We consider the most basic sequential game, called the 'game of trust'
, in which the first mover starts by deciding between cooperation and
non-cooperation, whereas the second mover can only react in case of co
operation by either exploiting the other player or by dividing the rew
ards equally. The unique subgame perfect equilibrium predicts non-coop
eration although this is payoff-dominated by fair cooperation. We anal
yze experiments by which we want to assess whether there is trust in t
he other player's fairness, and, if so, whether it enables mutually be
neficial cooperation. When positions were auctioned, there was no trus
t. When positions were allocated randomly, the considerably often occu
rring trust was rarely rewarded.