Stereoscopic vision is the ability to use binocular disparity – difference in the viewpoints of the eyes - to perceive depth in our environment. Poor depth perception is one of the main symptoms of amblyopia, therefore of clinical interest to improve it for stereoanomalous persons. To properly assess the effectiveness of clinical intervention, a robust and precise measure of stereopsis is needed, triggering the minimal amount of learning. The present study aims at measuring the test-retest reliability and learning in two new tests assessing stereovision: a global stereoacuity test using a stereoscope (eRDS) and a local stereoacuity test using virtual reality (Vivid Vision). 56 adults (18 to 35 yo) with normal vision but likely-impaired depth perception were recruited and allocated into two groups. They underwent six sessions of repeated measures either the on eRDS stereotest or on the Vivid Vision stereotest. We found that the eRDS test is efficient at detecting very high stereoacuities. High correlations exist between the different test-retest sessions and learning occurs in the first testing times. We observed lower test-retest correlations for the Vivid Vision test, and we found no evidence of learning through the testing sessions. However, stereoacuities under 20 arcsec are not detected with that test, and participants showed high variability in their responses. This study demonstrates that our two new tests are reliable and can measure precise stereoacuities, each with different advantages and limitations.