A psychophysical analog to cortical receptive-field end-stopping has been demonstrated previously in spatial filters tuned to a wide range of spatial frequencies (Yu & Levi, 1997a). The current study investigated tuning characteristics in psychophysical spatial filter end-stopping. When a D6 (the sixth derivative of a Gaussian) target is masked by a center mask (placed in the putative spatial filter center), two end-zone masks (placed in the filter end-zones) reduce thresholds. This “end-stopping” effect (the reduction of masking induced by end-zone masks) was measured at various spatial frequencies and orientations of end-zone masks. End-stopping reached its maximal strength when the spatial frequency and/or orientation of the end-zone masks matched the spatial frequency and/or orientation of the target and center mask, showing spatial-frequency tuning and orientation tuning. The bandwidths of spatial-frequency and orientation tuning functions decreased with increasing target spatial frequency. At larger orientation differences, however, end-zone masks induced a secondary facilitation effect, which was maximal when the spatial frequency of end-zone masks equated the target spatial frequency. This facilitation effect might be related to certain types of contour and texture perception, such as perceptual pop-out.

Immunocytochemistry was used to study the distribution of gamma-aminobutyric acid (GABA) throughout the central visual nuclei and retina in Rana pipiens. In the retina, GABA immunoreactivity (both somata and fibers) was observed in all layers except the outer nuclear layer (ONL). Contrary to earlier reports, about 30% of total neurons within ganglion cell layer (GCL) expressed GABA immunoreactivity. Double-labeling studies indicated that about half of the GABA-containing perikarya in the GCL were retinal ganglion cells (RGCs). In the diencephalon, intensely labeled GABA-immunoreactive neurons and nerve fibers were observed within the neuropil of Bellonci (nB) and corpus geniculatum (CG), while only immunoreactive puncta were found in the rostral visual nucleus (RVN). In the pretectal region, the posterior thalamic nucleus (nPT) contained the most intensely labeled GABA immunoreactive perikarya and nerve fibers in the entire brain. Lightly immunoreactive perikarya were also found in the large-celled nucleus lentiformis mesencephali (nLM), as well as in the pretectal gray, which contains neurons postsynaptic to the retinal terminal zones within nLM. In the optic tectum (OT), both immunoreactive perikarya and fibers were found within superficial layers 8 and 9, whereas only densely packed immunoreactive perikarya were evident in the deep tectal layers (i.e. 2, 4, 6). The nucleus of the basal optic root (nBOR) contained a small number of lightly labeled GABA-immunoreactive perikarya, mostly located in the dorsal half of the nucleus. A large number of perikarya within the nucleus isthmi (NI) were also immunostained.

In the kitten, as little as a week of monocular lid suture during early life causes a remarkable remodeling of the geniculocortical projections serving the deprived eye (Antonini & Stryker, 1993a, 1996). While the physiological effects of monocular deprivation have been shown to be due to competitive interactions between the projections serving the two eyes, it is not known whether these morphological changes are due to competitive interactions or to sensory disuse. We addressed this question by analyzing the morphology of geniculocortical arbors in kittens deprived of patterned vision by binocular lid suture for 1 week or 2 weeks ending at 6 weeks of age. Such deprivation would be expected to affect the afferents serving the two eyes equally, giving neither eye a competitive advantage. The arbors were anterogradely filled with Phaseolus lectin iontophoretically injected into lamina A of the lateral geniculate nucleus. The lectin was visualized immunohistochemically, and single geniculocortical arbors were serially reconstructed in three dimensions. Arbors reconstructed in binocularly deprived animals were compared with arbors serving the deprived and nondeprived eye in animals monocularly deprived by lid suture of one eye for a week and with arbors obtained in age-matched normal controls. Geniculocortical arbors in binocularly deprived animals did not suffer the drastic remodeling of the deprived arbors in monocularly deprived animals. Indeed, arbors in binocularly deprived animals were indistinguishable from arbors in normal kittens or nondeprived arbors in short-term monocularly deprived animals. These results support the notion that competitive mechanisms rather than sensory disuse are responsible for gross morphological remodeling of geniculocortical arbors.

