Purpose To evaluate the intradevice repeatability and agreement for peripapillary retinal nerve fiber layer (pRNFL) measurements in healthy eyes with two different scan directions and two different number of B scans. Methods pRNFL was measured with a spectral domain optical coherence tomography on 54 healthy participants. Three-dimensional optic disc scans (6 mm x 6 mm) were performed on the right eye of the participants. Two repeated scans were performed in four different settings: H1: Horizontal scan with 512 A-scans x 96 B-scans; H2: Horizontal scan with 512 A-scans x 128 B-scans; V1: Vertical scan with 512 A-scans x 96 B-scans; V2: Vertical scan with 512 A-scans x 128 B-scans. The pRNFL thickness was evaluated in twelve clock-hour sector in a circle of 3.45 mm diameter centred at the optic disc. Repeatability and agreement were assessed with within subject standard deviation (Sw) and Bland-Altman test respectively. Results The repeatability of pRNFL measurements varied depending on the scan direction and sectors. The repeatability for the horizontal sectors were better with H1 and H2, with sector 9 having the best Sw (< 3 μm). The repeatability for the vertical sectors were better with V1 and V2 with sector 5 and 9 having the best Sw (< 4 μm). The repeatability with vertical scan was more symmetric among the sectors than with horizontal scans. The repeatability metrics of the sectors did not vary much between H1 and H2 (difference < 2 μm) and between V1 and V2 (difference < 3.2 μm). Comparing horizontal and vertical scans, the vertical sectors had larger limits of agreement of about 45 μm. Conclusion The reliability of the pRNFL thickness measurements is dependent on the direction of the scan and independent on the numbers of B-scans. Vertical scans for pRNFL gives more homogeneous repeatability across the different sectors.