We present OH (1612 MHz) and H2O (22.2 GHz) maser observations made wi
th the VLA at an angular resolution of approximately 1'' toward three
OH/IR stars: OH 12.8-0.9, OH37.1-0.8 and OH42.3-0.1. These stars exhib
it the H2O maser components in a velocity range that exceeds the OH ve
locity range, a situation in disagreement with the standard model for
maser emission in late-type stars. We find that the OH and H2O maser p
ositions agree within approximately 0.5'' and that most likely both ma
sers arise from the same source, ruling out the possibility of a fortu
itous superposition of two different sources. We discuss some of the p
ossible mechanisms that could explain the anomalous H2O maser emission
. The most likely explanation is that these stars have axisymmetric wi
nds and that the H2O emission originates in the polar regions with lar
ger velocities than the regions closer to the equator where the OH emi
ssion arises. Very long baseline observations are required to test thi
s explanation.