problem with zero-length element

[srisri_iitk]

while simulating transmitting boundary i have used zerolength elements fixed at one node and the other node connected to soil domain
I am getting some results if i use uniform excitation command but if i use multisupport excitation by converting the acceleration time history to displacement time history, the nodes which are connected to soil domain have very high accelerations but remaining nodal accelerations are ok
can any body suggest me some thing regarding this
for zero-length elements if we fix one node, does the second node get connected to the other adjacent node or will it be act like fixed node?