In primates, short wavelength sensitive cones (S cones) and medium- or long-wavelength-sensitive cones (L/M cones) are two separate populations. Each cone type has a different developmental timecourse, contributes to different intra-retinal circuits, and transmits different types of information to the brain. However, in fetal human retina a significant population of cones express both S and L/M opsin (S+L/M cones), raising questions about whether S+L/M cones die or change opsin expression during development. We have utilized fetal, postnatal and adult human retinae to study the immunohistochemical distribution and morphology of S+L/M cones during development. Because S cones appear to be at higher density in fetal compared to adult retinae, we used antibodies to S opsin and alpha-transducin to estimate the proportion of S-cones, and TUNEL labelling to detect apoptotic death in the L/M, S or S+L/M population during development. S cones were present in central retina from fetal week (Fwk)11 and covered the retina by Fwk20. L/M cones appeared in the foveal cone mosaic 3-4 weeks after S-opsin was first detected, and covered the retina by birth. S+L/M cones were detected in all retinae older than Fwk14. They were most numerous at the retinal eccentricity where L/M opsin was just appearing; i.e. at the 'front' of L/M opsin expression. In this region, five morphological types of cones were present. (1) Heavily labelled S cones had thick cell bodies, a thick basal axon and pedicle, and a nucleus at any level of the outer nuclear layer (ONL). (2) Heavily labelled L/M cones were wine goblet shaped with a small round cell body, a large nucleus at the outer ONL edge, and a thin axon with a prominent synaptic pedicle. (3) Goblet-shaped S+L/M cones. (4) Goblet-shaped cones lightly labelled for S-opsin. (5) Cones that were not immunoreactive to either opsin. Only type 1 S cones were present peripheral to the L/M expression front, and their labelling intensity, morphology and distribution indicates that these are the 'true blue' cones of the adult mosaic. Only type 2 L/M cones were present in the foveal cone mosaic. Types 3 and 4 were most numerous within 500-750 microm of the L/M expression front, but type 3 S+L/M cones were also scattered throughout more central regions in fetal, infant and adult retinae. S+L/M cones comprised 5-10% of opsin immunoreactive cones at the L/M front in fetal and early postnatal retinas but 0.01-0.03% throughout P8mo and adult retinae. We found no evidence of significant levels of apoptosis in L/M cones at the expression front, suggesting that this decrease was not due to cell death. The findings suggest that goblet-shaped cones destined to express L or M opsin may initially and transiently express S opsin. Near the optic disc, at Fwk17 S cone density was around 2000 cells mm(-2), which dropped 50% by Fwk20 and stabilized at around 500 cells mm(-2) by birth. Double labelling with alpha-transducin showed that throughout this period 8-10% of all cones expressed S opsin. TUNEL labelling found no significant apoptosis in the S cone population. The decrease in S cone density near the optic disc occurs in the absence of apoptosis, and is likely due to other developmental events acting on the photoreceptor layer, including displacement of cones towards the fovea.