it is well-known that adding extra capacity to queues in networks where individuals choose their own route can sometimes severely degrade performance, rather than improving it. we will discuss some simple examples of queueing networks where this is the case under probabilistic routing, but where under state-dependent routing the worst case performance is no longer seen. this raises the question of whether giving arrivals more information about the state of the network leads to better performance more generally. this is joint work with heti afimeimounga, lisa chen, mark holmes, bill solomon, and, latterly, niffe hermansson and elena yudovina. (jointly sponsored by the mathematics and ece departments.)