Yield phenomena during the tensile testing of pure titanium sheets of 0.2 mm and 0.4 mm in thickness were investigated in detail focusing on the effects of grain size (d), thickness (t), t/d ratio and tensile direction. With decreasing grain size, the yielding behavior changed from continuous yielding to one accompanied with yield point drop. Upper and lower yield stresses and 0.2%-proof stress followed the Hall–Petch relationship; however, coarse-grained specimens (d ≧ 20 µm) showed larger scatter in 0.2%-proof stress than the others. Consequently, the Hall–Petch coefficient (k) and friction stress (σ0) derived from 0.2%-proof stress are not accurate enough. The values of k and σ0 derived from various yield stresses and tensile directions were in the range of 250–600 MPa·µm0.5 and 30–180 MPa, respectively. Therefore, the validity of stress for yield stress was a concern, and the combination of the lower yield stress in the fine-grain range (d ≦ 20 µm) and 0.2%-proof stress excluding work-hardening in the coarse-grain range (d > 20 µm) was suggested to obtain reliable values of k and σ0, resulting in values of 370–460 MPa·µm0.5 and σ0 65–140 MPa, respectively. Moreover, it was revealed that k decreased and σ0 increased with the increasing angle of tensile direction to the rolling direction, regardless of thickness. The anisotropy of k is presumed to be affected by the grain boundary character rather than the Schmid factor, and neither rigidity nor the length of the Burgers vector are responsible. Meanwhile, the anisotropy of σ0 is verified to be affected by Schmid factors. Furthermore, it was clarified that the t/d ratio hardly affects upper and lower yield stresses (t/d > 14) nor 0.2%-proof stress (1.5 ≦ t/d ≦ 14).