The decline of the European eel (Anguilla anguilla) stock has led the European Commission to initiate an Eel Recovery Plan (Council Regulation N0 1100/2007), in which each member state was required to establish Eel Management Plans. Various measures in the French plan aim at minimizing the impact of obstacles on the colonisation of continental water by glass eels. In this context, a method to assess the passability of an obstacle seems essential. Tag-recaptures methods are often used for analysis. In this study, we developed a specific tag-recaptures method appropriate to elvers and an associated Bayesian model to assess (i) the passability of a ladder and (ii) to quantify the effect of various environmental factors on the passability. Nine tag-recaptures campaigns were carried out from 2009 to 2011 on an obstacle named “Pas-du-Bouc” (France) equipped with an eel ladder and trap since 2008. The site is located about 3 kilometers upstream from the tidal limit. About 4400 young eels (total length between 55 and 240mm) were tagged by batch (a batch corresponding to a campaign and a release location) using coloured Visible Implant Elastomers (VIE), enabling the identification of the batches. Fishes were released in 50 batches either 100m downstream, or directly down the ladder. The trap was then visited every two days and the batch of trapped eels were identified thank to the VIE. A Bayesian model was then developed to analyse the results. This model aims at predicting the probability for an elver of climbing the ladder according to some batch characteristics (release location), eels characteristics (length and sanitary conditions) and environmental conditions (especially temperature and river flow). The probability of climbing was subdivided into two steps: first the probability for an eel in the river to find and climb the ladder and (ii) the probability for an elver that have found the ladder to climb it. About 50% of eels released in front of the ladder were recaptured, while only 22% were recaptured when released downstream. Moreover, it is generic and has already been applied to two other obstacles. Concerning the Bayesian model, it was able to predict well the probability for an elver to climb the ladder, and it especially demonstrates the major influence of the river flow on the probability for an eel to reach the ladder. The model will be applied to other sites in order to distinguish environmental factors that are site specific or more general.