The influence of neurons projecting from the pretectal nuclear complex to the ipsilateral dorsal lateral geniculate nucleus (LGNd) was investigated in awake cats. Responses from relay cells in the A-laminae of the LGNd were extracellularly recorded and analyzed during saccadic eye movements and visual stimulation in association with reversible inactivation of the ipsilateral pretectum with the GABA agonist, muscimol. Pretectal inactivation (PTI) resulted in spontaneous nystagmic eye movements in the dark with slow phases directed away from the injected side. In the control situation, all Y-cells and about two thirds of X-cells were excited during saccades or saccade-like visual stimulation but one third of X-cells were inhibited. During PTI all recorded X-cells were inhibited, either during saccades or saccade-like visual stimulation. The PTI-associated inhibition was stronger than in inhibited X-cells in control experiments only during saccades but not during stimulation with a moving pattern while the eyes were stationary. In Y-cells a reduction in the response peak width at half-height was seen during PTI, again only during saccades but not during stimulation with a moving pattern. These results indicate that during saccades the pretecto-geniculate pathway has a stronger influence on X LGNd relay cells than on Y-cells. The findings are discussed in terms of saccadic suppression and postsaccadic facilitation.

Color vision is spectrally opponent, suggesting that spectrally opponent neurons, such as the horizontal cells in fish and turtle retinae, play a prominent role in color discrimination. In the accompanying paper (Kraaij et al., 1998), it was shown that the output signal of the horizontal cell system to the cones is not at all spectrally opponent. Therefore, a role for the spectrally opponent horizontal cells in color discrimination seems unlikely. In this paper, we propose that the horizontal cells play a prominent role in color constancy and simultaneous color contrast instead of in color discrimination. We have formulated a model of the cone/horizontal cell network based on measurements of the action spectra of the cones and of the feedback signal of the horizontal cell system to the various cone types. The key feature of the model is (1) that feedback is spectrally and spatially very broad and (2) that the gain of the cone synapse strongly depends on the feedback strength. This makes the synaptic gain of the cones strongly dependent on the spectral composition of the surround. Our model, which incorporates many physiological details of the outer retina, displays a behavior that can be interpreted as color constancy and simultaneous color contrast. We propose that the horizontal cell network modulates the cone synaptic gains such that the ratios of the cone outputs become almost invariant with the spectral composition of the global illumination. Therefore, color constancy appears to be coded in the retina.

Horizontal cells and cone photoreceptors in the vertebrate retina are interconnected by a complex network of synapses leading to the generation of color-coded responses in chromaticity horizontal cells. A simple cascade model of excitatory feedforward and inhibitory feedback synapses had been suggested to underlie these observations. In this study, the photoresponses of cones and horizontal cells were recorded intracellularly from the turtle eyecup. Three different approaches were adopted in order to test quantitatively the cascade model. Comparing linearity functions between these neurons indicated multiple excitatory inputs to each type of horizontal cells. The depolarizing photoresponses of R/G C-type horizontal cells were considerably faster than those of L-type horizontal cells but slower than those recorded from L-cones. This observation disagrees with the basic assumption of the cascade model that assign the depolarizing photoresponses of R/G C-type horizontal cells to a negative feedback pathway from L-type horizontal cells onto M-cones. Finally, the action spectra of each of the three types of horizontal cells could not be solely accounted for by input from one spectral type of cones. Only by assuming excitatory and inhibitory inputs from all spectral types of cones, the action spectra of all types of horizontal cells could be reconstructed. These findings suggest that the negative feedback pathways from horizontal cells onto cones in the turtle retina cannot solely account for the chromatic properties of the horizontal cells and support a direct inhibitory inputs from cones to turtle horizontal cells.

We have examined the role of neurotrophins in promoting survival of mammalian rod bipolar cells (RBC) in culture. Retinas taken from 8- to 10-day-old Long-Evans rats were dissociated and cultured in media supplemented with either nerve growth factor (NGF), neurotrophin-3 (NT-3), brain-derived neurotrophic factor (BDNF), ciliary neurotrophic factor (CNTF), or basic fibroblast growth factor (FGF-2). Survival was measured by the number of cells that were immunoreactive for α-, β-, γ-PKC, a bipolar cell-specific marker. Compared to untreated cultures, CNTF had no effect on RBC survival, while NGF and NT-3 increased survival only slightly. BDNF, however, increased survival by approximately 300%. Similar results were obtained with FGF-2. Both nerve growth factor (NGF) and an antibody (anti-REX) which interferes with binding to the 75-kD low-affinity neurotrophin receptor (p75NTR) eliminated BDNF-promoted survival, but had no effect on FGF-2-mediated survival. Interestingly, p75NTR was expressed by retinal glia (Müller cells), but not by the bipolar cells themselves, providing for the possibility that BDNF might induce Müller cells to produce a secondary factor, perhaps FGF-2, which directly rescues RBCs. In support of this hypothesis, an antibody that neutralizes FGF-2 attenuated the trophic effects of BDNF, and dramatically reduced survival in cultures with no added growth factors, indicating that there may be an endogenous source of FGF-2 that promotes survival of RBCs in culture. We suggest that BDNF increases production or release of FGF-2 by binding to p75NTR on Müller cells.

