During visual fixation, small eye movements keep the retinal image continuously in motion. It is known that neurons in the visual system are sensitive to the spatiotemporal modulations of luminance resulting from this motion. In this study, we examined the influence of fixational eye movements on the statistics of neural activity in the macaque's retina during the brief intersaccadic periods of natural visual fixation. The responses of parvocellular (P) and magnocellular (M) ganglion cells in different regions of the visual field were modeled while their receptive fields scanned natural images following recorded traces of eye movements. Immediately after the onset of fixation, wide ensembles of coactive ganglion cells extended over several degrees of visual angle, both in the central and peripheral regions of the visual field. Following this initial pattern of activity, the covariance between the responses of pairs of P and M cells and the correlation between the responses of pairs of M cells dropped drastically during the course of fixation. Cell responses were completely uncorrelated by the end of a typical 300-ms fixation. This dynamic spatial decorrelation of retinal activity is a robust phenomenon independent of the specifics of the model. We show that it originates from the interaction of three factors: the statistics of natural scenes, the small amplitudes of fixational eye movements, and the temporal sensitivities of ganglion cells. These results support the hypothesis that fixational eye movements, by shaping the statistics of retinal activity, are an integral component of early visual representations.

The mammalian neural retina contains single or multiple intrinsic circadian oscillators that can be directly entrained by light cycles. Dopaminergic amacrine (DA) cells represent an especially interesting candidate as a site of the retinal oscillator because of the crucial role of dopamine in light adaptation, and the widespread distribution of dopamine receptors in the retina. We hereby show by single-cell, end-point RT-PCR that retinal DA cells contain the transcripts for six core components of the circadian clock: Bmal1, Clock, Cry1, Cry2, Per1, and Per2. Rod photoreceptors represented a negative control, because they did not appear to contain clock transcripts. We finally confirmed that DA cells contain the protein encoded by the Bmal1 gene by comparing immunostaining of the nuclei of DA cells in the retinas of wildtype and Bmal1−/− mice. It is therefore likely that DA cells contain a circadian clock that anticipates predictable variations in retinal illumination.

Simple (dorsal light reflex) and complex (predator-prey interactions) visually mediated behaviors were used concurrently with morphological examination to assess restoration of visual function following optic nerve crush in bluegill (Lepomis macrochirus) × pumpkinseed (Lepomis gibbosus) hybrid sunfish. Regenerating optic nerve axons projected into the stratum opticum-stratum fibrosum et griseum superficiale by week 2, the stratum griseum centrale by week 4, and stratum album centrale by week 6. Initial projections into the laminae were diffuse and less stratified compared to controls. By week 12, the projection pattern of regenerating nerve fibers closely resembled the innervation of normal tecta. Visual improvements were correlated with increasing projections into the tectum. The dorsal light reflex improved from a 45° vertical deviation following nerve crush to 4.5° by week 16. Initial predator-prey interactions were exclusively mediated by the control eye. As regeneration progressed, there was a gradual expansion of the visual field. The reaction distance and attack angles within the visual field of the experimental eye were initially less than controls, however, these differences disappeared by week 10. Improvements in visual function were closely correlated with an increase of regenerating ganglion cell axons into the optic tectum indicating sufficient synaptogenesis to mediate both simple and complex visual behavior.

The following two Review Articles that were published in 2006 in Volume 23 of Visual Neuroscience, were incorrectly categorized in the table of contents as Research Articles.

Spatial vision was studied in the bluegill sunfish, Lepomis macrochirus (9.5–14 cm standard length) to assess the limitations imposed by the optics of the eye, the retinal receptor spacing and the retinotectal projection during regeneration. Examination of images formed by the dioptric elements of the eye showed that spatial frequencies up to 29 c/° could be imaged on the retina. Cone spacing was measured in the retina of fresh, intact eyes. The spacing of rows of double cones predicted 3.4 c/° as the cutoff spatial frequency; the spacing between rows of single and double cones predicted 6.7 c/°. Contrast sensitivity functions were obtained psychophysically in normals and fish with one regenerating optic nerve. Fish were trained to orient to gratings (mean luminance = 25 cd/m2) presented to either eye. In normals, contrast sensitivity functions were similar in shape and bandwidth to those of other species, peaking at 0.4 c/° with a minimum contrast threshold of 0.03 and a cutoff at about 5 c/°, which was within the range predicted by cone spacing. Given that the optical cutoff frequency exceeds that predicted by cone spacing, it is possible that gratings could be detected by aliasing with the bluegill's regular cone mosaic. However, tests with high contrast gratings up to 15 c/° found no evidence of such detection. After crushing one optic nerve in three trained sunfish, recovery of visual avoidance, dorsal light reflex and orienting to gratings, were monitored over 315 days. At 64–69 days postcrush, responses to gratings reappeared, and within 2–5 days contrast sensitivity at low (0.15 c/°) and medium (1.0 c/°) spatial frequencies had returned to normal. At a high spatial frequency (2.93 c/°) recovery was much slower, and complete only in one fish.

