In connection with statements made at the preceding session, some results presented at the Nice Conference on systematic and random errors affecting the elements a and P as well as a3/P2 may be recalled. If many orbits have been calculated for one binary, it is instructive to plot log P, log a, and log a3/P2 versus log Δt/P where Δt is the time interval covered by observations. As a general rule, when Δt ≃ P the systematic trends vanish, and the scatter of the plotted elements reaches a low value, not considerably changing thereafter. It may well be of interest, then, to have several orbits based on the same observations but computed by different authors and methods.