In-plane hole g-factors measured in quantum point contacts based on p-type heterostructures strongly depend on the orientation of the magnetic field with respect to the electric current. This effect, first reported a decade ago and confirmed in a number of publications, has remained an open problem. In this work, we present systematic experimental studies to disentangle different mechanisms contributing to the effect and develop the theory which describes it successfully. We show that there is a new mechanism for the anisotropy related to the existence of an additional B+k 4 − σ+ effective Zeeman interaction for holes, which is kinematically different from the standard single Zeeman term B−k 2 − σ+ considered until now.PACS numbers: 71.70. Ej, 73.22.Dj, 71.18.+y A quantum point contact (QPC) is a narrow quasione-dimensional (1D) constriction linking two twodimensional (2D) electron or hole reservoirs. Experimental studies of QPCs started with the discovery of the conductance quantization in steps of. The steps are due to the quantization of transverse channels [3]. Effects of many-body correlations in QPCs were identified by a "0.7-anomaly" in the conductance, an enhancement of the g-factor in the 1D limit [4], and by a zero bias anomaly [5]. G-factors in n-type QPCs have been measured in numerous experiments; a relatively recent one is reported in Ref. [6].The in-plane electron g-factor in a QPC takes the same value for any direction of the in-plane magnetic field. Even in InGaAs, which has appreciable spin-orbit coupling, no in-plane g-factor anisotropy has been observed [7]. Contrary to this, measurements for holes in QPCs based on GaAs p-type heterostructures indicate a huge anisotropy. All previously reported values of the g-factor for magnetic fields applied perpendicular to the QPC are consistent with g ⊥ = 0 within experimental error, while the g-factor g || for the parallel direction is nonzero [8][9][10].Regardless of numerous studies, the g-factor anisotropy effect in QPCs remains unclear.One mechanism to explain the g-factor anisotropy was suggested in Ref. [9]. This mechanism is based on the crystal anisotropy of the cubic lattice. While it is not negligible, the contribution of this mechanism is too small to explain the observed anisotropy.In this work, we identify a new mechanism for the g-factor anisotropy unrelated to the crystal lattice. It is instructive to use classification in powers of crystal anisotropy η defined below. The new mechanism is leading in η and the mechanism [9] is subleading. The new mechanism is negligible at very low hole densities. However, at real physical densities it is the major anisotropy mechanism. Previous measurements were performed in 2D hole systems formed at a single heterojunction [8, 9], which can be modeled as a triangular potential well. There is also a measurement with an asymmetric quantum well [10] which can be modeled as a square potential with an electric field along the z-axis. The z-axis is perpendicular to the plane of the 2D hole system. The z-as...