Due to a software translation incompatibility, errors were introduced in the final version of the manuscript submitted for the above publication.

A class 1 aldehyde dehydrogenase (ALDH) catalyzes oxidation of retinaldehyde to retinoic acid in bovine retina. We used immunocytochemistry and in situ hybridization to localize this enzyme in adult and fetal bovine retinas. Specific ALDH immunoreactivity was present in the cytoplasm of wide-field amacrine cells restricted in distribution to the dorsal part of the adult retina. The somata diameters ranged from ∼8 μ to ∼15 μ, and the cells increased in density from ∼125 cells/mm2 near the horizontal meridian to ∼425 cells/mm2 in the superior far periphery. The ALDH-positive cells had somata on both sides of the inner plexiform layer (IPL) and processes in two IPL strata. The majority of ALDH-positive cells were unreactive with antibodies against known amacrine cell enzymes and neurotransmitters, including GABA and glycine. The ALDH-positive amacrine cells also did not react with anti-cellular retinoic acid-binding protein, which was present in a subset of GABA-positive amacrine cells. In flat-mounted retinas processed by in situ hybridization, the larger ALDH-positive amacrine cells tended to be more heavily labeled. In addition to amacrine cells, Müller cell processes in the inner retina were weakly immunoreactive for ALDH; however, these glial cells did not contain ALDH mRNA. The pattern of ALDH expression in fetal bovine retinas was documented by immunocytochemistry. No ALDH reactivity was found before 5.5 months; for the remainder of the fetal period, ALDH immunoreactivity was present in amacrine cells similar to those in adult retina. The ALDH-positive amacrine cells in bovine retina are novel, being limited in distribution to the dorsal retina and unlabeled with other amacrine cell-specific markers. Identification of ALDH in amacrine cells provides additional evidence that cells of the inner retina are involved in retinoid metabolism.

We recently proposed a model of cytogenesis which assumes that primitive ancestral mammals and premammalian vertebrates had a retinal composition that consisted of about seven neurons per Müller cell, comprising 1–2 cone photoreceptors, 1–2 rod photoreceptors, 2–3 bipolar cells, 1–2 amacrine cells, less than 1 ganglion cell, and less than 1 horizontal cell (Reichenbach & Robinson, 1995). The Australian lungfish (Neoceratodus forsten) closely resembles the lobe-finned ancestors of land vertebrates, and has an extremely plesiomorphic nervous system. The present study, therefore, has examined the relative frequencies of retinal neurons and Müller cells (identified by immunolabelling for glutamine synthetase) in the lungfish retina. It was found that for each Müller cell there is an average of 1.9 cone photoreceptors, 1.7 rod photoreceptors, 3.1 amacrine/bipolar/horizontal cells, and 0.6 ganglion cells; amounting to a ratio of 7.3 neurons per Müller cell. These results support our conjecture that the sequence of cytogenesis in mammals is constrained by a developmental program that predates the evolution of mammals. The study also provides the first detailed morphological descriptions of lungfish Müller cells and their relationship with adjacent neurons. It was found that individual Müller cells in lungfish have a volume (more than 12,000 μm3) that is an order of magnitude higher than in mammals, yet the proportion of total retinal volume occupied by these cells (20%) is very similar.

Alpha ganglion cells from the midperiphery of the rabbit retina were recorded intracellularly under visual control, in a superfused everted eyecup, and labeled with HRP. Their physiology and large somata with broad dendritic arbors identified them as uniform populations of ON- and OFF-center alpha ganglion cells, which typically displayed transient/sustained light-evoked responses. When dark adapted, the light-evoked responses from both ON- and OFF-center alpha ganglion cells were more sustained than those generally seen under light-adapted conditions. During dark-adapted (scotopic) conditions, stimulation with dim full-field illumination and small spots, either positioned over the soma or displaced 450 μm from the soma, all elicited pure center responses. After light adaptation (photopic conditions), the displaced small spots that previously evoked center responses elicited antagonistic surround responses from both ON- and OFF-center cells. Thus, as originally described in cat retina (Barlow et al., 1957), the receptive-field organization of ganglion cells changed between dark and light adaptation, and an absence or presence of surround antagonism was indicative of scotopic versus photopic states.

The receptive-field properties of retinal ganglion cells (RGCs) provide information about early visual processing. In the primate retina, P cells form the largest class of RGCs (Rodieck, 1988). A detailed exploration of the dynamics of the two subdivisions of the P-cell receptive field—the center and the surround—was undertaken. In the preceding paper (Benardete & Kaplan, 1996), the first-order responses of the center and the surround of P cells were described, which were obtained with a new technique, the multiple m-sequence stimulus (Benardete & Victor, 1994). In this paper, the investigation of P-cell responses measured as S-potentials in the lateral geniculate nucleus (LGN) is continued, and significant nonlinear, second-order responses from the center and the surround are described. These responses are quantified by fitting a mathematical model, the linear-nonlinear-linear (LNL) model (Korenberg, 1973; Korenberg & Hunter, 1986; Victor, 1988) to the data. In a second series of experiments, demonstration that steady illumination of the surround modifies the gain of the center to contrast signals (see also Kaplan & Shapley, 1989) is made. In P ON cells, increasing the steady illumination of the surround decreases the gain and speeds up the center's first-order response. In P OFF cells, increasing the steady illumination of the surround increases the gain of the center while speeding up the response. The results of both sets of experiments are related to the known anatomy and physiology of the P cell.

arises from the integration of signals from strongly oscillatory cells within the LGN. The model also predicts the incidence of 50-Hz oscillatory cells within the cortex. Oscillatory discharge around 30 Hz is explained in a second model by the presence of intrinsically oscillatory cells within cortical layer 5. Both models generate spike trains whose power spectra and mean firing rates are in close agreement with experimental observations of simple and complex cells. Considered together, the two models can largely account for the nature and incidence of oscillatory discharge in the cat's visual cortex. The validity of these models is consistent with the possibility that oscillations are generated independently of intracortical interactions. Because these models rely on intrinsic stimulus-independent oscillators within the retina and cortex, the results further suggest that oscillatory activity within the cortex is not necessarily associated with the processing of high-order visual information.