A long cycle time task is assumed to consist of a series of non-repeti
tive unique sub-tasks whose standard times average at about 11/2 minut
es. 'Forgetting' is therefore a consequence of a specific sub-task rea
ppearing in the next cycle after a whole cycle time of other activitie
s is completed. Learning behavior of long cycle tasks is therefore pre
dicted on the learning of its constituent sub-tasks. A method for pred
icting the learning curve parameters for the sub-tasks (the learning c
onstant, and execution time of the first repetition) are proposed and
tested. The extent of 'forgetting' is empirically determined as a func
tion of the learning constant and interruption length. Finally, a mode
l is developed for predicting execution times for long cycle tasks.