Oblique magnetic resonance imaging of the temporal lobe (tilted orientation) requires a stable reference line with minimum variability. In the clinical setting, where several observers carry out examination of the patients, there is a need to assure minimum inter-observer variability, in order to obtain comparable tilted anatomical planes. This is particularly relevant when performing quantitative imaging (qMRI) of the hippocampus, amygdala and para-hippocampal cortices. In this study, eight experienced observers tested the stability of four sagittal reference lines by manually tracing the posterior commissure-obex (PC-OB) line, the line tangential to the anterior surface of the pons at its most convex point and the lines orthogonal to the main axis of both hippocampi, in ten exams of healthy subjects. The stability of the tracing was assessed by comparing the inter-observer variability expressed by the variances of the measurements. The observers' performance was assessed by comparing the precision of the tracing for each line. We tested the results statistically using Bartlett's test (analysis of the variances of the four lines) followed by Fischer-Snedecor (in order to compare the two lines that had the smallest variance). The PC-OB line and the line tangential to the anterior surface of the pons had smaller inter-observer variances than the orthogonal lines (p < 0.01). In addition, the variance of the PC-OB line was smaller than that of the line tangential to the pons (p < 0.01). There were no significant intra-observer differences in the precision of tracing of any of the lines. We show quantitatively that the PC-OB line is the scout reference that yields the smallest inter-observer variance. Thus, this line should be preferred to improve the reproducibility of temporal lobe imaging while performing tilted coronal and axial sequences, to make quantitative assessments of the hippocampus, amygdala and para-hippocampal cortices.