Fatty acid amide hydrolase (FAAH) is the degradative enzyme for anandamide (AEA), an endogenous ligand for the vanilloid receptor (TRPV1) and cannabinoid receptor 1. As FAAH and TRPV1 are integral membrane proteins, FAAH activity could modulate the availability of AEA for TRPV1 activation. Previous studies in this laboratory reported an extensive endocannabinoid system in goldfish retina. Immunocytochemistry was used to determine the relative distributions of FAAH-immunoreactivity (IR) and TRPV1-IR in goldfish retina. Here, we show the first example in an intact neural system in which TRPV1-IR co-localizes in subpopulations of FAAH-immunoreactive neurons, in this case amacrine cells. These cells are rare and consist of three subtypes: 1. ovoid cell with granular-type dendrites restricted to sublamina a, 2. pyriform cell with smooth processes in sublamina b, and 3. fusiform cell with smooth processes that project to sublaminae a and b. The varied appearances of reaction product in the dendrites suggest different subcellular localization of TRPV1, and hence function of FAAH activity regarding TRPV1 stimulation among the cell types. Ovoid and pyriform amacrine cells, but not fusiform cells, labeled with GAD-IR and constituted subsets of GABAergic amacrine cells. TRPV1 amacrine cells, though rare, are represented in the ON, OFF and ON/OFF pathways of the retina. As TRPV1 stimulation increases intracellular calcium with numerous downstream effects, co-localization of TRPV1 and FAAH suggests an autoregulatory function for anandamide. Due to the rarity of these cells, the three vanilloid amacrine cell types may be involved in global effects rather than feature extraction, for example: sampling of ambient light or maintaining homeostasis.

Many retinal ganglion cells are coupled via gap junctions with neighboring amacrine cells and ganglion cells. We investigated the extent and dynamics of coupling in one such network, the OFF α ganglion cell of rabbit retina and its associated amacrine cells. We also observed the relative spread of Neurobiotin injected into a ganglion cell in the presence of modulators of gap junctional permeability. We found that gap junctions between amacrine cells were closed via stimulation of a D1 dopamine receptor, while the gap junctions between ganglion cells were closed via stimulation of a D2 dopamine receptor. The pairs of hemichannels making up the heterologous gap junctions between the ganglion and amacrine cells were modulated independently, so that elevations of cAMP in the ganglion cell open the ganglion cell hemichannels, while elevations of cAMP in the amacrine cell close its hemichannels. We also measured endogenous dopamine release from an eyecup preparation and found a basal release from the dark-adapted retina of approximately 2 pmol/min during the day. Maximal stimulation with light increased the rate of dopamine release from rabbit retina by 66%. The results suggest that coupling between members of the OFF α ganglion cell/amacrine cell network is differentially modulated with changing levels of dopamine.

Horizontal cells are second order neurons that receive direct synaptic input from photoreceptors. In teleosts horizontal cells can be divided into two categories, cone-connected and rod-connected. Although the anatomy and physiology of fish cone horizontal cells have been extensively investigated, less is known about rod horizontal cells. This study was undertaken to determine whether light and/or the circadian clock regulate gap junctional coupling between goldfish rod horizontal cells. We used fine-tipped, microelectrode intracellular recording to monitor rod horizontal cells under various visual stimulation conditions, and tracer (biocytin) iontophoresis to visualize their morphology and evaluate the extent of coupling. Under dark-adapted conditions, rod horizontal cells were extensively coupled to cells of like-type (homologous coupling) with an average of ∼120 cells coupled. Under these conditions, no differences were observed between day, night, the subjective day, and subjective night. In addition, under dark-adapted conditions, application of the dopamine D2-like agonist quinpirole (1 μM), the D2-like antagonist spiperone (10 μM), or the D1-like antagonist SCH23390 (10 μM) had no effect on rod horizontal cell tracer coupling. In contrast, the extent of tracer coupling was reduced by ∼90% following repetitive light (photopic range) stimulation of the retina or application of the D1-agonist SKF38393 (10 μM) during the subjective day and night. We conclude that similarly to cone horizontal cells, rod horizontal cells are extensively coupled to one another in darkness and that the extent of coupling is dramatically reduced by bright light stimulation or dopamine D1-receptor activation. However, in contrast to cone horizontal cells whose light responses are under the control of the retinal clock, the light responses of rod horizontal cells under dark-adapted conditions were similar during the day, night, subjective day, and subjective night thus demonstrating that they are not under the influence of the circadian clock.

