In order to test for non-Gaussianities with respect to scale-dependencies we use so-called surrogate maps, in which possible phase correlations of the Fourier phases of the original WMAP data and simulations, respectively, are destroyed by applying a shuffling scheme to the maps. A statistical comparison of the original maps with the surrogate maps then allows to test for the existence of higher order correlations in the original maps, also and especially on well-defined Fourier modes. Using Minkowski functionals as an image analysis technique we calculate the deviation between the original data and 500 surrogates for different hemispheres in the sky and find ecliptic hemispherical asymmetries between northern and southern ecliptic sky. We find strong deviations from Gaussianity in the WMAP data when considering the low-l range with l = [2,20]. The analysis technique of the scaling indices leads to a slightly lower deviation. Although the underlying foreground reduction methods of the maps differ from each other, we find similar results for the WMAP 7yr ILC map and the WMAP 7yr (needlet-based) NILC map in the low-l range. Our results point once more to a cosmological nature of the signal. For a higher l range with l = [120,300] the results differ between the two image analysis techniques and between the two maps which makes an intrinsic nature of the signal on this l range less likely. When we decrease the size of the analysed sky regions for the low-l study, we do not find signatures of NG in the northern sky. In the south we find individual spots which show deviations from Gaussianity. In addition, we investigate non-Gaussian CMB simulations that depend on the f_NL parameter of the local type. These simulations cannot account for the detected signatures on the low-l range.