We study the rotational distortions of the vacuum dipole magnetic field in the context of geometrical models of the radio emission from pulsars .
We find that at low altitudes the rotation deflects the local direction of the magnetic field by at most an angle of the order of r _ { n } ^ { 2 } , where r _ { n } = r / R _ { lc } , r is the radial distance of the radio emission and R _ { lc } is the light cylinder radius .
To the lowest ( ie. second ) order in r _ { n } , this distortion is symmetrical with respect to the plane containing the dipole axis and the rotation axis ( ( \vec { \Omega } , \vec { \mu } ) plane ) .
The lowest order distortion which is asymmetrical with respect to the ( \vec { \Omega } , \vec { \mu } ) plane is third order in r _ { n } .
These results confirm the common assumption that the rotational sweepback has negligible effect on the position angle ( PA ) curve .
We show , however , that the influence of the sweepback on the outer boundary of the open field line region ( open volume ) is a much larger effect , of the order of r _ { n } ^ { 1 / 2 } .
The open volume is shifted backwards with respect to the rotation direction by an angle \delta _ { ov } \sim 0.2 \sin \alpha r _ { n } ^ { 1 / 2 } where \alpha is the dipole inclination with respect to the rotation axis .
The associated phase shift of the pulse profile \Delta \phi _ { ov } \sim 0.2 r _ { n } ^ { 1 / 2 } can easily exceed the shift due to combined effects of aberration and propagation time delays ( \approx 2 r _ { n } ) .
This strongly affects the misalignment of the center of the PA curve and the center of the pulse profile , thereby modifying the delay-radius relation .
Contrary to intuition , the effect of sweepback dominates over other effects when emission occurs at low altitudes .
For r _ { n } \lesssim 3 \cdot 10 ^ { -3 } the shift becomes negative , ie. the center of the position angle curve precedes the profile center .
With the sweepback effect included , the modified delay-radius relation predicts larger emission radii and is in much better agreement with the other methods of determining r _ { n } .