Predator-prey interactions play an important role in determining the dynamics of pelagic ecosystems. Human intervention in such interactions can have effects that cascade throughout these ecosystems. Recently, concerns have arisen due to the commercial harvesting of Euphausia superba, the Antarctic krill, a keystone prey species in the Southern Ocean food web. It has been difficult to evaluate these concerns because of problems associated with determining the availability of krill to their natural predators. Here, we report a novel method for assessing prey availability to an important krill predator, Pygoscelis antarctica, the chinstrap penguin. Acoustic techniques were used to analyze the three-dimensional distribution of krill within a 1852 X 1852 X 100 m volume of ocean. Our study revealed the presence of at least six, distinct krill aggregations and substantial vertical and horizontal patchiness at scales of 10s to 100s of metres. Monte Carlo simulations revealed significant spatial concordance between the surface distribution of penguins and krill distributed in the 30 - 40 m depth layer; no spatial concordance was detected in shallower layers. We conclude that fine-scale, depth-dependent patterns of kill patchiness must be incorporated into analyses of krill availability to predator populations if we hope to predict predator responses to a changing food environment.