Midwater predators often have double eyes consisting of a large upward-poin
ting part with a narrow field of view and high resolution, and a small down
ward-pointing part with a wide field of view and low resolution. III crusta
ceans with compound eyes the different eye parts are of basically similar c
onstruction, but in fishes the downward-pointing regions may employ unusual
optical systems with unknown image-forming capabilities. It has been sugge
sted that the upward-directed parts are used to detect silhouettes of anima
ls against the residual daylight, whereas the lower parts look out for lumi
nescent organisms. Here I calculate the sizes that apposition compound eyes
would need to attain in order to fulfil these tasks, and the way that size
should vary with depth. It is concluded that silhouette detection is much
the more demanding task, and becomes increasingly difficult as light levels
decrease. For this reason the upward-pointing parts must increase rapidly
with depth. This is not the case with luminescence detectors, where the tas
k is most difficult near the surface because of upwelling background light,
and becomes easier with depth. On the whole these predictions fit well wit
h the sizes and shapes of real midwater eyes, especially in the case of the
hyperiid amphipods.