False negative detections may bias the surveys for rare species and reduce the reliability of models based on the proportion of occupied patches. We assessed the detectability of the Eurasian otter Lutra lutra through the standard survey method by analysing the detection history of 28 sampling stretches surveyed monthly between March 2001 and January 2003. Each survey negative for otter spraints was considered as a false negative if the otter had been recorded in the previous and/or following month (respectively, cFN and FN). Otter marking intensity (MI) (MI=N° of spraints per kilometre) was calculated and assumed to represent an index of its relative abundance. Spraints were found in 81.7% of all surveys. Yearly MI ranged from 1.02 to 101.4 spraints per kilometre. In 2002, mean MI was significantly lower than in the previous year, while no clear seasonal trend could be outlined. The minimum number of surveys required to establish the occurrence of the otter, as estimated by a probability model, was 2.6 and was inversely related to MI. For a sub-sample of 18 sampling stretches, the relation between the frequency of both cFN and FN and five variables of potential interest for otters was tested by means of stepwise linear multiple regressions, yielding two highly significant models, which both included only MI as the explanatory variable. The frequency of both FN and cFN was correlated to MI and the resulting equations used to assess the percentage of surveys positive for otters in both years. After the correction for non-detections, otter site occupancy did not vary between the 2 years, except for one river when applying the more conservative estimate of false negatives (cFN). Multiple visits and the assessing of MI should become standard components of otter surveys. This approach has broad applicability and may be applied to assess the large-scale distribution of other rare or elusive mammalian carnivores.