While frailty has often been assumed to be a transitional state between robustness and ADL-dependence, no research has supported this assumption with empirical data. This study documents individual health trajectories in order to assess whether frailty (a) can be considered as a necessary precursor to ADL-dependence, (b) is of short duration or long-lasting and, (c) is reversible or irreversible. The octogenarians participating in the Swiss Interdisciplinary Longitudinal Study on the Oldest Old were assessed a maximum of nine times over ten years. The transitions between robustness (non-frail persons without ADL incapacities), ADL-independent frailty (frail without ADL incapacities) and ADL-dependence (frail with one ADL incapacity or more) were analyzed using Markov chain methods. At baseline, 42.4%, 49.4% and 8.1% of the sample were classified respectively as robust, ADL-independent frail, and ADL-dependent. Among those who became ADL-dependent, 76.6% were ADL-independent frail at the previous wave. Stability was the predominant pattern for ADL-independent frail people: on the average 58.8% of them maintained the same health status at the following wave. Recovery from ADL-independent frailty was possible (17.9%) but was often of short duration: among ADL-independent frail persons who recover to become robust, 51.9% became ADL-independant frail again one wave later. Frailty was the dominant health status in these very elderly individuals and can be considered as a relatively long transitional period between robustness and ADL-dependence. Recovery from ADL-independent frailty to robustness occurs but appears to be quite short-lived.