The concept of grain boundary engineering (GBE) has been proposed based on the fact that many studies have demonstrated that boundaries associated with low value coincident site lattice (CSL) misorientations show higher resistance to intergranular fracture and corrosion, reduced susceptibility to impurity segregation and superior ductility. It is commonly accepted that for fcc metals of low to medium stacking fault energy metals, including Ni and many Ni–alloys, the most important CSL boundary for the GBE process is a Σ3 boundary, the occurrence of which is dominated by the formation of annealing twins. Moreover, it has been found that repetitive thermo–mechanical processing can be used to increase further the fraction of Σ3 (and Σ3ⁿ (n >1)) boundaries. However, the mechanism for this is not yet clear. Therefore, an investigation on the evolution of Σ3 boundaries during recrystallization is important for understanding the mechanisms of GBE for those materials. In the present paper the evolution of Σ3 boundaries during recrystallization in a 96% cold–rolled sample of pure nickel f 99.996% purity has been explored using orientation maps obtained using electrn backscatter diffraction (EBSD). Each orientation map was taken from the same area after annealing for various times. Based on the EBSD data the Σ3 boundaries can be divided into two groups: "twin" type and "non–twin" type. These groups can be differentiated using a parameter of deviation angle (Δθ) of boundary misorientation to the ideal twin misorientation (60°<111>). During recrystallization incoherent twin boundaries are found to develop from coherent twin boundaries. It is found also that most Σ3ⁿ (n >1) boundaries are formed by impingement of a nucles with its n–order twins, and that the chance for such impingement events decreases significantly with increasing n. Most non–twin type Σ3 boundaries arise from impingement of Σ1 and twin type Σ3 boundares. Non–twin type Σ3 boundaries may be more effective than twin type Σ3 boundaries to develop a beneficial grain boundary network.