Despite the widespread use of the doubly labelled water (DLW) method in energetic studies of free-ranging animals, effects of the method on study animals are rarely assessed. We studied behavioural effects of two alternative DLW protocols. During two consecutive breeding seasons, 42 parent black-legged kittiwakes received either the commonly used two-sample (TS) or the less invasive single-sample (SS) DLW treatment. A third group served as a non-treated control. We evaluated the effect of treatment with respect to the time birds took to return to their nest after treatment and recaptures, and the nest attendance during DLW measurement periods. We found that TS kittiwakes took on average 20 times longer to return to their nest than SS kittiwakes after initial treatment, and nest attendance was reduced by about 40% relative to control birds. In contrast, nest attendance did not differ between control and SS kittiwakes. Estimates of energy expenditure of SS kittiwakes exceeded those of TS kittiwakes by 15%. This difference was probably caused by TS birds remaining inactive for extended time periods while at sea. Our results demonstrate that the common assumption that the TS DLW method has little impact on the behaviour of study subjects is in some circumstances fallacious. Estimates of energy expenditure derived by the SS approach may thus more accurately reflect unbiased rates of energy expenditure. However, the choice of protocol may be a trade-off between their impact on behaviour, and hence accuracy, and their differences in precision. Adopting procedures that minimize the impact of TS protocols may be useful.