Fig. 2

a, b Steps in the conversion of the frontal eye, with cells arranged in rows, to a two-dimensional receptor array; simplified to show one row each of pigment cells (black), photoreceptors (blue), retinal interneurons (red) and output neurons, i.e., ganglion cell homologs (yellow). a The frontal eye in side view, showing the retinal and ganglion cell homologs being produced in sequence (arrows) from a precursor which then terminally differentiates into a photoreceptor. This is a reasonable assumption, given the morphology, but has yet to be proven. All cells contact the ventricular surface in this arrangement, so only one-dimensional receptor arrays are possible. b For comparison, a hypothetical, more advanced condition where the precursors have proliferated in two dimensions, i.e., in both the plane of this section and perpendicular to it. Their retinal and output-cell progeny are now produced by basal divisions, and so lack any connection to the ventricular surface of the neural epithelium. This yields a multilayered two-dimensional array where every photoreceptor has direct access to an output pathway. c–e A hypothetical sequence to show how an amphioxus-type frontal eye (c, colors as above) might expand by increasing the length of each transverse row and begin separating into two fields (d) to produce, with the organizational changes shown in (b), a flat layered array (shown in purple to indicate a sandwich of blue, red and yellow layers, with a rim of pigment) positioned on either side of the head of a hypothetical early vertebrate (e). The frontal eye as shown in (c) could itself be reduced from a larger structure in ancestral protochordates, but there is no evidence currently to suggest it was ever multilayered. Abbreviations: neuropore, np; rostral nerve, rn