Kallivayalil et al .
have used the Hubble Space Telescope to measure proper motions of the LMC and SMC using images in 21 and five fields , respectively , all centered on known QSOs .
These results are more precise than previous measurements , but have surprising and important physical implications : for example , the LMC and SMC may be approaching the Milky Way for the first time ; they might not have been in a binary system ; and the origin of the Magellanic Stream needs to be re-examined .
Motivated by these implications , we have reanalyzed the original data in order to check the validity of these measurements .
Our work has produced a proper motion for the LMC that is in excellent agreement with that of Kallivayalil et al. , and for the SMC that is in acceptable agreement .

We have detected a dependence between the brightness of stars and their mean measured motion in a majority of the fields in both our reduction and that of Kallivayalil et al .
Correcting for this systematic error and for the errors caused by the decreasing charge transfer efficiency of the detector produces better agreement between the measurements from different fields .
With our improved reduction , we do not need to exclude any fields from the final averages and , for the first time using proper motions , we are able to detect the rotation of the LMC .
The best-fit amplitude of the rotation curve at a radius of 275 arcmin in the disk plane is 120 \pm 15 km s ^ { -1 } .
This value is larger than the 60–70 km s ^ { -1 } derived from the radial velocities of HI and carbon stars , but in agreement with the value of 107 km s ^ { -1 } derived from the radial velocities of red supergiants .

Our measured proper motion for the center of mass of the LMC is ( \mu _ { \alpha } , \mu _ { \delta } ) = ( 195.6 \pm 3.6 , 43.5 \pm 3.6 ) mas century ^ { -1 } ; that for the SMC is ( \mu _ { \alpha } , \mu _ { \delta } ) = ( 75.4 \pm 6.1 , -125.2 \pm 5.8 ) mas century ^ { -1 } .
The uncertainties for the latter proper motion are 3 times smaller than those of Kallivayalil et al .