The Syngnathidae are specialized diurnal feeders that are known to possess a retinal fovea and use independent eye movements to locate, track, and strike individual planktonic prey items. In this study, we have investigated the spectral sensitivities of three syngnathid species: a pipefish and two seahorses. We used spectrophotometry to measure the spectral transmission properties of ocular lenses and microspectrophotometry to measure the spectral absorption characteristics of visual pigments in the retinal photoreceptors. The pipefish, Stigmatopora argus, together with the seahorse Hippocampus subelongatus, is found in “green-water” temperate coastal seagrass habitats, whereas the second seahorse, H. barbouri, originates from a “blue-water” tropical coral reef habitat. All species were found to possess short wavelength absorbing pigment(s) in their lenses, with the 50% cut-off point of S. argus and H. subelongatus at 429 and 425 nm respectively, whereas that of H. barbouri was located at 409 nm. Microspectrophotometry of the photoreceptors revealed that the rods of all three species contained visual pigment with the wavelength of maximum absorption (λmax) at approximately 500 nm. The visual pigment complement of the cones varied between the species: all possessed single cones with a λmax close to 460 nm but H. barbouri also possessed an additional class of single cone with λmax at 430 nm. Three classes of visual pigment were found in the double cones, the λmax being approximately 520, 537, and 560 nm in the two seahorses and 520, 537, and 580 nm in the pipefish. The spectral sensitivities of the syngnathids investigated here do not appear to conform to generally accepted trends for fishes inhabiting different spectral environments. The influence of the specialized feeding regime of the syngnathids is discussed in relation to our findings that ultra-violet sensitivity is apparently not necessary for zooplanktivory in certain habitats.

Many cell types in the retina are coupled via gap junctions and so there is a pressing need for a potent and reversible gap junction antagonist. We screened a series of potential gap junction antagonists by evaluating their effects on dye coupling in the network of A-type horizontal cells. We evaluated the following compounds: meclofenamic acid (MFA), mefloquine, 2-aminoethyldiphenyl borate (2-APB), 18-α-glycyrrhetinic acid, 18-β-glycyrrhetinic acid (18-β-GA), retinoic acid, flufenamic acid, niflumic acid, and carbenoxolone. The efficacy of each drug was determined by measuring the diffusion coefficient for Neurobiotin (Mills & Massey, 1998). MFA, 18-β-GA, 2-APB and mefloquine were the most effective antagonists, completely eliminating A-type horizontal cell coupling at a concentration of 200 μM. Niflumic acid, flufenamic acid, and carbenoxolone were less potent. Additionally, carbenoxolone was difficult to wash out and also may be harmful, as the retina became opaque and swollen. MFA, 18-β-GA, 2-APB and mefloquine also blocked coupling in B-type horizontal cells and AII amacrine cells. Because these cell types express different connexins, this suggests that the antagonists were relatively non-selective across several different types of gap junction. It should be emphasized that MFA was water-soluble and its effects on dye coupling were easily reversible. In contrast, the other gap junction antagonists, except carbenoxolone, required DMSO to make stock solutions and were difficult to wash out of the preparation at the doses required to block coupling in A-type HCs. The combination of potency, water solubility and reversibility suggest that MFA may be a useful compound to manipulate gap junction coupling.

