Experiments designed to test the predictions of noncooperative game th
eory about the way that surplus will be divided in alternating offer g
ames with discounting have produced conflicting results. Neither the S
tahl/Rubinstein division nor the alternative equal division are genera
lly supported. The experiments reported here suggest a regularity in t
he data that is consistent with all the previous experiments: First pl
ayers who have a great advantage in the sense that the Stahl/Rubinstei
n division would give them a large share of the initial pie demand mor
e than half of the pie, but do not fully exploit their advantage. Firs
t players who are at a disadvantage demand equal divisions. Journal of
Economic Literature Classification Numbers: C72, C91. (C) 1994 Academ
ic Press, Inc.