The applicability of the potential approximation in the case of open universes is tested .
Great Attractor-like structures are considered in the test .
Previous estimates of the Cosmic Microwave background anisotropies produced by these structures are analyzed and interpreted .
The anisotropies corresponding to inhomogeneous ellipsoidal models are also computed .
It is proved that , whatever the spatial symmetry may be , Great Attractor-like objects with extended cores ( radius \sim 10 h ^ { -1 } ) , located at redshift z = 5.9 in an open universe with density parameter \Omega _ { 0 } = 0.2 , produce secondary gravitational anisotropies of the order of 10 ^ { -5 } on angular scales of a few degrees .
The amplitudes and angular scales of the estimated anisotropy decrease as the Great Attractor size decreases .
For comparable normalizations and compensations , the anisotropy produced by spherical realizations is found to be smaller than that of ellipsoidal models .
This anisotropy appears to be an integrated effect along the photon geodesics .
Its angular scale is much greater than that subtended by the Great Attractor itself .
This is understood easily taking into account that the integrated effect is produced by the variations of the gravitational potential , which seem to be important in large regions subtending angular scales of various degrees .
As a result of the large size of these regions , the spatial curvature of the universe becomes important and , consequently , significant errors ( \sim 30 per cent ) arise in estimations based on the potential approximation .
As it is emphasized in this paper , two facts should be taken into account carefully in some numerical estimates of secondary gravitational anisotropies in open universes : ( 1 ) the importance of scales much greater than those subtended by the cosmological structures themselves , and ( 2 ) the compatibility of the potential approximation with the largest scales .