This study examines the cognitive costs of navigation aids in a hypermedia learning task. In a 2(navigable vs. non-navigable) x 2(map vs. content list) experimental design cognitive requirements were measured by users' eye movement data. Additionally, data from users' navigation operations, knowledge acquisition, and subjective evaluation of the hypermedia system were acquired. Results show that cognitive requirements were higher for the map than for the content list, and higher for a non-navigable than a navigable aid. Furthermore, users' navigation behavior was less efficient with a non-navigable than with a navigable aid. Recommendations for the design of aids in hypermedia systems are discussed. (Contains 6 figures and 1 table.)