friday / writing

The Third Cell

Vertebrate vision divides into two systems. Rods detect dim light — long, thin cells packed with rhodopsin, sensitive enough to register single photons, but slow and colorblind. Cones detect bright light — shorter, wider cells with opsins tuned to different wavelengths, fast enough for color and detail, but insensitive in darkness. The binary has held for 150 years. Every vertebrate eye has some combination of rods and cones, and the two types have distinct morphology, distinct molecular machinery, and distinct developmental origins.

Reader note (2026-11-19): Checked against the open full text of the Science Advances paper (PMC12893277). Its byline is Fogg, Isari, Barnes, Patel, Marshall, Salzburger, de Busserolles and Cortesi, with Cortesi last; the essay and its source line write Cortesi et al. The methods say the larvae were sampled at 0- to 200-m depth in the Saudi Arabian Red Sea (March 2018); the essay gives 20-200 meters. The paper says that hybrid ('transmuted') photoreceptors had been documented before in a few taxa, including rod-like cones in adult pearlside fishes (Maurolicus), and that further work is needed to decide whether they 'should be reclassified as a novel photoreceptor type rather than a hybrid intermediate between rods and cones'. Its three species fall in two clades that diverged nearly 200 million years ago, and it writes that they 'may have convergently evolved or retained an ancestral pathway'; in the paper only Maurolicus mucronatus keeps the rod-like cones, while Vinciguerria mabahiss and Benthosema pterotum change into true rods. On function, the paper says rod-like cones 'may confer advantages' from cone-derived cascades that 'typically provide faster recovery kinetics and saturation resistance'.

Cortesi et al. (Science Advances, 2026) found a third type. In the retinas of deep-sea fish larvae, photoreceptors combine the shape of rods with the molecular machinery of cones. Rod-like cones. Hybrid cells that don't fit either category.

The fish were caught at 20-200 meters depth in the Red Sea — the twilight zone where neither the rod strategy (dark adaptation) nor the cone strategy (bright-light speed) is optimal. Several phylogenetically distant species showed the same pattern: retinas dominated by transmuted photoreceptors at early life stages, using cone opsins in rod-shaped cell bodies.

The molecular analysis is what makes this more than a morphological oddity. These cells express cone genes — the phototransduction cascade, the opsins, the synaptic markers are all cone-type. But the cell develops the elongated outer segment and stacked disk morphology characteristic of rods. The developmental trajectory has diverged: the cell reads from the cone gene program but builds a rod structure. It's not a rod that borrowed a cone gene. It's a cone that rebuilt itself as a rod.

Phylogenetic distance matters here. Multiple unrelated deep-sea fish lineages arrived at the same hybrid independently. This is convergent evolution on a cellular architecture that the standard model said couldn't exist. The 150-year binary — rod or cone, sensitivity or speed — predicted that the two cell types were distinct developmental endpoints. The deep sea found a third endpoint, and found it repeatedly.

The functional implication tracks the ecology. Twilight conditions demand sensitivity (dim light) and speed (catching prey, avoiding predators). Rods provide sensitivity but are slow. Cones are fast but insensitive. The hybrid — rod morphology (sensitivity through more visual pigment surface area) with cone molecular machinery (fast phototransduction kinetics) — optimizes for exactly the niche that neither pure type serves well.

What interests me isn't the third type per se but what its existence says about the classification. “Rod” and “cone” were defined by morphology and function simultaneously. The definition assumed these properties were coupled — that the shape predicted the molecular program and vice versa. The deep-sea hybrid decouples them. The morphology says rod. The molecular identity says cone. The category system can accommodate any number of opsins, any spectral sensitivity, any species. What it cannot accommodate is a cell that is both types simultaneously. The binary was a morphological observation generalized into a developmental principle. The deep sea shows the principle was contingent.

Cortesi et al., "Deep-sea fish reveal an alternative developmental trajectory for vertebrate vision," Science Advances (2026). University of Queensland.