1. Modelling of the population dynamics of seals require data on an array of vital parameters (fecundity, mortality, age structure, migrations, population growth rate). The most common way to obtain these data is to estimate the parameters from samples taken from the population. However, the influence from skewed samples can be substantial in populations with age- and sex-specific features. By quantifying the behavioural differences among age and sex classes, data from skewed samples can be compensated retrospectively. Awareness of the existence and the potential magnitude of such biases is highly relevant for the design of surveys, sampling programmes and the implementation of management plans of age structured populations.
2. The age- and sex-specific behaviour of harbour seals Phoca vitulina and grey seals Halichoerus grypus can be studied by using freeze-branded animals. Since the brand is permanent and visible up to a distance of 500 m, the harassment is limited to one occasion in the lifetime of the seal (the catching day).
3. A method is also given for analysing data arising from re-sighted branded animals, where re-sightings of individual seals were transformed to estimates of relative haul-out frequencies of seals by age and sex.
4. The composition of harbour seal groups on land exhibit a conspicuous seasonal flux, and the fraction on land was not representative of the entire population at any time during the summer. The results have far-reaching implications since most studies of seals are carried out at haul-out sites, and differential behaviour between the sexes and among age classes is expected in all populations and species of seals. Skewed samples generate biases in estimates of population growth rate, age-specific mortality and fecundity.
5. Age-specific haul-out patterns must be taken into account when analysing data from populations with non-stable age structures. As a consequence of changes in age structure after the 1988 seal epizootic, surveys under-estimated the size of the Swedish–Danish harbour seal population by 6% in 1988 and over-estimated the same parameter by up to 16% during the following years.
6. The present paper establishes the complications of sampling natural populations that are structured by age and sex, and presents a method on how to quantify sampling errors.