To elucidate the specific properties of retinae with grouped photoreceptors the neural pathways in the outer and inner plexiform layer were studied. Photoreceptor bundles in this species consist of more than 100 rods and up to 50 cones, and are usually regarded as functional units. Golgi impregnation in thick and thin sections and light microscopy were used to identify bipolar cell types linking photoreceptors to amacrine and/or ganglion cells. Nine different types were distinguished based on their dendritic morphology and the position of the axon terminal in the inner plexiform layer. Small cells have dendritic fields smaller than the diameter of a photoreceptor bundle and are contacted mostly by cones. The dendritic field size of bushy cells matches that of a photoreceptor bundle; they are contacted mainly by rods. Flat cells receive about equal input from rods and cones; their dendritic field size exceeds the bundle diameter. Within the three major classes there are subtypes addressing three sublaminae of the inner plexiform layer, the proximal On-centre region (sl b), the distal Off-centre region (sl a) and a central sublayer (sl c) probably with transient activity. These observations suggest that cone vision has a spatial acuity better than the “bundle grain”. In rod dominated vision the resolution matches that of the bundles; for this pathway, the hypothesis of the bundle as a functional unit is confirmed. The mesopic flat cell pathway has a resolution inferior to the “bundle grain”; it may therefore be dedicated to movement detection.

We evaluated the contributions of somatic and dendritic impulses to the receptive field dimensions of amacrine cells in the amphibian retina. For this analysis, we used the NEURON simulation program with a multicompartmental, multichannel model of an On-Off amacrine cell with a three-dimensional structure obtained through computer tracing techniques. Simulated synaptic inputs were evenly spaced along the dendritic branches and organized into eight annuli of increasing radius. The first set of simulations activated each ring progressively to simulate an area summation experiment, while a second approach activated each annulus individually. Both sets of simulations were done with and without the presence of Na channels in the dendrites and soma. Unexpectedly, the receptive field dimensions observed in the area summation simulations was often smaller than that predicted from the summation of the annular simulations. Collisions of action potentials moving in opposite directions in the dendrites largely accounted for this contraction in receptive field size for the area summation studies. The presence of dendritic Na channels increased the size of the receptive field beyond that achieved in their absence and allowed the physiological size of the receptive field to approximate the physical dimensions of the dendritic tree. This receptive field augmentation was the result of impulse generating ability in the dendrites which enhanced the signal observed at the soma. These simulations provide a plausible mechanistic explanation for physiological recordings from amacrine cells that show similar phenomena.

A GABA- or glycine-induced increase in Cl− permeability can produce either a depolarization or hyperpolarization, depending on the Cl− equilibrium potential. It has been shown that retinal neurons express the chloride cotransporters, Na-K-2Cl (NKCC) and K-Cl (KCC), the primary molecular mechanisms that control the intracellular Cl− concentration. We thus studied (1) the localization of these cotransporters in the fish retina, and (2) how suppression of cotransporter activity in the fish retina affects function. Specific antibodies against NKCC and KCC2 revealed that both cotransporters were expressed in the outer and inner plexiform layers, and colocalized in many putative amacrine cells and in cells of the ganglion cell layer. However, the somata of putative horizontal cells displayed only NKCC immunoreactivity and many bipolar cells were only immunopositive for KCC2. In the outer retina, application of bumetanide, a specific inhibitor of NKCC activity, (1) increased the steady-state extracellular concentration of K+ ([K+]o) and enhanced the light-induced decrease in the [K+]o, (2) increased the sPIII photoreceptor-dependent component of the ERG, and (3) reduced the extracellular space volume. In contrast, in the outer retina, application of furosemide, a specific inhibitor of KCC activity, decreased sPIII and the light-induced reduction in [K+]o, but had little effect on steady-state [K+]o. In the inner retina, bumetanide increased the sustained component of the light-induced increase in [K+]o. These findings thus indicate that NKCC and KCC2 control the [K+]o and extracellular space volume in the retina in addition to regulating GABA- and glycine-mediated synaptic transmission. In addition, the anatomical and electrophysiological results together suggest that all of the major neuronal types in the fish retina are influenced by chloride cotransporter activity.

