects encode a List of sim
ple action phrases (e.g., thumb through a book, knock at the door) by perfo
rming these actions during learning. In three experiments, we investigated
the size of the levels-of-processing effects in SPTs as compared with those
in standard verbal learning tasks (VTs). Subjects under SPT and VT conditi
ons learned lists of action phrases in a surface or a conceptual orienting
task. Under both encoding conditions, the subjects recalled fewer items wit
h surface orienting tasks than with conceptual orienting tasks, but the lev
els-of-processing effects were strongly reduced in the SPT condition. In th
e SPT condition, items that were encoded in a surface orienting task were s
till substantially recalled. The items were recalled almost as well as the
conceptually encoded items in the VT condition. The distinct reduction of t
he levels-of-processing effect is caused by the fact that, in SPT encoding
even with a verbal surface orienting task, subjects process conceptual info
rmation in order to perform the denoted action. We attribute the small conc
eptual advantage, which remains with SPT despite the conceptual processing
for performing, to the fact that items are not as well integrated into memo
ry as they are when conceptual processing is focused on the action componen
t, rather than on the semantic contexts. This lower integration reduces the
accessibility of items in the verbal surface task, even with SPT encoding.