Resonant inelastic x-ray scattering is used to investigate the electronic origin of orbital polarization in nickelate heterostructures taking LaTiO3 − LaNiO3 − 3x(LaAlO3), a system with exceptionally large polarization, as a model system. We find that heterostructuring generates only minor changes in the Ni 3d orbital energy levels, contradicting the often-invoked picture in which changes in orbital energy levels generate orbital polarization. Instead, O K-edge x-ray absorption spectroscopy demonstrates that orbital polarization is caused by an anisotropic reconstruction of the oxygen ligand hole states. This provides an explanation for the limited success of theoretical predictions based on tuning orbital energy levels and implies that future theories should focus on anisotropic hybridization as the most effective means to drive large changes in electronic structure and realize novel emergent phenomena.The electronic structure of transition metal oxides (TMOs) is dominated by the active 3d TM orbitals and how these hybridize with the neighboring oxygen 2p ligand orbitals. Building heterostructures from one-unitcell-thick layers of different TMOs offers the opportunity of tuning the TM 3d states configuration and potentially realizing emergent phenomena with new or improved properties [1][2][3][4][5][6]. LaNiO 3 based heterostructures are a prototypical example of such an endeavor [6][7][8][9][10][11][12][13][14][15][16][17][18][19][20], motivated by several predictions of novel superconducting, magnetic and topological states [21][22][23][24]. Bulk LaNiO 3 has nominally Ni 3+ ions with filled t 2g states and one e g electron. Efforts to realize the predicted novel states are crucially dependent on inducing orbital polarization in the Ni e g level [25], i.e. breaking its degeneracy and pushing the system towards a half filled x 2 − y 2 (or 3z 2 − r 2 ) configuration ( Fig. 1(b)). Initial efforts to realize strong orbital polarization in heterostructures reported up to ∼ 20% change in orbital occupancy compared to the bulk [7-12, 14, 19], corresponding to ∼ 0.3 eV splitting of the e g states as inferred from cluster calculations of x-ray absorption spectra [14], in fact first principle calculations indicate that ∼ 10% strain is needed to drive ∼ 40% change in orbital polarization (see supplemental material of Ref. 17). Novel trilayer LaTiO 3 − LaNiO 3 − 3x(LaAlO 3 ) (LTNAO) superlattices were recently produced obtaining ∼ 50% change in orbital polarization via charge transfer, polar charge and electronic confinement effects [15][16][17][18]. To date, changes in orbital polarization have been conceptualized as a consequence of tuning the 3d orbital energy level through crystal field engineering. However, this approach has led to a dramatic mismatch between theoretical predictions and experimental results [14,[21][22][23]26], raising questions about the precise electronic character, and the driving mechanism behind, orbital polarization in these materials.We apply resonant inelastic x-ray scattering (RIXS) [27,28] ...