Stemmata, the eyes of holometabolous insect larvae, have gained little attention, even though they exhibit remarkably different optical solutions, ranging from compound eyes with upright images, to sophisticated single-chamber eyes with inverted images. Such optical differences raise the question of how major transitions may have occurred. Stemmata evolved from compound eye ancestry, and optical differences are apparent even in some of the simplest systems that share strong cellular homology with adult ommatidia. The transition to sophisticated single-chamber eyes occurred many times independently, and in at least two different ways: through the fusion of many ommatidia [as in the sawfly (Hymenoptera)], and through the expansion of single ommatidia [as in tiger beetles (Coleoptera), antlions (Neuroptera) and dobsonflies (Megaloptera)]. Although ommatidia-like units frequently have multiple photoreceptor layers (tiers), sophisticated image-forming stemmata tend to only have one photoreceptor tier, presumably a consequence of the lens only being able to efficiently focus light on to one photoreceptor layer. An interesting exception is found in some diving beetles [Dytiscidae (Coleoptera)], in which two retinas receive sharp images from a bifocal lens. Taken together, stemmata represent a great model system to study an impressive set of optical solutions that evolved from a relatively simple ancestral organization.