New methods of estimating the size of salmonids Salmo spp. and eels Anguilla anguilla consumed by otters Lutra lutra are presented, based on feeding trials involving captive, tame otters. These methods involve modelling the size-related differential recovery of key fish bones recovered in faeces (spraints) and avoid two sources of errors that may have occurred using previous techniques which relied solely on a series of correlations between fish length and the length of individual vertebrae. Sources of errors avoided are: (a) that vertebrae recovered in spraints could not always be assigned to the correct correlation equation for mean, maximum, or minimumsized bones; and (b) that the number of fish represented in a spraint was assumed to be one, in the case of similarly-sized vertebrae, and two in the case of bones varying considerably in size. We tested the use of salmonid atlas vertebrae to determine the largest minimal numbers estimate, the length frequency distribution of salmonids consumed, and to estimate the proportions of trout and salmon in the diet. Eels do not contain an atlas which is resistant to digestion and so equivalent estimates of the minimum numbers of fish represented in spraints are not possible. A model was therefore developed to estimate the distribution of lengths of fish consumed from the lengths of thoracic vertebrae recovered in the spraints using a series of equations. For each type of prey (salmonids or eels), the proportions of identifiable bones recorded in spraints were related to the original size of ingested fish. Such models greatly enhance the value of otter spraint analyses, particularly for size-selection studies.