Foraging animals are expected to adjust their path according to the hierarchical spatial distribution of food resources and environmental factors. Studying such behaviour requires methods that allow for the detection of changes in pathways' characteristics across scales, i.e. a definition of scale boundaries and techniques to continuously monitor the precise movement of the animal over a sufficiently long period. We used a recently developed application of fractals, the changes in fractal dimension within a path and applied it to foraging trips over scales ranging across five orders of magnitude (10 m to 1000 km), using locations of wandering albatrosses (Diomedea exulans) recorded at 1 s intervals with a miniaturized global positioning system. Remarkably, all animals consistently showed the same pattern: the use of three scaledependent nested domains where they adjust tortuosity to different environmental and behavioural constraints. At a small scale (ca. 100 m) they use a zigzag movement as they continuously adjust for optimal use of wind; at a medium scale (1–10 km), the movement shows changes in tortuosity consistent with food-searching behaviour; and at a large scale (greater than 10 km) the movement corresponds to commuting between patches and is probably influenced by large-scale weather systems. Our results demonstrate the possibility of identifying the hierarchical spatial scales at which long-ranging animals adjust their foraging behaviour, even in featureless environments such as oceans, and hence how to relate their movement patterns to environmental factors using an objective mathematical approach.