How stable is perceived direction of gravity over extended periods in darkness?

Previous studies reported linear drift of perceived vertical for brief (≤10 min) observation periods. Here, we repeated estimates of direction of gravity up to 60 min to evaluate whether the drift is sustained, shows saturation or even reverses over time. Fifteen healthy human subjects repetitively adjusted a luminous line along subjective visual vertical (SVV) and horizontal (SVH) over periods of 5 min (constituting one block). We obtained seven blocks within 60 min in each subject for SVV and SVH. In between the first six blocks, subjects remained in darkness for 5 min each, whereas the lights were briefly turned on before block 7. We noted significantly (p < 0.05) increased errors in perceived direction of gravity by block 2 (SVV) and 3 (SVH). These increases disappeared after turning on the lights before block 7. Focusing on blocks 2-6, significant drift started from similar offset positions and pointed to the same direction in a majority of runs in 9/15 (SVV) and 11/15 (SVH) subjects. When pooling data from all blocks, orthogonality of errors was lost in all subjects. Trial-to-trial variability remained stable over the seven runs for SVV and SVH. Only when pooling all runs, precision was significantly (p < 0.05) higher for the SVH. Our findings suggest that perceived direction of gravity continues to fluctuate over extended recording periods with individuals showing unique patterns of direction-specific drift while variability remains stable. As subjects were upright during the entire experiment and as drift persisted over several blocks, sensory adaptation seems unlikely. We therefore favor a central origin of this kind of drift.