Connexin 35/36 is the most widespread neuronal gap junction protein in the retina and central nervous system. Electrical and/or tracer coupling in a number of neuronal circuits that express this connexin are regulated by light adaptation. In many cases, the regulation of coupling depends on signaling pathways that activate protein kinases such as PKA, and Cx35 has been shown to be regulated by PKA phosphorylation in cell culture systems. To examine whether phosphorylation might regulate Cx35/36 in the retina we developed phospho-specific polyclonal antibodies against the two regulatory phosphorylation sites of Cx35 and examined the phosphorylation state of this connexin in the retina. Western blot analysis with hybrid bass retinal membrane preparations showed Cx35 to be phosphorylated at both the Ser110 and Ser276 sites, and this labeling was eliminated by alkaline phosphatase digestion. The homologous sites of mouse and rabbit Cx36 were also phosphorylated in retinal membrane preparations. Quantitative confocal immunofluorescence analysis showed gap junctions identified with a monoclonal anti-Cx35 antibody to have variable levels of phosphorylation at both the Ser110 and Ser276 sites. Unusual gap junctions that could be identified by their large size (up to 32 μm2) and location in the IPL showed a prominent shift in phosphorylation state from heavily phosphorylated in nighttime, dark-adapted retina to weakly phosphorylated in daytime, light-adapted retina. Both Ser110 and Ser276 sites showed significant changes in this manner. Under both lighting conditions, other gap junctions varied from non-phosphorylated to heavily phosphorylated. We predict that changes in the phosphorylation states of these sites correlate with changes in the degree of coupling through Cx35/36 gap junctions. This leads to the conclusion that connexin phosphorylation mediates changes in coupling in some retinal networks. However, these changes are not global and likely occur in a cell type-specific or possibly a gap junction-specific manner.

The spatial and temporal interactions in the receptive fields of On-Off directionally selective (DS) ganglion cells endow them with directional selectivity. Using a variety of stimuli, such as sinusoidal gratings, we show that these interactions make directional selectivity of the DS ganglion cell robust with respect to stimulus parameters such as contrast, speed, spatial frequency, and extent of motion. Moreover, unlike the directional selectivity of striate-cortex cells, On-Off DS ganglion cells display directional selectivity to motions not oriented perpendicularly to the contour of the objects. We argue that these cells may achieve such high robustness by combining multiple mechanisms of directional selectivity.

Lungfish (order Dipnoi) evolved during the Devonian period and are believed to be the closest living relatives to the land vertebrates. Here we describe the previously unknown morphology of the lungfish eye in order to examine ocular adaptations present in early sarcopterygian fish. Unlike many teleosts, the Australian lungfish Neoceratodus forsteri possesses a mobile pupil with a slow pupillary response similar to amphibians. The structure of the eye changes from juvenile to adult, with both eye and lens becoming more elliptical in shape with growth. This change in structure results in a decrease in focal ratio (the distance from lens center to the retina divided by the lens radius) and increased retinal illumination in adult fish. Despite a degree of lenticular correction for spherical aberration, there is considerable variation across the lens. A re-calculation of spatial resolving power using measured focal ratios from cryosectioning reveals a low ability to discriminate fine detail. The dipnoan eye shares more features with amphibian eyes than with most teleost eyes, which may echo the visual needs of this living fossil.

Absorbance spectra of rods and some cones were measured by microspectrophotometry in 22 fish species from the brackish-water of the Baltic Sea, and when applicable, in the same species from the Atlantic Ocean (3 spp.), the Mediterranean Sea (1 sp.), or Finnish fresh-water lakes (9 spp.). The main purpose was to study whether there were differences suggesting spectral adaptation of rod vision to different photic environments during the short history (<104 years) of postglacial isolation of the Baltic Sea and the Finnish lakes. Rod absorbance spectra of the Baltic subspecies/populations of herring (Clupea harengus membras), flounder (Platichthys flesus), and sand goby (Pomatoschistus minutus) were all long-wavelength-shifted (9.8, 1.9, and 5.3 nm, respectively, at the wavelength of maximum absorbance, λmax) compared with their truly marine counterparts, consistent with adaptation for improved quantum catch, and improved signal-to-noise ratio of vision in the Baltic light environment. Judged by the shape of the spectra, the chromophore was pure A1 in all these cases; hence the differences indicate evolutionary tuning of the opsin. In no species of fresh-water origin did we find significant opsin-based spectral shifts specific to the Baltic populations, only spectral differences due to varying A1/A2 chromophore ratio in some. For most species, rod λmax fell within a wavelength range consistent with high signal-to-noise ratio of vision in the spectral conditions prevailing at depths where light becomes scarce in the respective waters. Exceptions were sandeels in the Baltic Sea, which are active only in bright light, and all species in a “brown” lake, where rod λmax lay far below the theoretically optimal range.