Pyramidal neurons in superficial layers of cerebral cortex have extensive horizontal axons that provide a substrate for lateral interactions across cortical columns. These connections are believed to link functionally similar regions, as suggested by the observation that cytochrome-oxidase blobs in the monkey primary visual cortex (V1) are preferentially connected to blobs and interblobs to interblobs. To better understand the precise relationship between horizontal connections and blobs, we intracellularly labeled 20 layer 2/3 pyramidal neurons in tangential living brain slices from V1 of macaque monkeys. The locations of each cell body and the cell's synaptic boutons relative to blobs were quantitatively analyzed. We found evidence for two cell types located at characteristic distances from blob centers: (1) neurons lacking long-distance, clustered axons (somata 130–200 μm from blob centers) and (2) cells with clustered, long-distance axon collaterals (somata <130 μm or >200 μm from blob centers). For all cells, synaptic boutons close to the cell body were located at similar distances from blob centers as the cell body. The majority of boutons from cells lacking distal axon clusters were close to their cell bodies. Cells located more than 200 μm from blob centers were in interblobs and had long-distance clustered axon collaterals selectively targeting distant interblob regions. Cells located less than 130 μm from blob centers were found within both blobs and interblobs, but many were close to traditionally defined borders. The distant synaptic boutons from these cells were generally located relatively near to blob centers, but the neurons closest to blob centers had synaptic boutons closer to blob centers than those farther away. There was not a sharp transition that would suggest specificity for blobs and interblobs as discrete, binary entities. Instead they appear to be extremes along a continuum. These observations have important implications for the function of lateral interactions within V1.

Centrifugal modulation of visual responsiveness of tectal cells by the isthmo-optic nucleus (ION) through the retina was studied in homing pigeons. Visual activity evoked by computer-generated stimuli was reduced by an average of 59% in tectal cells whose receptive fields (RFs) either overlapped with, or were close to, those of isthmo-optic cells whose activity was blocked by the injection of lidocaine through micropipettes. Activity usually recovered to 87% of pre-drug controls in 8–17 min (average 12.3 min) after stopping lidocaine injections. Those tectal cells whose RFs were far from those of ION cells did not show clear-cut changes in their visual responsiveness to isthmo-optic lidocaine application. The spatial relationship between receptive fields of tectal and isthmo-optic cells, saline controls, as well as the specificity, reproducibility and reversibility of effects of ION-injected lidocaine on tectal activity, show that this chemical action is pharmacological, not toxicological. Neuronal circuitry underlying centrifugal modulation of tectal activity by isthmo-optic cells is discussed.

In this study, we examined functional contributions of major subdivisions of the lateral geniculate nucleus to the cat's lateral suprasylvian visual area (LS) in relation to the patchy horizontal distributions of association inputs. Multiple-unit activity driven via the contralateral eye was assessed during reversible blockade of the retinotopically corresponding part of layer A, the C layers as a group, or the medial interlaminar nucleus (MIN). Inactivating each of these targets reduced activity at some cortical sites, with inactivation of layer A having, on average, the largest effect. Activity was rarely abolished by inactivation of a single target, indicating that most LS sites receive multiple inputs. Dependence on layer A was strongly correlated with the horizontal distribution of association inputs from area 18. Closely spaced injections of anatomical tracers into extensive regions of area 18 resulted in patches of terminal label in lateral suprasylvian cortex. Activity inside the patches was relatively dependent on layer A, whereas that outside the patches was not. Dependence on the MIN and layer A were negatively correlated, suggesting that inputs dominated by the MIN and layer A were concentrated in independent sets of patches. These results indicate that the anatomically observed patchy projections reflect the functional consequences of geniculate lamination. The A layers are high-acuity relays, whereas the MIN is probably a specialization for dim-light vision (Lee et al., 1984; Lee et al., 1992). We propose that the partial overlap of inputs dominated by the A layers and the MIN allows dynamic shifts in their relative contributions to LS responses, optimizing the balance of high-acuity and high-sensitivity channels over a wide range of illumination conditions.

