The paper reports results of experiments on three-person quota games w
ithout the grand coalition and with zero values for the one-person coa
litions. The experimental procedure used is the demand commitment mode
l. This model generates finite extensive games with perfect informatio
n whose neutral equilibrium point predicts quota agreements. The term
neutral means that at every decision point all locally optimal choices
are taken with equal probability. Quota agreements do not perform bet
ter than equal division payoff bounds, but nevertheless the experiment
s show that quotas have some behavioral relevance. The demand commitme
nt model seems to favor learning of the quota concept if subjects have
the opportunity to gain experience by frequently playing the same gam
e,