The Touvre spring is the only outlet of a large karstic system, the La Rochefoucauld karst, located in the West of France. This system is particularly interesting from a modeler's point of view, for it allows to investigate how karstic losses and gains can be accounted for within a rainfall-runoff model. Indeed, this spring is fed by the losses of three rivers, which are themselves gaged after recharging the karstic aquifer: this setting is exceptional in that simulations of surface catchment losses (estimated by the GR4J rainfall-runoff model) can be compared with measurements of spring discharge at the outlet of the karstic system. It provides an opportunity to test the estimates of catchment leakages', which are often suspected but seldom measurable, against observations. We show here that the conceptualization of surface losses in the GR4J rainfall-runoff model yields simulated leakage volumes that are consistent with the spring flow volumes and that, provided a routing store is added to GR4J to account for the specific retarding behavior of the karstic aquifer, an integrated model can be proposed to represent the behavior of the whole system adequately. The Touvre case study shows that a parsimonious, empirical rainfall-runoff representation of a karstic system is possible. It also provides an independent corroboration for the conceptualization of surface catchment leakages proposed in the GR4J rainfall-runoff model.