The spectral sensitivity of cones in isolated goldfish retina was determined with whole-cell recording techniques. Three spectral classes of cones were found with maximal sensitivities around 620 nm, 540 nm, and 460 nm. UV-cones were not found because our stimulator did not allow effective stimulation in the UV range. The spectral sensitivity of the cones closely matched the cone photopigment absorption spectra at the long wavelength side of the spectrum, but deviated significantly at shorter wavelengths. Surround stimulation induced an inward current in cones due to feedback from horizontal cells. The spectral sensitivity of this feedback signal was determined in all three cone classes and found to be broader than the spectral sensitivity of the cones recorded from, and to be spectrally nonopponent. These data are consistent with a connectivity scheme between cones and horizontal cells in which the three horizontal cell systems feed back to all cone systems and in which all horizontal cell systems receive input from more than one cone system.

Many rodents do not see well. For a target to be resolved by a rat or a mouse, it must subtend a visual angle of a degree or more. It is commonly assumed that this poor spatial resolving capacity is due to neural rather than optical limitations, but the quality of the retinal image has not been well characterized in these animals. We have modified a double-pass apparatus, initially designed for the human eye, so it could be used with rodents to measure the modulation transfer function (MTF) of the eye's optics. That is, the double-pass retinal image of a monochromatic (λ = 632.8 nm) point source was digitized with a CCD camera. From these double-pass measurements, the single-pass MTF was computed under a variety of conditions of focus and with different pupil sizes. Even with the eye in best focus, the image quality in both rats and mice is exceedingly poor. With a 1-mm pupil, for example, the MTF in the rat had an upper limit of about 2.5 cycles/deg, rather than the 28 cycles/deg one would obtain if the eye were a diffraction-limited system. These images are about 10 times worse than the comparable retinal images in the human eye. Using our measurements of the optics and the published behavioral and electrophysiological contrast sensitivity functions (CSFs) of rats, we have calculated the CSF that the rat would have if it had perfect rather than poor optics. We find, interestingly, that diffraction-limited optics would produce only slight improvement overall. That is, in spite of retinal images which are of very low quality, the upper limit of visual resolution in rodents is neurally determined. Rats and mice seem to have eyes in which the optics and retina/brain are well matched.

We report VEP studies which delineate interactions between chromatic and luminance contrast signals. We examined responses to sinusoidal luminance gratings undergoing 4-Hz square-wave contrast reversal, upon which standing gratings with various admixtures of luminance and chromatic contrast were alternately superimposed and withdrawn. The presence of the standing grating induced a VEP component at the fundamental frequency of the contrast-reversal grating. This VEP component appeared without any appreciable lag, and did not vary in amplitude over the 4 s during which the standing grating was present. The observed fundamental response differed from the fundamental component that would be expected from the known interaction between the luminance component of the standing grating with the modulated grating (Bodis-Wollner et al., 1972; Bobak et al., 1988), in three ways: (1) The fundamental response was not nulled for standing gratings that were isoluminant or near-isoluminant. (2) The chromatic dependence of the fundamental response implied an S-cone input to the interaction. (3) No single mechanism (driven by a linear combination of cone signals) could account quantitatively for the size of this response, particularly when the standing grating strongly modulated two cones in phase.

Microspectrophotometry studies show that zebrafish (Danio rerio) possess four cone photopigments. The purpose of this study was to determine the cone contributions to the zebrafish photopic increment threshold spectral-sensitivity function. Electroretinogram (ERG) b-wave responses to monochromatic lights presented on a broadband or chromatic background were obtained. It was found that under the broadband background condition, the zebrafish spectral-sensitivity function showed several peaks that were narrower in sensitivity compared to the cone spectra. The spectral-sensitivity function was modeled with L − M and M − S opponent interactions and nonopponent S- and U-cone mechanisms. Using chromatic adaptation designed to suppress the contribution of the S-cones, a strong U-cone contribution to the spectral-sensitivity function was revealed, and the contributions of the S-cones to the M − S mechanism were reduced. These results show that the b-wave component of the ERG receives input from all four cone types and appears to reflect color opponent mechanisms. Thus, zebrafish may possess the fundamental properties necessary for color vision.