Large field motion detection in goldfish, measured in the optomotor response, is based on the L-cone type, and is therefore color-blind (Schaerer & Neumeyer, 1996). In experiments using a two-choice training procedure, we investigated now whether the same holds for the detection of a small moving object (size: 8 mm diameter; velocity: 7 cm/s). In initial experiments, we found that goldfish did not discriminate between a moving and a stationary stimulus, obviously not taking attention to the cue “moving.” Therefore, random dot patterns were used in which the stimulus was visible only when moving. Using black and white random dot patterns with variable contrast between 0.2 and 1, we found that the fish could see motion only with high (0.8) contrast. In the decisive experiment, a red-green random dot pattern was used. By keeping the intensity of the red dots constant and reducing the intensity of the green dots, a narrow intensity range was found in which goldfish could no longer discriminate between the moving random dot stimulus in random dot surround and the stationary random dot pattern. The same was the case when a red moving disk was presented in green surround. This is the evidence that object motion is red-green color blind, i.e., color information cannot be used to detect the moving object. Calculations of the cone excitation values revealed that the M-cone type is decisive, as this cone type (and not the L-cone type) is not modulated by that particular red-green pattern in which the moving stimulus was invisible.

The exposure of the developing human embryo to ethanol results in a spectrum of disorders involving multiple organ systems, including the visual system. One common phenotype seen in humans exposed to ethanol in utero is microphthalmia. The objective of this study was to describe the effects of ethanol during retinal neurogenesis in a model organism, the zebrafish, and to pursue the potential mechanisms by which ethanol causes microphthalmia. Zebrafish embryos were exposed to 1% or 1.5% ethanol from 24 to 48 h after fertilization, a period during which the retinal neuroepithelium undergoes rapid proliferation and differentiation to form a laminated structure composed of different retinal cell types. Ethanol exposure resulted in significantly reduced eye size immediately following the treatment, and this microphthalmia persisted through larval development. This reduced eye size could not entirely be accounted for by the accompanying general delay in embryonic development. Retinal cell death was only slightly higher in ethanol-exposed embryos, although cell death in the lens was extensive in some of these embryos, and lenses were significantly reduced in size as compared to those of control embryos. The initiation of retinal neurogenesis was not affected, but the subsequent waves of cell differentiation were markedly reduced. Even cells that were likely generated after ethanol exposure—rod and cone photoreceptors and Müller glia—were delayed in their expression of cell-specific markers by at least 24 h. We conclude that ethanol exposure over the time of retinal neurogenesis resulted in persistent microphthalmia due to a combination of an overall developmental delay, lens abnormalities, and reduced retinal cell differentiation.

The morphological characteristics of the eyes and the retinae of lanternfish larvae of Lampanyctus crocodilus, Benthosema glaciale, and Myctophum punctatum were analyzed in pre-flexion, flexion, and post-flexion stages. Pre-flexion larvae of L. crocodilus, the species with the shallowest depth distribution, had spherical eyes located antero-laterally on a strongly laterally-compressed head, suggesting a forward binocular visual field. B. glaciale and M. punctatum larvae live deeper in the water column and had eyes elongated in the dorsal-ventral plane. The eyes of B. glaciale were prominent, projecting slightly outward from a laterally-compressed head, suggesting a strongly laterally-directed visual field. M. punctaum had stalked elongated eyes projecting from a dorso-ventrally flattened head. The eyes can be freely rotated allowing lateral, anterior and dorsally-directed vision. A prominent choroidal gland was situated beneath the ventral portion of the eye in M. punctatum and B. glaciale, while a smaller gland was present in the dorsal and ventral portions of the eye of L. crocodilus. In pre-flexion stage larvae, the retina of all three species was differentiated with numerous rod photoreceptors in the peripheral retinal areas and fewer cone photoreceptors mainly distributed in the central retina. This distribution suggests concomitant enhancement of scotopic sensitivity in the vertical visual plane and improved photopic acuity in the lateral and forward visual directions. The concurrent development of cones and rods, as observed in the pre-flexion stage of myctophid larvae, is consistent with meeting the special demands of visual planktivory in sub-surface waters. During larval development a gradual increase of ROS length was also accompanied by a progressive loss of cones that were almost totally absent in post-flexion larvae. This can be interpreted as an adaptive response to an impending deep mesopelagic adult life.