Mammalian rods respond to single photons with a hyperpolarization of about 1 mV which is accompanied by continuous noise. Since the mammalian rod bipolar cell collects signals from 20–100 rods, the noise from the converging rods would overwhelm the single-photon signal from one rod at scotopic intensities (starlight) if the bipolar cell summed signals linearly (Baylor et al., 1984). However, it is known that at scotopic intensities the retina preserves single-photon responses (Barlow et al., 1971; Mastronarde, 1983). To explore noise summation in the rod bipolar pathway, we simulated an array of rods synaptically connected to a rod bipolar cell using a compartmental model. The performance of the circuit was evaluated with a discriminator measuring errors in photon detection as false positives and false negatives, which were compared to physiologically and psychophysically measured error rates. When only one rod was connected to the rod bipolar, a Poisson rate of 80 vesicles/s was necessary for reliable transmission of the single-photon signal. When 25 rods converged through a linear synapse the noise caused an unacceptably high false positive rate, even when either dark continuous noise or synaptic noise where completely removed. We propose that a threshold nonlinearity is provided by the mGluR6 receptor in the rod bipolar dendrite (Shiells & Falk, 1994) to yield a synapse with a noise removing mechanism. With the threshold nonlinearity the synapse removed most of the noise. These results suggest that a threshold provided by the mGluR6 receptor in the rod bipolar cell is necessary for proper functioning of the retina at scotopic intensities and that the metabotropic domains in the rod bipolar are distinct. Such a nonlinear threshold could also reduce synaptic noise for cortical circuits in which sparse signals converge.

The goal of this study was to determine the effects of inactivating layer A or the C layers of the cat lateral geniculate nucleus on the supragranular layers of area 18, including cells antidromically activated from the lateral suprasylvian visual area (LS). Isolated cells were visually driven via the contralateral eye while the retinotopically corresponding regions of layer A or, in some cases, the C layers were reversibly inactivated with injections of cobaltous chloride. Simple cells were frequently encountered and were on average more dependent upon layer A than were complex cells, a result qualitatively similar to that found previously in area 17 (Malpeli, 1983; Malpeli et al., 1986). However, the influence of the C layers on area 18 was much more apparent than for area 17. In area 18, as in area 17, the dependence of simple cells on particular geniculate layers appears to follow the terminal patterns of the major direct geniculate inputs. Those simple cells most dependent on layer A were located in lower layer 3. Simple cells in upper layer 3, like complex cells, showed little dependence on layer A, but were strongly dependent upon the C layers. All cells antidromically activated from LS were simple cells with rapidly conducting axons. They had, on average, the same moderately strong dependence on layer A as the patches of LS receiving area 18 input (Lee et al., 1997), supporting the conclusion that the influence of layer A in these patches is largely transmitted via association inputs from area 18. These results demonstrate that simple cells play a major role in association pathways.

Birds have a well-developed centrifugal pathway from the isthmo-optic nucleus to the retina. In the present study, receptive-field properties of the isthmo-optic (IO) neurons and suppressive effect of remote stimuli far beyond the “classical” receptive fields were examined electrophysiologically in the Japanese quail. Receptive fields emitting more than half the maximal ON–OFF response to a small spot of light measured 4.3 ± 1.9 deg (n = 37) in diameter. A stationary spot of light was presented at a remote point (35–76 deg) away from the receptive-field center, besides a stimulus for the receptive-field center, with various onset time difference. The peripheral spot, when turned on 50–100 ms before the center spot, maximally suppressed the ON and OFF responses to the center spot. In most of the IO neurons examined (13 out of 15 units), even small remote stimuli (2–5 deg in diameter) significantly suppressed the ON–OFF responses to the center stimulus. The suppressive effects of remote stimuli were seen to be extended to, at least, two thirds of the entire visual field of a single eye. Such wide suppressive fields indicate that the IO neurons may compete with each other for activity in a very long-range scale. Simulation using a simple static model based on three basic principles of processing of the centrifugal visual system— topographical input, long-range competition, and local gain enhancement—suggests that the system could function as an autonomous mechanism for attentional object selection.

The impulse discharge of single ganglion cells was recorded extracellularly in superfused eyecup preparations of the tiger salamander (Ambystoma tigrinum). Contrast flashes (500 ms) were applied at the center of the receptive field while the retina was light adapted to a background field of 20 cd/m2. The incidence of cell types in a sample of 387 cells was: ON cells (4%), OFF cells (28%), and ON/OFF cells (68%). Quantitative contrast/response measurements were obtained for 83 cells. On the basis of C50, the contrast necessary to evoke a half-maximal response, ON/OFF cells fell into 3 groups: (1) Positive Dominant (26%), (2) Balanced (23%), and (3) Negative Dominant (51%). Positive Dominant cells tended to be relatively contrast insensitive. On the other hand, many Negative Dominant cells showed remarkably low C50 values and very steep contrast/response curves. Contrast gain to negative contrast averaged 8.5 impulses/s/% contrast, some four times greater than that evoked by positive contrast. In most ON/OFF cells, the latency of the first spike evoked by a negative contrast step was much shorter (40–100 ms) than that evoked by a positive contrast step of equal contrast. OFF cells typically showed higher C50 values, larger dynamic ranges, and longer latencies than those of Negative Dominant ON/OFF cells. Thus, different pathways or mechanism apparently mediate the off responses of OFF and ON/OFF cells. In sum, the light-adapted retina of the tiger salamander is strongly biased in favor of negative contrast, as shown by the remarkably high contrast sensitivity and faster response of Negative Dominant cells, the remarkably low incidence of ON cells, and the insensitivity of Positive Dominant cells. Some possible underlying influences of bipolar and amacrine cells are discussed.

Changes in intracellular calcium are involved in phototransduction processes in both vertebrate and invertebrate photoreceptors. During this phototransduction process in the Limulus ventral eye, there is a biochemical change in the protein phosphatase, calcineurin, such that it becomes capable of activation by calcium and calmodulin. Here we show that the calcium/calmodulin-dependent calcineurin-like activity in light-adapted ventral eye was completely inhibited by the CaN autoinhibitory peptide, CaN A457–482 and the Merck analog of the membrane-permeable, immunosuppressant drug, FK 506, L-683, 590, but not an inactive analogue, L-685, 818. Whole-cell, voltage-clamp recordings of spontaneous quantal bump activity present in dark-adapted photoreceptors injected with either CaN A457–482 (500 μM) or superfused with L-683, 590 (20 μM) or L-685, 818 revealed that both CaN A457–482 and L-683, 590, but not L-685, 818, caused rapid decreases in quantal bump amplitude, rise time and fall time, resulting in smaller, sharper bumps. This was correlated with enhanced phosphorylation of arrestin in light-adapted ventral eye photoreceptors exposed to L-683, 590 or less reliably okadaic acid. Both CaN A457–482 and L-683, 590 markedly affected the light-stimulated inward currents recorded from light-adapted ventral photoreceptors, causing a “terracing” of the inward current, and an intensity-dependent delay in the time required to reach peak amplitude. Consequently, inhibition of calcineurin markedly affects two major rhodopsin-dependent electrophysiological processes, and implicates CaN as an integral component in the phototransduction cascade.

We studied how intrinsic membrane properties affect the gain and temporal pattern of response in bipolar cells dissociated from retinae of tiger salamanders. Currents specified by a pseudorandom binary sequence, an m-sequence, superimposed on various means, were injected into the cells. From the resultant membrane voltage response for each mean current, impulse responses were estimated. From each impulse response, transfer function, gain, and time constant were calculated. The bipolar cells acted as quasilinear adaptive filters whose gain and response speed are determined by the mean input current. Near resting potential, gain, and time constant were maximum. Dynamics were slow and low-pass, characterized by an approximately exponential impulse response. With depolarization, gains were reduced sharply, responses were much faster, and dynamics became band-pass, as indicated by an undershoot in the impulse response. For any given mean current, the shape of the impulse response did not depend on the amplitude of the m-sequence currents. Thus, bipolar cells behaved in a quasilinear fashion. The adaptive behavior was eliminated by blocking a potassium current, which implicates the role of a voltage-gated potassium conductance. Computer simulations on a model neuron including a delayed-rectifier reconstructed the observed behavior, and provided insight into other, less readily observable, parameters. Thus, bipolar cells, even when isolated, possess mechanisms which regulate, with unsuspected elaborateness, the sensitivities and dynamics of their responsiveness. Implications for adaptation and neuronal processing are discussed.